Sociometer perspectives.

From the sociometer perspective, self-esteem is dependent on social embeddedness. More specifically, self-esteem is assumed to monitor an individual’s social inclusion as a prerequisite for survival and reproduction [22]. Low self-esteem is regarded as an indicator of rejection, whereas high self-esteem is regarded as the result of social success in terms of approving social reactions [21, 23]. Romantic relationships in Western cultures are one of the rare relationships that cannot only increase survival by (possibly) being of supportive nature, but also especially serve the goal to establish a family. Given that one basic assumption of the sociometer perspective is that self-esteem serves the ultimate goal of reproduction, romantic relationships should be one of the best contexts to study this perspective. In line with this, Knee and colleagues [19] argued that successfully navigating romantic relationships plays a leading role for an individual’s self-esteem. In one of their studies, they examined self-esteem fluctuations on days after perceived positive and negative relationship events. Across all individuals, self-esteem was lower after perceived rejection situations than after positive situations [19]. Similarly, Murray, Griffin, Rose, and Bellavia [20] reported that perceiving or fearing disapproval and rejection in a relationship predicted decreases in self-esteem. This effect especially manifested in low self-esteem individuals, who were indeed chronically less positively regarded by their partners than high self-esteem individuals [20]. These studies support the assumption that changes in self-esteem may be triggered by social experiences of disapproval and rejection within romantic relationships.

In the following, we will introduce the fictitious couple Jenny and James to illustrate the complex interplay between self-esteem and conflict in a dyadic context. The sociometer perspective can help explain several scenarios regarding their interactions. First, if Jenny and James perceive multiple conflict situations with each other, this could lead both of them to feel rejected, leading to decreases in both their self-esteem. Second, perceiving his partner Jenny being withdrawing or unwilling to compromise during a fight might lead James to think that she rejects him, which decreases his self-esteem. And third, if James constantly perceives and reports that Jenny shows dysfunctional conflict behavior, this might tell something about James’ devaluation of Jenny. Put differently, James might not appreciate Jenny’s behavior, which in turn leads her to feel disapproved, and eventually decreases her self-esteem.

Self-broadcasting perspectives.

The self-broadcasting perspective posits that individuals behave in accordance with their self-evaluations, thus evoking congruent reactions in their social encounters [24]. Low self-esteem individuals are assumed to chronically strive for approval [20]. At the same time, they seem to underestimate their partner’s regard as a consequence of a heightened attention towards hostile partner signals in the sense of chronically high rejection sensitivity [25]. Chronically high rejection sensitivity is defined as “the disposition to anxiously expect, readily perceive, and overreact to rejection” [26, p. 545]. In line with this, individuals with low self-esteem and high rejection sensitivity have been found to underestimate their partner’s satisfaction [20, 27], and to be more sensitive to actual rejection behavior by the partner [20, 28]. Additionally, low self-esteem individuals seem to distance themselves from their partner (i.e., avoidance) to minimize the risk of rejection, whereas high self-esteem individuals seem to seek closeness, even after conflict situations [14]. Another consequence from the dilemma in which low self-esteem individuals find themselves is that their increased fear of rejection might cause them to engage in dysfunctional behavior during conflicts [15, 20]. Possibly as a consequence, partners of women afraid of being rejected are indeed more rejecting [26].

In accordance with self-broadcasting perspectives, self-esteem has been found to affect self-perceived popularity among peers [29, 30]. In romantic relationships, small positive effects of self-esteem on relationship satisfaction changes have been reported [25, 31]. In line with this, Murray and colleagues [20] found that one individual’s sensitivity to rejection spilled over to their partner, leading their partner’s relationship satisfaction to decrease over the year. As fear of rejection is associated to self-esteem [27], this finding might indicate a spill-over effect of self-esteem on the partner’s satisfaction.

