A socioemotional network perspective on momentary experiences of family conflict in young adults

A family systems and network perspective on conflict and emotions

One potentially fruitful way to quantify the importance of family conflict for youth psychopathology is to focus on the interplay between the emotional lives of family members and everyday family conflict experiences (Larson & Almeida, 1999). Emotions can be functional, alerting us to important changes in our environments and facilitating our responses to those changes (Izard, 2009). Yet, emotional intensity and its duration must often be modulated to enable optimal engagement in situations (Cole et al., 2004). Emotion regulation is adaptive when emotions exhibit flexibility, persisting until goals are achieved but rising and falling when appropriate to accommodate changing environmental demands (Thompson, 1994). In all families, conflict may arise and negatively valenced emotions may be triggered during interchanges between family members. Here, we define negatively valenced emotions as short-lived affective responses that are experienced as unpleasant, with valence expressed on a continuum from pleasant to unpleasant (Barrett & Russell, 1999). In this study, we specifically focus on three discrete emotions: anger, anxiety, and sadness. In well-functioning families, conflict can be repaired to avoid long-duration conflict and consequences (Leger et al., 2018). In other families, spiraling patterns of interactions where one reaction triggers another's reaction may emerge, leading to instances where both family conflict and negatively valenced emotions are sustained for long periods of time (Minuchin, 1985).

An intensive repeated measures approach

Capturing the exchanges between conflict and emotions remains challenging but is being facilitated by recent efforts to quantify family functioning on relatively short timescales (e.g., day-to-day; Fosco & Lydon-Staley, 2020; Leonard et al., 2022). Examining fluctuations on timescales shorter than monthly or yearly changes reflects that all types of families have “bad days” and “good days” on which they are more or less conflictual than usual (Chung et al., 2009; Cummings et al., 2003). These fluctuations in family conflict have implications for emotional experiences, with days of higher-than-usual family conflict, for example, being days of higher-than-usual negatively valenced emotions.

In addition to providing insights into how short timescale (e.g., daily) fluctuations in family conflict are associated with fluctuations in emotions, intensive repeated measures data facilitate the construction of feedback loops like those posited by family systems theory (Brinberg & Lydon-Staley, 2023). This ability to capture complex feedback loops and spirals with intensive repeated measures is perhaps best seen in an emerging emotion network perspective (Lydon-Staley et al., 2019; Pe et al., 2015). An emotion network perspective contextualizes the experiences of a discrete emotion within a broader emotion system comprising multiple discrete emotions and their associations with one another across time. Emotion networks capture the intuitive notion that one's current emotional state may give rise to other emotions. For example, one may feel embarrassment over an angry outburst—a form of secondary emotional response that has long been emphasized in emotion theory (Gross & Muñoz, 1995) and observed in empirical studies that capture the moment-to-moment transfer of emotions across time and states (Pe & Kuppens, 2012). Some degree of emotion transfer across time and states is normative. Emotion networks that support the spread of negatively valenced emotions across time and states, however, may place people at risk for psychopathology. Indeed, individuals with negatively valenced emotion networks that facilitate the substantial spread of emotions across time and states tend to exhibit a high number of depressive symptoms (Shin et al., 2022).

This emotion network perspective posits that emotions persist across time in strongly connected networks, as activity in one emotion is transferred to another emotion via causal associations. This spreading of emotional activity through networks is analogous to the behavior of two sets of domino tiles: one in which the tiles are far apart and a second in which the tiles are close together. In a densely connected emotion network, the set of domino tiles are close together, and when one tile falls it causes other tiles to topple, rippling through the system (Borsboom & Cramer, 2013). In a more sparsely connected emotion network, the set of domino tiles are far apart and when one tile falls the other tiles remain standing.

Embedding family conflict experiences in emotion networks

Emotion networks have been expanded to include not only discrete emotions but also social experiences that act as both causes and consequences of emotions (Yang, Ram, Gest, et al., 2018; Yang et al., 2018). Social experiences that occur in the family context, in particular fluctuations in conflict, make natural additions to a socioemotional network. We may posit, in line with family systems theory and recent empirical findings (e.g., LoBraico et al., 2020), that family conflict and discrete emotions are both causes and consequences, that impact one another across time.

