his study investigates the interrelations among Relationship Maintenance, Satisfaction, Jealousy, and Violence within young Peruvian couples, particularly in the context of the post-pandemic environment, utilizing a network analysis approach.
Methods
A total of 832 participants aged 18 to 30 years (M = 20.94, SD = 2.29), comprising 645 females (77.50%) and 187 males (22.50%), were included in the study. The primary objective was to elucidate the relationships among network nodes, with a specific focus on the connections between dimensions of Relationship Maintenance and the constructs of Jealousy and Violence. The research also aimed to identify the central node within the network and to analyze gender-specific connections among nodes, employing the SMOTE algorithm to achieve gender data balance.
Results
The findings indicated a direct correlation between Complementarity and Jealousy, suggesting that intense shared interests may foster unhealthy dependence. An inverse relationship was observed between Companionship and Violence. Satisfaction emerged as a critical factor, underscoring its significance in the success of romantic relationships. Furthermore, the study revealed that men tend to prioritize Companionship and Sharing, potentially influenced by cultural norms, while women emphasize the Companionship-Complementarity connection, indicative of mutual support.
Conclusions
The research highlights the essential role of maintenance variables in influencing Satisfaction, Jealousy, and Violence within romantic relationships. The impact of the pandemic on romantic dynamics is evident, reinforcing the importance of Satisfaction. Future research should focus on gender equity and further investigate these interrelations.
Peer Review Reports
Introduction
The National Institute of Statistics and Informatics in Peru has reported a troubling trend: between 1993 and 2017, the population of separated and divorced individuals has increased, while the number of married individuals has declined [1]. This shift has significant implications for understanding romantic relationships in the Peruvian context, particularly given the importance young individuals place on these relationships as a contributor to personal happiness [2]. In light of these developments, the exploration of romantic relationships in Peru warrants a thorough investigation, particularly concerning factors such as marital satisfaction, integration, stability, and relational maintenance behaviors [3].
Global relationship dynamics have undergone notable changes in recent decades. Countries such as Canada and the United States have experienced rising divorce rates [4], often attributed to societal changes [5] and economic pressures [6], which have sparked scholarly debates. In the post-COVID-19 era, nations like Spain and Italy have also reported declining marriage rates [7]. The challenges posed by the pandemic, including enforced isolation and socio-economic uncertainties, have exacerbated relational strains globally [8, 9]. While these global trends provide a contextual backdrop, it is crucial to understand how they intersect with or diverge from the unique socio-cultural dynamics in Peru. By juxtaposing these global trends with the Peruvian context, we can derive insights into the distinctive factors shaping romantic relationships in the region and inform localized interventions and support mechanisms. This comparative approach enriches our investigation, allowing us to contextualize our findings within a global narrative.
A comprehensive examination of the intricate phenomenon of romantic relationships, particularly in the Peruvian context, necessitates an in-depth exploration of the variables involved. This article will introduce the topic, delve into relationship maintenance behaviors and factors affecting relationships such as jealousy and violence in Peru, explore the methodological approach, discuss clinical relevance, and conclude with the study objectives.
