Stability and Change in Newlyweds’ Social Networks Over the First Years of Marriage

Abstract

Marriage sanctifies the relationship between two spouses, but what happens to their relationships with family, friends, and others who comprise their social networks? Scholarly accounts disagree about whether couples’ networks strengthen, weaken, or remain stable in the years after marriage. To reconcile competing perspectives, marriage licenses from lower-income communities were used to recruit 462 spouses (231 couples) in their first marriages. Each spouse independently provided data on 24 network members with whom they interact regularly (over 11,000 network members). These data were used to calculate 14 dimensions of each spouse’s social network, and networks were assessed in this way three times over the first 18 months of marriage. Over time, spouses’ networks grew to include more of each other’s family members, more married and financially secure individuals, and more members with whom they reported good relationships. For husbands, proportions of their own friends and their wives’ friends declined. Proportions of own family and members providing support did not change. With rare exceptions, these changes were not moderated by premarital parenthood, cohabitation, or relationship duration. Thus, regardless of a couples’ premarital history, getting married itself appears to be associated with specific changes in spouses’ social networks. Yet whether those changes broaden or narrow their networks depends on where in the network one looks. Illuminating how relationships between spouses are shaped by relationships outside the marriage therefore requires multifaceted assessments that are capable of distinguishing among discrete elements of couples’ social networks.

Individuals’ social networks consist of lovers, friends, family, coworkers, and all the relationships among them. When two individuals marry, a legal and often religious ceremony recognizes and sanctifies one of these relationships. What happens to the rest of them? A prevalent view asserts that, after marriage, shared holidays and life transitions (e.g., having children) make couples’ networks closer and more integrated (Cox et al., 1997; Milardo, 1982). A contrary view describes marriage as weakening social networks via withdrawal from friends and forced interactions with potentially unpleasant in-laws (Johnson & Leslie, 1982; Williamson et al., 2013). A third perspective, informed by the convoy model (Antonucci & Akiyama, 1987), emphasizes the continuity of social networks by suggesting that people are likely to stay in contact with their closest network members throughout their lives. To date, social network research on couples has seldom collected the longitudinal data necessary to resolve the tension among these accounts. Addressing this gap, the current analyses draw upon multiple waves of detailed social network data collected from a diverse sample of recently married couples to examine how their networks change or remain stable across the early years of marriage.

Sources of Change in Couples’ Social Networks

Early network theories identified marriage as a vehicle for integrating individuals into society (e.g., Berkman et al., 2000). Upon marrying, spouses frequently take the same last name, combine finances, and develop a joint identity (Emery et al., 2021). The growing connections between two individuals may also have implications for their social networks, including the development of connections with and between in-laws, the formation of new shared friends, or the shedding of weak ties that demand too much time and energy. A widespread assumption is that these changes are beneficial, leading to greater integration throughout the network (e.g., Kearns & Leonard, 2004). Wedding rituals, for example, take place in front of family and friends of both spouses, and so may facilitate connections not only between partners but among their loved ones as well (Chesser, 1980). Getting married may therefore be accompanied by changes in the composition of spouses’ networks, with each partner including more of the other’s family in particular, as suggested by prior work showing that family networks tend to increase in size after marriage (Kalmijn, 2012). This may be especially true for wives, who typically engage in more caretaking responsibilities for in-laws than do husbands (Lee et al., 2020).

The early years of marriage may also enhance the quality of spouses’ network relationships. Relationships outside marriage represent potential sources of material, emotional, and informational resources, so newlyweds should favor higher quality network relationships to ensure support for the many transitions (e.g., moving in together, having children) that characterize this stage of life (Cox et al., 1997; Haggerty et al., 2022). The early years of marriage may therefore be a time when couples gravitate toward network members who provide more support for and approval of the marriage, i.e., those who are best positioned to help the couple through periods of instability. Again, wives may be particularly likely to deepen bonds with these stronger ties, as women are typically socialized to deepen relationships with a few close others, whereas men tend to emphasize larger collections of weaker ties (Halevy & Kalish, 2021). Because marriage typically occurs after individuals have navigated the transitions of emerging adulthood (Robinson, 2015), marriage may also motivate strengthening ties to network members who have also made these transitions, i.e., those who are married, employed, or financially secure.