Both family systems theory and an emotion network perspective share notions of the importance of a return to homeostasis following perturbation (Minuchin, 1985; Yang, Ram, Gest, et al., 2018; Yang et al., 2018). Experiences of family conflict and negatively valenced emotions are normative. Instead, sustained and repeated experiences of family conflict and negatively valenced emotions can place individuals at risk for psychopathology. By embedding variables that capture family conflict within socioemotional networks, we can use appropriate tools to capture the extent to which, after an increase in family conflict, the conflict perseverates over time as it impacts, and is impacted by, negatively valenced emotions.

In particular, impulse response analysis is a tool that may capture the complex spread of conflict activity through a socioemotional network (Lütkepohl, 2005). Once a temporal network that encodes the predictive associations between emotions and family conflict experiences across time has been constructed, an impulse response analysis simulates an exogenous impulse to certain socioemotional network nodes, one node at a time. After the system receives the perturbation, the propagation of this impulse through the network and the time-lagged symptom network edges can be charted. The impulse response function shows the hypothetical change in conflict or discrete emotion in response to a stimulated increase in conflict or emotion intensity over several time points, capturing the continuous interplay between nodes in a network and matching theoretical notions of feedback loops.

Family conflict in young adulthood

Changes in family experiences occur as youth transition from adolescence into young adulthood, become more independent from their families, move into new homes, finish school, and begin careers (Arnett, 2000). Yet, conflict remains an important dimension of family life during this time of instability and shifting of family relationships (Johnson et al., 2010). Indeed, family conflict sometimes decreases throughout adolescence and into young adulthood (Whiteman et al., 2011) and the family environment is theorized to lose its dominant function when environments outside of the family become more important (Arnett, 2000). Even in the context of these changes, however, family conflict remains associated with young adults' psychopathology (Lopez, 1991). Young adults who perceive their families to be cohesive feel less stress when transitioning to college compared to those who perceive their families to be less cohesive (Johnson et al., 2008). Similarly, those who report more frequent and intense conflict tend to experience more stress than young adults who report less family conflict (Lopez, 1991).

Zooming into the everyday lives of young adults will provide a complementary perspective to this long-timescale work and, by taking a socioemotional network perspective, nuanced insights into everyday exchanges between family conflict and young adult emotional experiences will be achieved. With the greater independence young adults have from their families relative to adolescence, both in terms of less time spent with families than with friends and romantic partners (Fuligni & Masten, 2010) and in terms of many leaving home for the first time to attend college or work (Arnett, 2000), a plausible hypothesis is that young adults may exert stronger spillover effects on family conflict than during adolescence. Indeed, although young adults take advantage of their individuality and independence from their families (Tsai et al., 2013), they may remain active in seeking parental support and guidance as they face the challenges associated with becoming an adult and, as a result, become drivers of family conflict as they bring their experiences with stressors into the family environment.

The present study

Guided by family systems theory, we evaluated the interplay between family conflict and three negatively valenced emotions (anger, anxiety, and sadness) at two timescales in a community sample of young adults, approximately 48% of whom were full-time college students. With daily reports, we draw from previous studies on adolescents showing that an increase in family conflict represents a stressor that is associated with increased negatively valenced emotions (Fosco & Lydon-Staley, 2020) and extend that work to young adults. Given the continued importance of the family in adulthood, despite some changes (Johnson et al., 2010), we hypothesize that days of higher-than-usual family conflict will be days of higher-than-usual negatively valenced emotions (H1).

To better understand the associations observed at the daily timescale, we next use intra-day measurements of family conflict and negatively valenced emotions to construct socioemotional networks that articulate the interplay between family conflict and negatively valenced emotions. We hypothesize that family conflict will show interplay with negatively valenced emotions across time, such that changes in family conflict at one time point will be associated with changes in emotions at the next time point and vice versa. Given the increased independence, young adults have from their families relative to adolescents, along with their continued tendencies to turn to the family for support when experiencing difficulties, we hypothesize that the influence of young adult negatively valenced emotions on family conflict will be more pronounced than the influence of family conflict on young adults' emotions (H2).