Maintenance behaviors are defined as the activities that couples engage in to preserve and prevent a decline in their romantic relationships [10, 11]. Previous research has indicated variations in the types and effectiveness of these behaviors across different cultural contexts, noting that these behaviors are more prevalent among females, with attachment variables potentially explaining these differences [12]. Five behaviors have been identified as foundational for successful romantic relationships: positivity, openness, assurance, social networks, and shared tasks [13]. In this context, maintenance behaviors enhance the quality of interaction and foster healthy bonds between couples [14]. Consequently, they serve as indicators of relational stability, helping to prevent deterioration or potential dissolution of the romantic relationship [15, 16]. This necessitates the presence of elements such as mutual commitment, effective communication, complementarity within the relationship, and expressions of affection and companionship [17, 18]. Indeed, companionship, humor, task collaboration, and verbal expressions of affection have been identified as the most valued aspects by adult couples in their romantic relationships [18]. Generally, companionship is vital in romantic relationships, as it embodies the desire for long-term commitment and the pursuit of mutually satisfying interactions [19]. Therefore, maintenance behaviors are critical indicators of satisfaction in romantic relationships, as demonstrated in previous research [20]. Relationship satisfaction refers to the subjective evaluation that an individual makes of their love relationship in the present [21], which is considered a key predictor of success and durability in romantic relationships [22]. In this regard, both satisfaction and love are essential components of romantic relationships [23]. This is particularly relevant as, in some countries, both dating and married couples experienced a significant decline in their levels of satisfaction and love following the conclusion of the COVID-19 emergency [24]. Such shifts underscore the urgent need to investigate how relationship dynamics, including maintenance behaviors and challenges such as jealousy and violence, have evolved in this new context. Relationship satisfaction has been linked to maintenance behaviors [14], as well as the quality of the relationship and overall life satisfaction [25]. However, evidence of a negative relationship with jealousy has also been documented [26].
Jealousy is regarded as an inherent emotion in romantic relationships, arising from real or imagined suspicions of a threat to affection within a valued relationship [27]. Other studies have identified factors such as insecurity and past relational traumas as contributors to jealousy [28]. Consequently, jealousy and distrust can be perceived as forms of negative relationship maintenance [29, 30]. Research has shown that expressions of jealousy vary by gender, with women typically expressing jealousy accompanied by feelings of sadness or depression, while men often express it through anger or aggression [31]. This aligns with a systematic review conducted by Pichon [32], which indicated a strong association between distrust, jealousy, and intimate partner violence.
Violence within relationships refers to attempts to exert dominance and control over another individual, whether physically, psychologically, or sexually [33]. Previous research has highlighted the multifaceted nature of relationship violence, linking it to factors such as power dynamics [34] and societal norms that some men may invoke to justify their use of gender-based violence [35]. The presence of violence in romantic relationships is a predictor of low levels of satisfaction, trust, and intimacy between partners, hindering the full development of the couple and obstructing the fulfillment of both partners’ needs [36].
While prior studies have examined the individual dynamics of relationship maintenance, satisfaction, jealousy, and violence, few have integrated these aspects into a comprehensive network analysis. The lack of research that holistically assesses the interplay of these variables presents a gap in our understanding. This study, therefore, aims to address this gap by employing a network analysis approach. To comprehensively explore the dynamics at play, we delve into multifaceted aspects of romantic relationships, such as complementarity, which is directly linked to satisfaction, relationship erosion, and physiological functioning post-conflict [37]. Affectivity significantly influences relationship satisfaction and romantic love, particularly in terms of intimacy, encompassing support provision, reception, and effective communication [38, 39]. Companionship plays a fundamental role, with shared novel activities enhancing relationship satisfaction [40]. Married women emphasize the value of companionship, offering presence, support, care, and trust for shared experiences and conversations [41]. Allocating quality time together emerges as pivotal, benefiting the relationship’s quality [42] and individual enjoyment, thereby promoting happiness during shared activities [43]. Understanding these dynamics necessitates acknowledging differing expectations and perceptions between genders [44]. Notably, women tend to exhibit higher dissatisfaction and contemplate separation, leading to increased divorce initiation rates [45]. Complex factors contribute to such dissatisfaction, including inequalities in labor division, varied expectations, and divergent notions of fairness and justice [46, 47].