From self-broadcasting perspectives, several scenarios are conceivable regarding the interplay between self-esteem and conflict in our fictitious couple. First, James having low self-esteem might perceive more negative conflict behavior in his partner Jenny than a high self-esteem individual would, independently of Jenny’s actual behavior. Second, James might distance himself from Jenny. This could eventually lead Jenny (the partner of a low self-esteem individual) to perceive rejection behavior in James, mirroring his attempt for self-protection. Third, James’ fear of rejection, triggered by his low self-esteem, might lead him to act less constructively during conflicts, evoking rejection behavior in Jenny. That is, Jenny would eventually really come to be disapproving of James, possibly as a cost of his unconstructive conflict behavior.

Predictions on sociometer perspectives.

Based on the sociometer perspective, we argue that perceiving withdrawal or unwillingness to compromise in a partnership should predict subsequent decreases in individuals’ self-esteem. Perceiving the partner lacking constructive conflict behavior (here termed as unconstructive behavior) and withdrawal during a fight seem to be obvious candidates to evoke perceived disapproval and rejection in individuals. We assumed that higher perceived conflict frequency in general might have the same effect. Thus, individuals who perceive more overall conflict and who perceive their partner showing more dysfunctional behavior during a fight should show stronger decreases in self-esteem (actor effect) than individuals who perceive less (dysfunctional) conflict. Moreover, we expected this effect to spill over to the partner, also leading to a subsequent self-esteem decrease in those individuals showing uncompromising or withdrawing conflict behavior (partner effect).

Level-change effects supporting sociometer hypotheses. A major strength of the current study is that level-change effects are not mere reflections of pre-post associations, but are specifically able to indicate whether the level of one construct has an influence on the change of another construct after controlling for the initial level of the second construct. If Jenny constantly perceives conflict in her relationship or James is little assuring during a fight (unconstructive behavior, withdrawal), this might increase feelings of disapproval and insecurity in Jenny in the long run. While we acknowledge that low self-esteem individuals might be particularly sensitive to such processes, we assume that constantly high levels of conflict may have the potential to erode self-esteem in most couples (actor level-change effect) [19].

Low self-esteem individuals have been shown to be less positively regarded by their partners [20]. Consequently, we expect that low partner regard would lead to subsequent decreases in self-esteem (partner level-change effect). That is, if James constantly perceives and reports that Jenny shows dysfunctional conflict behavior, this might be a hint on James’ potential devaluation of her, and Jenny’s self-esteem might decrease in reaction to this, too–opposite to the processes that have been found for high regard [21].

Predictions on self-broadcasting perspectives.

In line with the self-broadcasting perspective, we argue that low self-esteem predicts a stronger perception of uncompromising or rejecting partner behaviors (actor effect). Moreover, we expected low self-esteem individuals to report more conflict (actor effect), to behave less constructively during a fight, and to show more withdrawal. From a dyadic perspective, we expected partners of low self-esteem individuals to report more conflict in general, and to report more unconstructive and withdrawal tendencies in their partner (partner effect).

Level-change effects supporting self-broadcasting hypotheses. In accordance with self-broadcasting perspectives, we expected self-esteem to be negatively associated with subsequent changes in perceived conflict frequency and dysfunctional conflict behavior. Put differently, low self-esteem individuals should report higher increases in their perception of arguing frequently (high conflict frequency) and/or in their perception of little assurance by their partner during a fight (unconstructive behavior, withdrawal). For example, James’ low self-esteem might lead him to increasingly perceive conflict with Jenny over time, and to increasingly perceive her behavior during a fight as dysfunctional (actor level-change effect) [20].

Previous work has shown that dysfunctional behavior arises situationally due to feelings of rejection, with low self-esteem individuals being especially prone to experience these feelings [33]. We assumed that this process might lead the partner of a low self-esteem individual to report more negative conflict behavior in their partner in the long run (partner level-change effect). Put differently, James’ low self-esteem might subsequently increase Jenny’s perception of their conflict frequency or her perception of his dysfunctional conflict behavior (i.e., unconstructive behavior, withdrawal) during a fight.

Predictions on change-change effects.