Despite the persistence of this view, a contrasting perspective suggests that marriage can undermine spouses’ social networks. Because time is a limited resource, spending more time with in-laws may require spouses to spend less time with their own family and friends (Hill & Dunbar, 2003). Indeed, partners’ friendship networks in particular tend to be smaller in longer intimate relationships (Johnson & Leslie, 1982) and married individuals do socialize less with friends than single indviduals (Sarkisian & Gerstel, 2015), not only because of limited time but because spouses may fill supportive roles previously occupied by friends (Kalmijn, 2012). These changes may mean that, after couples marry, spouses have fewer people to rely on for support during times of stress, and they may lose access to the weaker ties that afford individuals access to resources (e.g., job opportunities) that closer ties do not provide (Giulietti et al., 2018; Granovetter, 1973). In addition to changes in composition, network quality may decline in the early years of marriage as partners spend less time with voluntary relationships (e.g., friends) and devote more energy to in-laws. To the extent that managing relationships with in-laws is a common area of disagreement among newlyweds (Williamson et al., 2013), family gatherings may force partners into repeated contact with network members they do not like.

Sources of Stability in Couples’ Networks

The prediction that couples’ social networks change after marriage rests on the assumption that getting married has ripple effects throughout spouses’ other relationships. In recent decades, the evolving meaning of marriage has called that assumption into question. Where marriage once represented a new beginning for a couple, marriage has increasingly come to be seen as “a status one builds up to, often by living with a partner beforehand, by attaining steady employment or starting a career, by putting away some savings, and even by having children” (Cherlin, 2004, p. 855). To the extent that many of the transitions formerly associated with marriage now occur prior to marriage, many of the associated social network changes may have already occurred by the time a couple gets married.

The fact that, on average, couples are marrying at older ages is consistent with the view that couples’ lives, and thus their social lives, may be more stable after marriage than was true for couples in earlier generations. Although Americans do change jobs more than 12 times in their lifetime on average, most of these transitions happen between ages 18 and 24 (United States Bureau of Labor Statistics, 2019). The median age at marriage is now nearly 30 for women and over 30 for men (United States Census Bureau, 2020). Meanwhile, internal migration in the United States has been steadily decreasing over the last several decades, with just over 1% of Americans moving across state lines every year (United States Census Bureau, 2019). Thus, by the time they marry, many couples will have already settled into a region and a career path, and thus will have already experienced the transitions most likely to lead to new social relationships. Network properties tied to employment (e.g., proportion of network employed or financially stable) and geography (e.g., maintaining contact with one’s own friends) may therefore be relatively stable during this period. Similarly, research informed by the convoy model suggests that, even as our weaker network ties may cycle in and out of our networks over time, a “convoy” of our closest social network relationships (e.g., family members and best friends) may be relatively constant throughout our lives (Antonucci & Akiyama, 1987; Wrzus et al., 2013).

Moderators of Change in Couples’ Social Networks

Marriage is frequently accompanied by signs of commitment (e.g., living together, having a child) that could be associated with changes in newlyweds’ networks. Yet, couples do not always wait for marriage to have kids or live together. To the extent that changes in couples’ social networks are associated with these transitions rather than the institution of marriage itself, then couples who have negotiated more of these transitions prior to getting married should experience fewer changes in their networks after marriage. In contrast, if network changes are uniquely associated with the institution of marriage, then couples who marry should all experience similar changes. Teasing apart these possibilities requires comparing changes in post-marital social networks across couples with different premarital histories.

Premarital parenthood.