To further understand heterogeneity in socioemotional networks and their associations with psychopathology, we construct person-specific socioemotional networks and use impulse response analysis to quantify between-person differences in how increases in family conflict perseverate over time. The perseveration of family conflict is indicative of difficulty returning to homeostasis following perturbation, resulting in sustained and repeated experiences of family conflict theorized to place people at risk for psychopathology (Minuchin, 1985). As such, we hypothesize that socioemotional networks that facilitate the perseveration of family conflict will be associated with greater experiences of anxiety and depression (H3). By testing these hypotheses, our study offers a fundamental contribution to quantifying and understanding the reciprocal dynamic between young adults' family conflict and negatively valenced emotions.

Participants

Participants were 77 young adults (M = 21.18 years, SD = 1.75, 63 women, 14 men; see Table S1 for additional participant demographic characteristics) recruited from Philadelphia through poster, Facebook, Craigslist, and university research site advertisements. Individuals were eligible if they met 5 criteria: (1) were between 18 and 25 years of age; (2) had consistent access to a home computer/laptop with internet; (3) owned a smartphone; (4) were willing to complete a 2-h laboratory visit; (5) were willing to install a free app on their smartphone and computer. Most participants (87%) reported their mother as their first caregiver and most participants lived full time (23%) or some of the time (55.4%) in the same household with people identified as their first caregiver. Table S2 presents the family structures of all participants.

Procedure

Participants attended a laboratory visit and then completed 21 days of daily diaries and experience-sampling. Links to the end-of-day daily diary surveys were sent via email at 6:30 pm. The experience-sampling protocol consisted of a morning survey and five further surveys delivered via LifeData. The morning survey was sent at 7 AM (n = 1), 8 AM (n = 51), 9 AM (n = 13), 10 AM (n = 10), and 11:10 AM (n = 3), with participant-specific times elicited at the laboratory session to maximize completion of the morning survey. The morning survey came at the same time each day. The other five momentary surveys were distributed randomly throughout the day, with at least 30 minutes between each prompt. Participants were compensated with a payment card (see Supporting Information for payment details). Data collection began in July 2019 and ended in March 2020 when laboratory visits were no longer possible due to COVID-19.

Measures

The present study used participants' reports of demographic characteristics, family conflict, anxiety, and depression from the laboratory visit, as well as daily and momentary reports of family conflict, family interactions, sadness, anger, and anxiety.

Data preparation and analysis

We first examined how daily fluctuations in family conflict were associated with daily fluctuations in anger, anxiety, and sadness before creating socio-emotional networks that articulated intraday associations between family conflict, anger, anxiety, and sadness.

Daily family conflict, anger, anxiety, and sadness

Results of the multilevel models examining daily associations between family conflict and emotions are shown in Table S4. Day's family conflict was associated with anger (b = 0.20, p < 0.001, d = 0.24), such that on days when family conflict was higher than usual, anger was higher as well. There were positive but not statistically significant associations between day's family conflict and anxiety (b = 0.11, p = 0.06) and sadness (b = 0.09, p = 0.09). At the between-person level, people from families experiencing higher than the sample average levels of family conflict on average across the 21 days experienced more anger (b = 0.48, p < 0.001, d = 1.08), anxiety (b = 0.43, p = 0.01, d = 0.60), and sadness (b = 0.36, p = 0.003, d = 0.71).

Socioemotional networks of momentary family conflict and emotion: a group model

We next turn to within-day fluctuations in family conflict and emotion. This allowed us to capture potential reciprocal associations between family conflict and emotions. Results from the multilevel vector autoregressive model are presented in Table S5. The edges of the temporal network (Figure 1a) represent the prediction of variables at one measurement occasion by variables at the previous measurement occasion (i.e., earlier in the day) for a prototypical individual in the sample. We find that all associations are positive, indicating that increases in a variable at the current time point are associated with increases in other variables at the next time point. One temporal association emerged between the emotions and family conflict. Previous prompt's anger predicted current moment's family conflict such that increases in youth anger were associated with increased family conflict.

The edges of the contemporaneous network (Figure 1b) represent the co-occurrence of family conflict, anger, anxiety, and sadness, controlling for both temporal effects and the effects of all other variables at the same measurement occasion. All edges in the contemporaneous network are positive, indicating the reinforcement of associations among variables. When anger is higher than usual, family conflict is also higher than usual. No such association was found between family conflict and anxiety or sadness. The three negatively valenced emotions created a feedback loop such that when an increase in one emotion is experienced, increases in two other emotions are also experienced.