Change-change effects denote longitudinal transactions between two constructs after controlling for their initial levels [34]. For example, change in perceived conflict frequency might predict subsequent change in self-esteem. While each change-change effect by itself might be subsumed under one certain perspective (e.g., the change-change effect of conflict frequency on self-esteem would belong to the sociometer perspective), we argue that it is most appropriate to interpret change-change effects from both directions at the same time. This way, they can inform us about bidirectional influences between multiple measurement occasions, making the interpretation of longitudinal dynamic transactions between two constructs possible. Support for the expectation of longitudinal transactions comes from accumulating negative circles of disapproving self-evaluations and relationship experiences [20].

With regard to change-change effects, we expected that an increase (decrease) in perceived conflict behavior might lead to a later decrease (increase) in self-esteem and vice versa (change-change effects) [20, 35]. That is, increases in James’ perceived conflict frequency and/or perceived partner dysfunctional conflict behavior across one year might lead to decreases in his self-esteem during the subsequent year, while decreases in James’ self-esteem between T1–T2 should predict increases in his perceived conflict frequency and/or perceived partner dysfunctional conflict behavior between T2–T3 (compared to an average person). Based on assumptions of mutual dynamics between partners [4, 5], we predicted these effects both within individuals (actor change-change effects) and between partners (partner change-change effects). Please note, however, that floor and ceiling effects are able to distort change-change effects.

Self-esteem.

Self-esteem was assessed via a computer-assisted self-administered interview at all measurement occasions (T1–T5). Three items from the Rosenberg Self-Esteem Scale [39] (“Sometimes I believe that I’m worthless (item recoded),” “I like myself just the way I am,” and “All in all, I am pleased with myself”) were answered on a five-point Likert scale ranging from 1 (not at all) to 5 (absolutely). One item was reversed before the analysis. Coefficient ω and its 95% confidence intervals for females’ self-esteem were .78 [.75, .81], .78 [.75, .81], .78 [0.76, 0.81], .74 [.70, .77], and .79 [.76, .81] at T1–T5, respectively. Coefficient ω estimates for males’ self-esteem were .74 [.71, .77], .74 [.71, .77], .77 [.74, .79], .71 [.66, .75], and .80 [.77, .82] at T1–T5, respectively.

Perceived conflict frequency in romantic relationships.

Relationship conflict was measured with multiple indicators. First, we used two items of an adapted Conflict Scale from the Network of Relationships Inventory [40] to measure perceived conflict frequency within the romantic relationship from both partners’ perspectives from 1 (never) to 5 (always) at all occasions. The English item translations would be, “How often do you and [name of current partner] disagree and quarrel?” and “How often are you and [name of current partner] annoyed or angry with each other?” Coefficient ω and its 95% confidence intervals for females’ perceived conflict frequency were .78 [.74, .81], .81 [.78, .83], .79 [.76, .81], .82 [.78, .85], and .81 [.78, .84] at T1–T5, respectively. Coefficient ω estimates of males’ perceived conflict frequency were .76 [.72, .79], .79 [.75, .82], .78 [.75, .81], .80 [.76, .83], and .79 [.76, .81] at T1–T5, respectively.

Perceived conflict styles in romantic relationships.

In contrast to perceived conflict frequency, we assessed perceived conflict styles of the romantic partner within conflict situations that we hypothesized to be associated to rejection perception. More specifically, we assessed individuals’ perceptions of their partner’s behavior within conflict situations via unconstructive and withdrawal behavior. The measures were introduced by the following sentences as documented in the Pairfam Scales Manual (the latest version can be retrieved from https://www.pairfam.de/dokumentation/scales-manual/), “What happens when you have a disagreement with [name of partner]? Please indicate how often each of you engaged in the following behaviors. When answering, please refer to the past six months,” followed by the introductory question to individuals’ perceptions of their partner’s unconstructive and withdrawal conflict styles, which was, “How often did your partner engage in any of these behaviors?”.

We recoded perceived constructive behavior from an adapted Constructive Behavior Scale [41] into perceived unconstructive behavior to address the negativity and unwillingness to compromise in relationship conflicts. The English item translations of perceived constructive (unconstructive) behavior in one’s partner as documented in the Scales Manual would be, “Listen to and ask questions of you in order to understand better (item recoded)” and “Endeavor to clarify his or her own position to you (item recoded).” Perceived withdrawal behavior of the partner was assessed via the Withdrawal Scale of the Conflict Resolution Inventory [42]. The English item translations of perceived withdrawal in one’s partner as documented in the Scales Manual would be, “Remain silent” and “Refuse to talk about the subject.” Both scales were assessed on 5-point scales from 1 (almost never or never) to 5 (very frequently).