Around 40% of all births occur outside of marriage and over 30% of first-married mothers already have a child at marriage (Lavner et al., 2020; Martin et al., 2019). Parenthood alters couples’ lives in ways that should lead to significant changes in their networks whether they are married or not. For example, the demands of parenthood affect couples’ daily routines, such that parents, compared to non-parents, have less time and energy to spend on commitments outside of childrearing (Neilson & Stanfors, 2018). With less flexible time, couples with children may prioritize closer network members, and those who provide more support and demand fewer resources, over more distal or less supportive network members. Scholars have found, for example, that parenthood is associated with smaller friendship networks and less contact with friends, but the family network remains relatively unchanged (Kalmijn, 2012; Rözer et al., 2017). Bost, Cox, Burchinal, and Payne (2002) similarly found that, immediately following the birth of a child, couples spend less time with their friends and more time with family, although they also feel less satisfied with the amount of support they receive. Thus, couples who have had children prior to marrying may begin their marriages with more family and fewer friends in their networks, compared to those without children.

Premarital cohabitation.

Cohabiting may also instigate a process of network change that begins only after marriage for couples who do not cohabit. In a study in which individuals were asked to name a maximum of five network members, Kalmijn (2003) found that couples who cohabited reported fewer friends but more shared friends, compared to couples who did not cohabit. This suggests that, prior to marriage, cohabiting partners may have already begun integrating their individual networks and shedding weaker ties. By the wedding, those who have cohabited may therefore have more integrated networks than those who did not cohabit, and may have already experienced changes that couples who did not cohabit have yet to experience.

Relationship duration.

If connections between partners’ families and friends develop when the relationship begins, rather than when the marriage begins, then the networks of couples in longer relationships could look very different by the time they do marry, with more shared family ties, greater approval of and support for the relationship, and more positive relationships with network members compared to the networks of couples who began their relationships more recently. Cross-sectional and retrospective reports indicate that as a couple’s commitment increases (e.g., from casually dating to seriously dating to engaged), overlap between partners’ networks increases and couples tend to spend more time with closer network members than with more distal ones (Milardo, 1982). Such findings suggest that couples in longer relationships should begin their marriages with closer and more integrated networks, and may therefore experience fewer changes after marriage, compared to couples in shorter relationships.

Overview of the Current Research

To date, several limitations in prior research on the social networks of married couples have prevented this work from resolving the tension between disparate accounts of network change during the early years of marriage. First, prior work has frequently relied on retrospective data, despite the notorious inaccuracy of partners’ reports of their own change over time (Frye & Karney, 2004). Second, the vast majority of this research has been cross-sectional, with changes inferred from between-group comparisons across ages and marital status groups, a strategy that is subject to cohort and selection effects. Third, most research has focused on a small number of network dimensions (e.g., received support or perceived approval), preventing a comprehensive description that might allow for both stability and change. Fourth, research on social networks has relied on global self-reports of network qualities, such as partners’ perceptions of how much emotional support they receive from their network as a whole (e.g., Sprecher & Felmlee, 1992). To the extent that these reports may be driven by a participant’s global feelings of well-being or recent interactions with more or less salient network members (Marsden, 2005), they may not correspond to qualities of the actual network. Fifth, the few studies that have examined specific network relationships have generally assessed only a small number of network members (typically 5 to 7). To characterize stability and change in couples’ networks over time, assessments of more network members are necessary (Hill & Dunbar, 2003), particularly because distal relationships may be more likely to change than closer ones (Wellman et al., 1997). Finally, to date social network research with couples has examined samples of mostly White, relatively affluent, college educated samples of convenience. Given that White individuals with higher SES are less likely than non-White and lower SES individuals to interact regularly with (Ajrouch et al., 2001) and exchange social support with (Taylor et al., 2013) extended family and friends, prior work may not generalize to more diverse samples in which social networks are more relevant to overall well-being (Huurre et al., 2007).

To assess changes in the properties of couples’ networks during the early years of marriage, these analyses overcome the limitations of prior research by drawing upon detailed social network interviews conducted over multiple assessments with a diverse sample of first-married, newlywed couples solicited from lower income communities. These data are appropriate for several reasons. First, the sample includes couples with a wide range of socioeconomic status (SES) and racial identity, expanding the scope of this work beyond the narrow samples used in prior studies. Second, the network interviews were extensive, soliciting reports on the characteristics of and relationships among 24 network members from each spouse in three interviews conducted at 9-month intervals. Third, this longitudinal study includes reports from both spouses in each couple, allowing for direct assessment of network integration across spouses. Analyses of these data address the following research questions:

Which qualities of the network remain stable and which change during the early years of marriage? By measuring multiple aspects of network composition and quality three times during the first 18 months of marriage, this study aims to resolve competing views about the relative degree of stability and change in network properties during the transition into marriage.