Given the unprecedented relational challenges posed by the post-COVID-19 era, there is a compelling need for advanced analytical methods to comprehend these complexities. While traditional scientific evidence indicates that correlational studies between jealousy and satisfaction, as well as aggression and jealousy, have been conducted using Pearson correlation [26, 48], newer methodologies such as network analysis offer deeper insights. However, no studies have been identified that correlate variables using network analysis, a method considered novel and potentially more efficient than latent variable modeling for studying psychological attributes [49, 50]. Network analysis has increasingly been applied as a novel approach to understanding the nature and treatment of various variables associated with mental health across different domains [51]. In network analysis, symptoms of mental health are represented as nodes that interact and mutually reinforce one another within a network [49]. This methodology allows for the representation of relationships within and between mental health variables [49]. Although network analysis in psychology was initially employed to analyze psychopathological variables, it has recently expanded to other areas of psychology, including intelligence [52], personality [53], emotional intelligence [54], academic self-efficacy [55], and even the domain of romantic relationships [56]. Network analysis assesses the strength and nature of associations between nodes, disregarding the assumption that the summation of scores on these variables adequately describes psychological characteristics [57]. This approach facilitates the identification of central nodes, which are those with stronger connections to other nodes [58]. Previous studies utilizing network analysis in other fields have underscored its ability to provide nuanced insights into complex systems, showcasing its potential utility in the realm of romantic relationships [56, 59]. Thus, the examination of the relationships between relationship maintenance, satisfaction, jealousy, and violence in young couples through network analysis holds clinical utility, as it enables an understanding of variable-to-variable interactions [60]. Furthermore, it may be beneficial for identifying interventions that could be effective in treating individual syndromes within couples [61].
Consequently, the present study aims to estimate the network structure of nodes among relationship maintenance, satisfaction, jealousy, and violence in young couples in Metropolitan Lima. Additionally, it seeks to identify the interconnections between nodes, the central node, and to compare the network based on gender.
Method
Participants
The study involved 832 young adults aged between 18 and 30 years (Mean = 20.94, SD = 2.29), comprising 645 females (77.50%) and 187 males (22.50%). All participants were engaged in a romantic relationship lasting a minimum of three months, deemed essential for achieving a requisite level of stability [62]. The duration of the romantic partnerships varied from 3 to 139 months (Mean = 22.56, SD = 19.97). All participants were from a middle socioeconomic stratum and resided in Metropolitan Lima. The sample size was determined using the powerly package, with 10 nodes, a statistical power of 0.80, and a density of 0.40, indicating that a minimum of 262 observations was recommended [63]. Participant selection was conducted using a non-probability technique known as snowball sampling [64], necessitated by the impact of the COVID-19 pandemic on traditional in-person and large-scale surveying practices in Peru.
Instruments
All instruments utilized in this study are adaptations of previously validated measures within the Peruvian context, ensuring their appropriateness for use. The psychometric properties of the WAST-2 were examined as a preliminary aspect for its inclusion (see Supplementary Information).
The Relationship Maintenance Scale (RMS) [65]
A 14-item Peruvian version of the RMS was employed [66]. The RMS consists of Likert-type items ranging from 1 (Strongly Disagree) to 5 (Strongly Agree) and measures four factors: Companionship, Affection, Complementarity, and Shared Interaction. For instance, items include statements such as “We share ideals,” “We feel chemistry in our relationship,” and “We discuss our experiences.” Validity testing was conducted through exploratory and confirmatory factor analysis using the WLSMV estimator, yielding optimal goodness-of-fit indices (CFI = 0.977, RMSEA = 0.058) according to prior studies [67]. Reliability was assessed using the omega coefficient, demonstrating acceptable to good internal consistency for Companionship (ω = 0.78), Affection (ω = 0.83), Complementarity (ω = 0.77), and Sharing (ω = 0.70) [68, 69].
Relationship Assessment Scale (RAS) [21]
A five-item Peruvian version of the RAS was utilized [70]. This unidimensional questionnaire employs a Likert-type scale ranging from 1 to 5, measuring the level of satisfaction in romantic relationships. The RAS-5 scores range from 5 to 25 points. Example items include “Do you feel that your partner meets your needs?” and “Overall, to what extent are you satisfied with your relationship?” Two approaches were employed to evaluate the validity of the questionnaire: Item Response Theory (IRT) and Confirmatory Factor Analysis (CFA), both demonstrating excellent goodness-of-fit (RMSEA < 0.08, CFI > 0.95). Reliability was assessed using empirical reliability (rxx = 0.86) and the omega coefficient (ω = 0.84), indicating a high level of internal consistency [69].