Coefficient ω estimates of females’ perceptions of their male partners’ unconstructive behavior were .72 [.67, .75], .68 [.63, .71], .76 [.72, .79], .72 [.69, .76], and .75 [.71, .78] at T1–T5, respectively. Coefficient ω and its 95% confidence intervals of males’ perceptions of their female partners’ unconstructive behavior were .66 [.61, .70] at T1, .68 [.63, .72] at T2, T3, and T4, and .70 [.65, .74] at T5. Coefficient ω estimates of females’ perceptions of their male partners’ withdrawal were .56 [.51, .61], .61 [.56, .66], .63 [.58, .67], .68 [.63, .72], and .66 [.62, .70]. Coefficient ω and its 95% confidence intervals of males’ perceptions of their female partners’ withdrawal were .61 [.56, .65], .63 [.58, .68], .64 [.59, .68], .68 [.63, .72], and .65 [.59, .69] at T1–T5, respectively.

Latent change modeling.

To get the most reliable measures, we modeled each anchor’s and partner’s self-esteem and perceived relationship conflict as latent variables, that is, controlling for measurement error [43]. Prior to all analyses, we standardized each variable within gender (except for relationship duration, which we standardized across the total sample). We first established univariate models to individually capture self-esteem and perceived conflict change patterns of females and males in continuing relationships. To assess latent changes, variable levels were decomposed into the initial variable level (i.e., level at T1) and the latent variable changes (i.e., the differences between latent trait levels at two neighboring measurement occasions) [44, 45].

We established strong measurement invariance by equating the latent variable structure, factor loadings, and intercepts for each construct across time in order to reasonably interpret the meaning of their mean-level changes (see S2 Table for descriptive model fits regarding metric and strong measurement invariance) [46]. Moreover, we equated change parameters across time to get the most parsimonious model. Residual variances of each item within a construct were allowed to correlate across time points to account for shared variance that was not accounted for by the latent variables [47].

Dyadic bivariate transactions.

In a second step, we integrated the univariate models into the dyadic bivariate design. That is, we combined the latent change models for females’ and males’ self-esteem with the respective latent change models for aspects of females’ and males’ perceived conflict. These models allowed us to observe several associations between the constructs under investigation, including initial correlations at T1 within individuals and across partners. Correlated changes, that is, correlations between residuals of change factors within concurrent time intervals, denote the degree to which constructs develop into similar or divergent directions, again both within and across partners.

Level-change effects between two constructs denote how levels in one construct at T1 predict subsequent changes in another construct beyond initial levels. More specifically, while classical crossed-lagged models examine pre-post associations, our approach enabled us to examine the effect of a certain characteristic level on another characteristic’s residual change after ruling out the second characteristic’s initial level. Actor and partner level-change effects from perceived relationship conflict on self-esteem made it possible to test the sociometer hypothesis, while effects of self-esteem on perceived relationship conflict pertain to the self-broadcasting perspective. Last, change-change effects denote how prior changes in one construct predict subsequent changes in the other [34]. We included age and relationship duration as covariates of all effects in our analyses. To yield the most parsimonious model, we equated paths for women and men [6, 48].

Data preparation was done using R version 3.4.4 [49]. The dyadic bivariate latent change models were implemented and run in Mplus version 8 [50], where missing data was handled using FIML [51]. All syntax files can be found at the open science framework (https://osf.io/9b7mr/). Combining CFI, RMSEA, and SRMR leads to a good overall model fit criterion for large sample sizes. As a rule of thumb, CFI (comparative fit index) should be > .93, and both RMSEA (root mean square error of approximation) and SRMR (standardized root mean square residual) < .07, respectively [52]. As our model was very complex and we wanted to ensure appropriate per-parameter power, we considered effects with a p level of < .05 as statistically significant.

Level-change effects supporting sociometer perspectives.