Are patterns of stability and change in network properties in the early years of marriage moderated by premarital parenthood, premarital cohabitation, and relationship duration? Assessing each moderator’s effects at the start of marriage and over time will address whether network change is precipitated by relationship events (e.g., cohabitation, parenthood) before the wedding or whether entering marriage itself initiates this process.

Method

Sampling

Between 2014 and 2015, recently married couples living in lower income communities were identified through marriage license applications obtained from the Recorder’s Office in Harris County, Texas, the third most populous county in the United States. Because sample recruitment occurred before the legalization of same-sex marriage, all couples were mixed-gender. Addresses were matched with census data to identify couples living in census block groups where a minimum of 30% of households were categorized as living below the poverty line (United States Census Bureau, 2013). Because eligibility was linked to neighborhood income rather than household income, this sampling frame was selected to maximize the likelihood of including poorer couples that have been often overlooked in prior research.

Through this procedure, 4,916 couples were identified for screening on the telephone or in person. Among the couples attempted for screening, 3,535 could not be reached, 224 refused screening, and 1,157 agreed to be screened. Interviewers screened couples to ensure they had married, partners were in their first marriage, partners spoke English or Spanish, and both partners were at least 18 years old. This screening identified 506 eligible couples, and 401 agreed to participate in the study. The 231 couples who were interviewed prior to the close of the baseline assessment period comprise the current sample. All procedures were approved by the RAND Corporation Institutional Review Board.

Participants

The final sample of 231 couples had all been married within 12 months of the first assessment (M = 5.5 months, SD = 2.0). At baseline, wives ranged in age from 18 to 56 years (M = 28.1, SD = 7.4) and husbands ranged in age from 18 to 53 years (M = 29.5, SD = 7.5). In terms of racial identification, 53% of wives and 52% of husbands identified as Hispanic, 35% of wives and 32% of husbands as Black, 9% of wives and 10% of husbands as White, and 4% of wives and 7% of husbands as Other/Multi-racial. The majority of wives (54%) and husbands (60%) reported receiving less than or the equivalent of a high school diploma or GED. Approximately 16% of wives and 12% of husbands reported a college degree or higher education. The median annual household income was $32,500 (SD = $29,590) and ranged from $0 to $175,000, reflecting the success of our sampling procedure at soliciting couples with a wide range of SES.

Procedure

At baseline (T1), two trained interviewers visited couples in their homes. The interviewers took spouses to separate areas to describe the study, obtain informed consent, and verbally administer the survey. Two portions of the interview are relevant for the current investigation. First, participants completed a standard interview that included all demographic and moderator measures. Second, participants completed a detailed social network interview. During this interview, participants were asked to name 25 network members (starting with their spouse) with whom they had any form of contact over the past 12 months. Spouses answered several questions about each individual they named, providing the basis for 14 variables.

This investigation utilizes data from three data collection waves, each spaced nine months apart. Couples were not interviewed if they reported a divorce or permanent separation between assessments. Data from all 231 couples were included in these analyses: Spouses were not required to complete social network interviews at each wave nor were they required to name 25 social network contacts to be included. At Time 1, 99% of wives and husbands completed social network interviews in addition to their baseline interviews. At Time 2, 93% of the baseline couples could be recontacted, 91% of those remained intact, and 99% of the intact couples completed a second network interview. Retention at T3, while still high, was lower than T1 and T2 in part because the field period was ended early due to Hurricane Harvey, which made landfall in Texas in August 2017. Consequently, 164 wives (of 165 who participated) and 160 husbands (of 161 who participated) completed the social network interview at T3. Across assessments, over 90% of spouses named the full 25 social network members in each interview.