The Woman Abuse Screening Tool (WAST-2) [71]
The Spanish version was utilized [72]. This instrument consists of two Likert-type items, with scores ranging from 0 to 4. The WAST-2 is unidimensional and designed to assess the presence of violent outbursts, tension, and difficulties in romantic relationships. The scale’s reliability was deemed acceptable for the study sample (ω = 0.66). Although the WAST-2 was originally designed to assess violence against women, its two items (1. “Overall, how would you describe your relationship with your partner? □High tension, □Some tension, □No tension; 2. “You and your partner resolve disagreements with: □A lot of difficulty, □Some difficulty, □No difficulty”) are sufficiently broad to measure episodes of violence in both genders.
The Brief Jealousy Scale (BJS) [73]
The BJS is derived from one of the dimensions of the Inventory of Emotional Communication in Romantic Relationships [74]. The BJS consists of nine items rated on a Likert scale from 1 (Not jealous at all) to 5 (Very jealous), assessing various scenarios in which an individual may experience jealousy. Example items include “If my partner spends significantly more time with another person, I would feel…” and “If I feel that my partner trusts another person more than I do, I would feel…”. The validity of the scale was established through confirmatory factor analysis, demonstrating acceptable goodness-of-fit (CFI = 0.97; SRMR = 0.03; RMSEA = 0.08) according to previous studies [67]. Additionally, reliability was determined using the omega coefficient (ω = 0.88), indicating good internal consistency [68].
Procedures
Prior to commencing the research, an evaluation of ethical considerations as stipulated in the Helsinki Declaration [75] and aspects related to conducting online research was conducted [76]. This was presented to the Research Ethics Committee of Universidad Privada del Norte (UPN) in Peru. Initially, the WAST-2 was analyzed, as it was the only instrument without a psychometric study in the Peruvian context. Given the brevity of the test, Item Response Theory (IRT) models were employed to assess differential functioning by gender. Specifically, the Expected Score Standardized Difference (ESSD) was utilized, which is based on expected scores and provides a measure of effect size in the latent trait [77]. An ESSD value greater than 0.30 indicates a small effect, greater than 0.50 indicates a moderate effect, and greater than 0.80 indicates a large effect. The results were favorable, permitting the inclusion of the WAST-2 (see Supplementary Information).
Due to the limitations and challenges arising from the COVID-19 pandemic in accessing participants through traditional means, a non-probabilistic snowball sampling method was employed. While this approach may have drawbacks, such as potential bias towards specific societal segments, it was essential for gathering information in the challenging post-pandemic landscape. This sampling technique was primarily chosen to maximize participant recruitment while leveraging the interconnected web of personal relationships and social circles. Participants were invited to participate through initial contacts who subsequently recommended other potential participants. While we acknowledge that this method may limit the generalizability of our results, it was a pragmatic solution given the post-pandemic circumstances.
Once contact was established with potential participants, they were provided with a consent form detailing the study’s objectives, assurances of anonymity, potential risks and benefits, and data handling protocols. Following this, they completed a sociodemographic questionnaire to contextualize responses and understand the diversity of the sample. After completing this preliminary questionnaire, participants proceeded to respond to self-reported questionnaires regarding their romantic relationships. These questionnaires included specific questions designed to capture the dynamics of romantic relationships. Participants were encouraged to respond honestly, with assurances that their responses would be treated without judgment or consequence. On average, the entire questionnaire suite took approximately 15 minutes to complete, with data collection occurring from March to June 2022. Comprehensive details of the data and R code were archived in the free OSF repository: https://osf.io/vbyhq/.
Data Analysis
Data analysis was conducted using the R programming language within the RStudio environment. The protocol recommended by the reporting standards for psychological network analyses was adhered to [78]. Consequently, network estimation, accuracy assessment, stability, and comparative analysis were performed.