Throughout the next paragraphs, we will present level-change effects with regard to sociometer perspectives. We will first describe effects of perceived conflict frequency, before we depict level-change effects of perceived partner conflict styles. We found a small within-person (actor) effect of perceived conflict frequency on subsequent self-esteem changes (β = –.09 [95% CI:–.13,–.05], p < .001) above and beyond the initial self-esteem level. In contrast, we did not find a between-person (partner) effect of perceived conflict frequency on changes in self-esteem. That is, while Jenny’s self-esteem would be unaffected by the conflict frequency reported by James, his perception of a high conflict frequency predicted a decrease in his self-esteem.

Significant actor effects of perceived partner unconstructive behavior (β = –.05 [95% CI:–.08;–.02], p = .001) and withdrawal (β = –.05 [95% CI:–.08; .02], p = .001) on subsequent self-esteem changes supported sociometer perspectives for conflict styles. That is, the more James perceived unconstructive behavior or withdrawal in Jenny, the steeper his self-esteem would decrease in comparison to the average person in our sample. Moreover, we found significant partner effects of perceived partner unconstructive behavior (β = –.04 [95% CI:–.06,–.01], p = .017) and perceived partner withdrawal (β = –.07 [95% CI:–.11,–.04], p < .001) on subsequent changes in the partner’s self-esteem controlled for their initial self-esteem level. These effects indicated the following: The more James perceived unconstructive or withdrawal tendencies in Jenny, the steeper Jenny’s self-esteem would consequently decrease. All effects were small, explaining only a little proportion of the overall variance in self-esteem changes.

Level-change effects supporting self-broadcasting perspectives.

Throughout this section, we will first present effects that support self-broadcasting perspectives, before summarizing non-significant effects. We found one small actor effect of self-esteem on subsequent changes in perceived partner unconstructive behavior (β = –.04 [95% CI:–.07;–.01], p = .009) after controlling for the initial level of perceived partner unconstructive behavior. That is, the lower James’ initial self-esteem was in comparison to the average person in our sample, the more his perception of Jenny’s unconstructive behavior subsequently increased (or, the less it decreased), above and beyond the control of her prior unconstructive behavior level. The other way round, the higher James’ initial self-esteem was, the more his perception of Jenny’s unconstructive behavior subsequently decreased. Please note that we found no partner effect of self-esteem on subsequent change in perceived partner unconstructive behavior. Put differently, while James’ self-esteem predicted changes in his perception of Jenny’s unconstructive behavior, Jenny’s self-esteem did not predict changes in James’ perception of her unconstructive behavior. Regarding conflict frequency and perceived partner withdrawal, we did not find any support for self-broadcasting perspectives in terms of level-change effects (see Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Dyadic latent bivariate level-change and change-change effects.

https://doi.org/10.1371/journal.pone.0248620.t003

Change-change effectss.

After having presented level-change effects with regard to sociometer and self-broadcasting perspectives, we will now present significant change-change effects. Please note that we did not subsume change-change effects under the respective perspectives. The reason is that in our opinion, bidirectional longitudinal transactions are best interpreted together. On the actor level, there were two significant change-change effects, both regarding conflict frequency: Increases in self-esteem predicted decreases in perceived conflict frequency (β = –.06 [95% CI:–.10,–.01], p = .009). That is, increases in Jenny’s self-esteem during T1–T2 predicted decreases in her reported conflict frequency during T2–T3. Moreover, increases in perceived conflict frequency predicted subsequent increases in self-esteem (β = .05 [95% CI: .01, .10], p = .022). That is, Jenny’s notion of an increasing conflict frequency during T1–T2 also predicted increases in her self-esteem during T2–T3. While the effects were very small, this pattern indicates longitudinal transactions between self-esteem and conflict frequency. Apart from that, we found no other change-change effects. That is, there were no hints of bidirectional longitudinal transactions between self-esteem and perceived unconstructive behavior or withdrawal, neither within individuals nor between partners.

Support for sociometer perspectives.