Over the course of the three data collection waves, 11 couples divorced while another seven did not have marital status data at the third wave because they were unable to be contacted. To evaluate the sensitivity of our results to the presence or absence of these couples, we ran analyses with and without divorced couples at T3. The results were not substantively different and so we only report analyses below that include all couples regardless of T3 marital status. Additionally, we assessed the differences between couples who provided complete data and couples who provided incomplete data. Across the 20 demographic, moderator, and dependent variables described in this report, participants who provided data at all time points differed significantly from those who did not on only two variables: There was a greater proportion of premarital cohabitors among those with complete data for both husbands (t(134.0) = 2.9, p < .01) and wives (t(131.5) = 2.5, p = .01), and wives with complete data had a greater proportion of the network who approved of their relationship at baseline (t(107.1) = 2.5, p = .02).

Measures

Network interview.

Participants were asked to identify 24 individuals besides their spouse with whom they interact regularly, and then answered several questions about each network member. The questions assessed each member’s gender, marital status, divorce status, parental status, employment status, financial status, relationship to the spouse, quality of the relationship with the spouse, as well as two questions regarding the spouse’s experience of receiving support from the person (i.e., one for emotional support, one for tangible/concrete support), two questions about giving support to the person (i.e., one for emotional support, one for tangible/concrete support), and perceived approval of the marriage by the network member. The full text of each question is provided in the online supplement: https://osf.io/ms2n6/.

Network measures.

Responses to these questions were used to derive indices that summarize the proportion of these 24 relationships (excluding the spouse) that reflect each of 14 different network dimensions. Ten dimensions capture aspects of network composition: proportion females, married, divorced, raising children under 18, employed, good finances, own family, spouse family, own friend, and spouse friend. An additional four dimensions capture aspects of network quality: proportion good relationships, members who approve of the marriage, members from whom they received support, and members to whom they gave support. Received support and gave support are each aggregates of the two support questions stated above. A network member was coded as 1 for received support if the respondent had received either concrete or emotional support (or both) from the network member (0 if not), and a network member was coded as 1 for gave support if the respondent had given either concrete or emotional support (or both) to the network member (0 if not).

Premarital cohabitation.

Premarital cohabitation was measured using a single item: “Did you and your spouse live together before you were married?” Both spouses provided the same answer to this question in 87% of couples. In light of numerous studies documenting that wives are significantly more accurate in recalling relationship events and dates than husbands, including those related to residence changes (e.g., Mitchell, 2010), wives’ responses to the premarital cohabitation question were used in couples that disagreed. Of the 231 couples, 169 (73.2%) wives indicated that they had cohabited before marriage.

Relationship duration.

Relationship duration was derived from a single item: “Thinking about your relationship since you first got together, how many years have you been together with [FILL SPOUSE NAME]?” Spouses’ responses correlated at r = .94. Again, when spouses disagreed, we used wives’ responses. To account for outliers, responses above the 97.5^th percentile were winsorized, which retains all participants’ data (Liao et al., 2016). Five responses greater than the 97.5^th percentile (i.e., 15 years) were recoded to 15. The mean relationship duration of this transformed variable for all couples was 4.4 years (SD = 3.6). Although relationship duration differed significantly between cohabitors (M = 5.0, SD = 3.7) and non-cohabitors (M = 2.9, SD = 2.8; t(144.2) = 4.5, p < .01), the two variables correlated at only r = .25, justifying our decision to include both in these analyses.

Premarital parenthood.

At baseline, spouses were asked four questions: 1) number of biological children with the spouse, 2) number of own biological children from a previous partner, 3) number of biological children the spouse has from a previous partner, and 4) number of other children that the participant or their spouse have (e.g., foster, adopted). Premarital parenthood was indicated when any of these values were greater than zero at the baseline assessment. Using this definition, husbands’ and wives’ reports agreed 99% of the time. In the three cases when spouses disagreed, we again used wives’ responses. Of the 231 couples, 149 (64.5%) were parents at T1, indicating at least one child born before the marriage or a wife who was pregnant at the time of the marriage. Due to a small number of first births (n = 19) during the 18 months of this study, we did not assess the effect of becoming a parent post-marriage.