Prior to the network analysis, an exploration of the variables or nodes of interest was conducted using Global Network Properties to describe the network. This included density (D), representing the proportion of existing connections in the graph; transitivity (C△), measuring the average tendency of nodes to form groups or communities within the network; and average shortest path length (APL), indicating the average number of links or connections required to traverse from one node to another. Finally, the small-world index (S) was calculated to evaluate the degree of association between nodes, with a recommended value exceeding 1 [79].
To estimate the network, the ggmModSelect function and Spearman correlation were employed within the RStudio environment. This combination was selected for its effectiveness in estimating asymmetric data [80]. Subsequently, centrality indices were examined. The Expect Influence index (EI) was prioritized, as it is most appropriate for networks containing negative signs [81]. Additionally, to evaluate nodes across different communities, the Bridge Expected Influence (BEI) index was utilized, representing the sum of edges (considering signs) between a node and other nodes outside its community [82]. Other centrality measures, such as closeness and betweenness, were not estimated due to their inadequacy for interpreting psychological variables [83] and their instability according to simulation studies [84]. It is important to note that the network is represented by nodes (circles) connected by edges (lines), with varying thickness indicating the strength of the interaction. Positive and negative correlations are denoted by green and red colors, respectively [85]. The Fruchterman-Reingold algorithm was employed to arrange the nodes, with stronger interactions centralized and weaker ones positioned on the periphery [86]. R² predictability indices were included in the estimations to indicate the percentage of variance explained by each node in relation to other nodes within the network [26].
The evaluation of edge weight accuracy involved the implementation of the bootstrapping technique, a rigorous statistical resampling method facilitated by the bootnet package. This method entailed iteratively modeling data randomly selected from the dataset, with edge values estimated in each iteration. To ascertain the precision of the edges, confidence intervals (CIs) were computed at a 95% level, with the width of the intervals reflecting the accuracy level [87]. Furthermore, a comprehensive visual representation, in the form of a plot, was devised to depict the frequency at which edges were unequivocally assigned a zero value.
The assessment of stability encompassed a meticulous analysis of a plot elucidating fluctuations in centrality indices following the removal of 70% of the data. Subsequently, a detailed comparative analysis was conducted, contrasting the resampled data against the original study data through the computation of their mean correlation. This intricate process culminated in the derivation of a comprehensive summary statistic, encapsulated within the Stability Correlation (CS), serving as a paramount metric discerning the extent to which data can be excised while maintaining a commendable correlation threshold of at least 0.70 with the centrality coefficients of the data. It is imperative to emphasize that the final CS value is anticipated to reside within the prescribed range of 0.25 ≤ CS ≤ 0.50, illuminating the robustness and reliability of the stability assessment [85].
A comparison was conducted according to gender, and given the significant differences between the groups, a statistical technique for unbalanced data was employed. The preferred technique for synthetic oversampling was the Synthetic Minority Over-sampling Technique (SMOTE), which has demonstrated good performance with extremely imbalanced [88] and categorical or ordinal data [89]. Additionally, the NetworkComparisonTest library package [90] was utilized. This package employs a permutation procedure to test the null hypothesis that both groups are identical, examining the differences after generating one thousand randomly obtained replicates. To determine effect size, Spearman correlations based on bootstrap were established, and the mean of the correlations obtained from one thousand resamples was reported. Furthermore, differences between the two networks were investigated by subtracting the values from the matrix, visualized in a corPlot graph, which facilitated the immediate identification of the most significant differences.
Results
Global Network Properties
The density analysis of the studied network revealed that 13 out of the 21 edges had a non-zero value, resulting in a density of 61.90%. A transitivity coefficient of 0.62 was found, indicating a favorable proportion of closed triangles in the network, exceeding the random transitivity of 0.53. Regarding the average shortest path length (APL), an average of 1.43 links is required to traverse from one node to another, suggesting high efficiency in information transmission. Finally, the small-world index obtained was 1.25, indicating proximity between nodes and efficient information propagation within the network.