Investigating level-change effects allowed us to elaborate on the directions of influence. In general, both perceived conflict frequency and conflict styles showed small but significant influences on self-esteem changes, supporting the sociometer perspective. Our findings complement previous results indicating small positive effects of relationship satisfaction aspects on both partners’ self-esteem changes [6]. In sum, our findings suggest that sociometer perspectives might be valid for both perceived conflict frequency and conflict styles.

While conflict frequency had a within-person effect on self-esteem, conflict styles showed both within-person and between-person effects. Turning to the within-person findings, we found negative actor effects of perceived conflict frequency and conflict styles on subsequent self-esteem changes. That is, the more conflict individuals perceived in their relationship and the less functional their perceived partner behavior was during these conflicts, the steeper was their decrease in self-esteem thereafter, after controlling for previous self-esteem levels. Although these effects were small, they might indicate that individuals experience both conflict frequency in romantic relationships and dysfunctional partner behavior during conflicts as indicators of social rejection [19, 20, 22].

Moreover, we observed partner effects of perceived conflict styles on changes in self-esteem. Referring to our example of Jenny and James, James’ perception of Jenny showing high levels of unconstructive behavior and withdrawal during conflict situations affected stronger decreases in Jenny’s self-esteem. These findings are in line with Srivastava and Beer [23], who reported that insecure attachment behavior predicted lower self-evaluations in the same individuals. They also conceptually converge with previous findings on negative effects of perceived or feared disapproval and rejection on the level of self-esteem [21], although no spill-over effects between partners were tested in the previous study. One explanation of these findings might look like this: James perceives and disapproves of Jenny’s dysfunctional behavior during conflicts. Subsequently, he becomes less satisfied with their relationship [20]. Jenny, in turn, interprets James’ dissatisfaction as rejection, leading her self-esteem to decrease [21]. In general, it might be true that the perception of social rejection affects both persons involved, that is, the person showing rejection behavior and the person perceiving it. As our study design does not allow to uncover underlying mechanisms, this possibility should be addressed in future studies.

Originally, the sociometer perspective was developed for state self-esteem [22]. That is, self-esteem is expected to fluctuate in close timely proximity after social rejection situations. While a number of studies supported this notion, many of them based their conception of self-esteem on the Rosenberg Self-Esteem Scale [19, 20, 25], which resembles a trait approach to self-esteem [55]. As our study investigated effects of perceived conflict on self-esteem across one-year time intervals, it seems warranted to infer theoretical and methodological implications for a trait perspective on the sociometer concept. In fact, Leary et al. [22] acknowledge that experiencing real or imagined rejection over a long period of time may lower trait self-esteem, that is, the “average resting position of the ‘indicator needle’ on the person’s sociometer” [22, p. 527]. Our findings imply the following: Social influences in terms of conflict on trait self-esteem seem to represent more than only short-term phenomena. Rather, they can impact the fairly stable trait self-esteem over the course of several months. What remains unexplained are the underlying processes.

Whole trait theory [56] was originally developed for the Big Five traits, but might prove useful in this regard. This theory posits that social-cognitive mechanisms can explain differential behavior in situations (i.e., distributions in states). In short, despite situational deviations, changes in general social cognitions and thus changes in average general behavioral tendencies (i.e., state aggregates) can alter the average trait level [56]. In our example, James’ state self-esteem might decrease after he feels rejected by Jenny. Manifesting a general social cognition of Jenny as rejecting may increase the frequency of these state reactions (i.e., decreased state self-esteem) in multiple situations, eventually lowering James’ trait self-esteem.

Besides, our results indicate that negative conflict behavior was not a within-person experience only, but seemed to have consequences for both partners. Couple consultants, for example, should bear in mind that one partner’s communication and interaction behavior possibly affect both individuals’ self-esteem, which in turn may impact different areas of their lives–for example, their mental and physical health [57, 58]. Our study emphasizes the importance of assessing both individuals’ perceptions in dyadic interactions.

Support for self-broadcasting perspectives.