Analytic Plan

Multilevel modeling (MLM) was conducted in SPSS Version 26 using Restricted Estimation Maximum Likelihood (REML) to estimate random effects (Corbeil & Searle, 1976). Couples were included if husbands and wives each provided data for at least one timepoint. We use a two-intercept approach and an interaction approach to handle dyadic data (Planalp et al., 2017; Raudenbush et al., 1995). The two-intercept approach provides separate husband and wife intercepts and slopes while the interaction approach tests whether husbands and wives differ significantly on a given parameter. Regardless of the approach, we simultaneously model the data from both dyad members while accounting for the covariance between husbands’ and wives’ responses (Planalp et al., 2017; Raudenbush et al., 1995). The models nest time (measured in waves; i.e., 1, 2, 3) within individuals within couples, but this can be conceptualized as a two-level model with time and individual nested within couple (Raudenbush et al., 1995). Although we use both the two-intercept and interaction approaches, our focus is on the parameters from the two-intercept approach, supplemented by gender differences derived from the interaction approach. The within-couple equation for the two-intercept approach takes the following form:

$$y_{idt}=b_{1d}\left(Husban{d}_{id}\right)+b_{2d}\left(Wif{e}_{id}\right)+b_{3d}\left(Tim{e}_{idt}\ast Husban{d}_{id}\right)+b_{4d}(Tim{e}_{idt}\ast Wif{e}_{id})+e_{idt}$$ (1)

where $b_{1d}$ represents the intercept of a network variable for husbands, $b_{2d}$ represents the intercept for wives, $b_{3d}$ represents the slope for husbands, and $b_{4d}$ represents the slope for wives.

Including variables that differ between dyads, the between-couple equations are:

$$b_{1d}=b_{10}+b_{11}\left(Cohabitatio{n}_d\right)+b_{12}\left(Duratio{n}_d\right)+b_{13}\left(Parenthoo{d}_d\right)+{\zeta }_{1d}$$ (2)

$$b_{2d}=b_{20}+b_{21}\left(Cohabitatio{n}_d\right)+b_{22}\left(Duratio{n}_d\right)+b_{23}\left(Parenthoo{d}_d\right)+{\zeta }_{2d}$$ (3)

$$b_{3d}=b_{30}+b_{31}\left(Cohabitatio{n}_d\right)+b_{32}\left(Duratio{n}_d\right)+b_{33}\left(Parenthoo{d}_d\right)$$ (4)

$$b_{4d}=b_{40}+b_{41}\left(Cohabitatio{n}_d\right)+b_{42}\left(Duratio{n}_d\right)+b_{43}\left(Parenthoo{d}_d\right)$$ (5)

In Equation 2, $b_{1d}$, the coefficient representing the intercept for husbands in the within-couple equation, now is further explained by three between-couple covariates. To aid interpretability, premarital cohabitation and premarital parenthood were contrast coded. The value −0.5 represented non-cohabitors while 0.5 represented cohabitors, and −0.5 represented premarital non-parents while 0.5 represented premarital parents. Likewise, relationship duration was mean-centered. To illustrate this equation, take proportion married as an example. The term $b_{10}$ represents the average intercept of proportion married across husbands at different levels of premarital parenthood (i.e., premarital parent and non-premarital parent) and premarital cohabitation (i.e., premarital cohabitor and non-premarital cohabitor), and at the average relationship duration. The $b_{11}$ term represents the difference in the intercept between a husband who did cohabit and one who did not cohabit, given the same relationship duration and premarital parenthood. The between-couple residual, ${\zeta }_{1d}$, captures between-dyad variation in husband intercepts. Equations 3 and 4 drop the error term as slopes were fixed across dyads. Models with random slopes faced nonconvergence issues. When there are not enough degrees of freedom to estimate between-couple variance, fixed slopes are used for model identification (Planalp et al., 2017). Given the 14 network variables under examination, we estimated 14 models for these analyses. Analyses were not preregistered. We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the study. Supplemental materials, including syntax, data file, and question text are available online: https://osf.io/ms2n6/.