Person-specific momentary socioemotional networks, anxiety, and depression

We next moved from a group model for the prototypical individual in the sample to an estimation of person-specific temporal networks. The estimation of person-specific networks requires more data than the estimation of a group model that pools data across individuals. A person-specific model of the associations among the four variables fit the data of 70 of the 77 (90.91%) individual participants well; the goodness of fit is indicated by at least three of the following four criteria: RMSEA ≤0.08, SRMRs ≤0.08, CFIs ≥0.95, NNFI ≥0.95 (Beltz et al., 2016; Yang, Ram, Gest, et al., 2018; Yang et al., 2018). Participants for whom good fits were not obtained had less data than participants for whom good fits were obtained, t(76) = −5.77, p < 0.001.

Taking these 70 participants forward, we quantified individual differences in participants' socioemotional networks and their implications for the extent to which family conflict persists across time after perturbation using impulse response analysis (see Figure 2a,b). Specifically, we simulated an increase in family conflict and examined how its intensity persisted across time as an activity associated with family conflict traveled through each person's socioemotional network. We quantified this persistence using recovery time for family conflict and used the common logarithm to reduce the positive skew of the family conflict recovery time variable. Family conflict recovery time was positively correlated with trait anxiety (Figure 2c; r(69) = 0.28, p = 0.02) but exhibited no significant association with depression, r(69) = 0.08, p = 0.49, meaning that more anxious people took longer to recover from family conflict.

Limitations and future outlook

In contrast to survey-based methods that rely on collection over long periods of time (e.g., 30 days; Schwarz, 2007), this study collected intensive repeated measures with daily diary and experience-sampling methodologies, reducing retrospective biases and enhancing ecological validity of the findings (Shiffman et al., 2008). The study sample consists mostly of young college-going women with heteronormative family structure. Determining the extent to which findings generalize to different samples will require more diverse samples. Estimating person-specific models captured heterogeneity in socioemotional networks that had implications for the experience of anxiety but required more data than approaches that aggregated measurements across people. Indeed, person-specific networks could not be constructed for six individuals in the sample due to the length of the time series required for estimation, a difficulty observed in applications beyond family dynamics research (e.g., tobacco withdrawal, Lydon-Staley et al., 2021). Daily and momentary family conflict were measured based on only two items regarding criticizing each other and family members fighting with each other. Although these are two major items that indicate family conflict, other categories such as getting angry and losing temper that are measured in the dispositional family conflict scale were not included. Additionally, although ecological momentary assessment increases accuracy and reduces retrospective bias in self-reports as compared to retrospective reports over longer periods (e.g., past 30 days; Ebner-Priemer & Trull, 2009), people may lack insight into their mental processes (Nisbett & Wilson, 1977). This could explain the stronger associations observed in socioemotional networks between anger, as compared to both sadness and anxiety, and family conflict if participants associate conflict with anger more and therefore downplay the concurrent experience of other emotions. Although plausible, this is unlikely to be the only driver of the observed anger-family conflict associations given that associations between anger and family conflict were observed across time (i.e., t − 1 edges in the socioemotional networks) and not just concurrently (i.e., during the same time point in the networks).

In assessing daily interactions with family, we did not specify which family member (e.g., caregiver, sibling) young adults had interacted with that could further disentangle the specific types of family interaction or family conflict that impact young adults. The term family member is also a generic description of members in the family that does not differentiate between members that are family of origin or family of choice. Family of choice members are particularly influential to adults that are in marginalized communities and to their psychopathology (Hull & Ortyl, 2019). Future assessment of family dynamics could fruitfully cover the multifaceted sides of family conflicts as well as more specific definitions of family structure.

The short-term dynamics manifesting on the timescales of hours and days as captured here are embedded in developmental processes occurring over longer timescales of weeks, months, and years (Zemp et al., 2018). Future research will benefit from measurement burst designs wherein bursts of intensive repeated measures protocols that take several weeks to complete are repeated several times over longer periods (e.g., years; Gao & Cummings, 2019; Mastrotheodoros et al., 2022). This will allow tests of how socioemotional networks change during the transition from adolescence into young adulthood and how changes in emotional networks parallel changes in psychopathology.