While there were no actor and partner level-change effects regarding conflict frequency and withdrawal, we found one small actor effect that supports self-broadcasting perspectives regarding unconstructive behavior. More specifically, a person’s self-esteem predicted changes in their perception of their partner’s unconstructive behavior, indicating that lower levels of self-esteem affected an increase in perceived unconstructive behavior. Murray and colleagues [21] had found the same effect on conflict–negativity and combined destructive conflict styles. So, while our study in part supports the authors’ [21] findings, our data also highlight the idea that self-esteem only affects specific dysfunctional conflict styles.

Interestingly, Murray and colleagues [20] found a spill-over effect of self-esteem related sensitivity to rejection on the partner’s relationship satisfaction, pointing towards between-person dynamics for this construct. While we did not find between-person effects supporting self-broadcasting perspectives, the within-person effect of low self-esteem on increases of perceived unconstructive partner behavior in our sample might indicate a similar process. More specifically, low self-esteem individuals might show behavior that their partners do not approve of, making them less satisfied over time [20]. Possibly triggered by this dissatisfaction, their partners behave less constructive during conflicts [compare 21]. This change in behavior would then be noticed by the low self-esteem individuals, manifesting in their increased perceived partner unconstructive behavior.

More generally, the pattern of results with regard to self-broadcasting perspectives was inconsistent. This tends to be in line with recent studies [6, 31, 48] reporting only small self-broadcasting effects for relationship satisfaction [59] and diverging effects for other indicators of relationship quality. Possibly, self-broadcasting effects might only hold in terms of certain relationship aspects, but not in terms of others. If this speculation holds true, it remains yet to be explained, however, which processes enforce these differences. To clear up this uncertainty, future research would need to test differences between relationship dynamics systematically with the help of multiple, comparably objective measures.

Change-change effects.

With regard to change-change effects, we found an interesting pattern for self-esteem and conflict frequency at the actor level. At first glance, the change-change effects occurred in both directions, implying support for sociometer as well as self-broadcasting perspectives. However, the direction of effects was less straightforward. In line with self-broadcasting perspectives, decreases in self-esteem predicted subsequent increases in perceived conflict frequency. In contrast, decreases in perceived conflict frequency were aligned to subsequent decreases in self-esteem (that is, increases in conflict frequency were aligned to subsequent increases in self-esteem). This finding is opposite to the predictions of the sociometer perspective. Possibly, increasing conflict frequency might not only have negative implications for individuals’ self-concepts. Jenny and James might find that they argue more frequently now than one year ago. However, they also observe a constructive culture of debate, making it possible to not only argue over different points of view, but to come up with solutions. Observing this dynamic in turn might let their self-esteem increase as a function of perceived self-efficacy for jointly solving problems. This interpretation is, however, mere speculation and needs to undergo further hypothesis testing.

To explain this finding from a statistical perspective, it might be helpful to bear in mind the other findings. The first information that sticks out were the high initial levels of self-esteem. Moreover, decreases in perceived conflict frequency were accompanied by increases in self-esteem during the same time span (correlated changes). Combining the negatively correlated changes between self-esteem and conflict frequency within the same time interval with high initial self-esteem levels, this might lead to inflated self-esteem measures at T2 for the individuals who reported decreases in conflict frequency across T1–T2. That is, while James’ perceived conflict frequency, for example, decreased, his self-esteem increased across T1–T2 up until the highest possible value. And although it would be plausible to assume that his self-esteem would increase even more across T2–T3, this is not possible. What follows is that his self-esteem can only decrease or not change after T2, leading to decreases in self-esteem between T2–T3 after decreases in conflict frequency between T1–T2. Assuming a highly probable regression to the mean, it is in fact very likely that his self-esteem decreases in this scenario, leading to the confusing pattern from above. In contrast, while James’ self-esteem increases between T1–T2, this perceived conflict frequency decreases during this time span, possibly and in line with the average mean levels in conflict frequency. In this case, it is possible that his perceived conflict frequency decreases further between T2–T3, leading to support of the self-broadcasting perspective. So, while these findings seem to be exciting, they require further investigation to rule out possible statistical artefacts due to ceiling effects of our sample’s high average self-esteem levels. In total, they still confirm the notion that romantic relationships are a meaningful context for self-esteem development, which is sensitive to perceived relationship conflict.