Results

Descriptive Statistics and Preliminary Analyses

The dimensions of husbands’ and wives’ networks correlated weakly across spouses (range: .03 – .46). Additionally, the mean absolute value within-spouse correlation between network dimensions was r = .12 at Time 1 (range: −.57 − .61), indicating that network dimensions were relatively independent of each other, justifying analyses that address each dimension separately. Patterns of bivariate correlations, means, and standard deviations for husbands and wives were similar at Time 2 and Time 3. Bivariate correlations and baseline descriptive statistics for all variables are provided in the supplemental materials (Table A1).

With respect to average network composition at baseline, husbands’ and wives’ networks primarily consisted of own family (35% for husbands and 39% for wives) and own friends (36% for husbands and 30% for wives). Only 9% of the husbands’ networks and 14% of the wives’ networks consisted of spouse family and less than 5% of husbands’ and wives’ networks consisted of spouse friends. About half of the network members that spouses listed at baseline were married or raising children under 18, and relatively few were divorced (16% for both spouses). For husbands and wives, around three-quarters of network members were employed, although fewer were financially stable (62% for husbands and 57% for wives).

With respect to average network quality at baseline, spouses reported good relationships with most of their network members, with wives reporting good relationships with 75% and husbands with 83%. Likewise, spouses generally perceived high rates of relationship approval from their networks (83% for wives and 91% for husbands at Time 1). Husbands received support from 28% of their network members and gave support to 32% while wives had higher proportions in both categories (35% and 38%, respectively).

Dyadic Growth Curve Models

Beta coefficients of the dyadic growth curve models for husbands and wives are shown in Table 1. Because we consider between-couple covariates, intercept and slope estimates control for the effect of premarital parenthood, relationship duration, and premarital cohabitation.

Table 1.

Mixed Model Intercepts and Slopes of Network Variables for Husbands and Wives

Measure (Proportion of Network)	Husband Intercept (b (SE))	Wife Intercept (b (SE))	Test of H/W Intercept Difference (b (SE))	Husband Slope (b (SE))	Wife Slope (b (SE))	Test of H/W Slope Difference (b (SE))
Composition
1. Female	0.35 (0.01)	0.65 (0.01)	0.31 (0.01)**	0.00 (0.01)	0.00 (0.01)	0.00 (0.01)
2. Married	0.47 (0.01)	0.48 (0.01)	0.01 (0.02)	0.04 (0.01)**	0.02 (0.01)**	−0.02 (0.01)
3. Divorced	0.15 (0.01)	0.16 (0.01)	0.01 (0.01)	0.01 (0.01)*	0.00 (0.00)	−0.01 (0.01)
4. Raising Children	0.48 (0.01)	0.48 (0.01)	−0.01 (0.01)	0.00 (0.01)	0.00 (0.01)	0.00 (0.01)
5. Employed	0.78 (0.01)	0.74 (0.01)	−0.04 (0.01)**	0.03 (0.01)**	0.01 (0.01)	−0.02 (0.01)
6. Good Finances	0.63 (0.02)	0.59 (0.02)	−0.04 (0.02)	0.02 (0.01)	0.03 (0.01)**	0.00 (0.01)
7. Own Family	0.35 (0.02)	0.37 (0.01)	0.02 (0.02)	0.00 (0.01)	0.01 (0.01)	0.00 (0.01)
8. Spouse Family	0.09 (0.01)	0.14 (0.01)	0.05 (0.01)**	0.01 (0.01)*	0.02 (0.01)**	0.00 (0.01)
9. Own Friend	0.36 (0.02)	0.32 (0.01)	−0.04 (0.02)*	−0.04 (0.01)**	−0.01 (0.01)	0.02 (0.01)
10. Spouse Friend	0.02 (0.00)	0.04 (0.01)	0.02 (0.01)**	−0.01 (0.00)**	−0.01 (0.00)	0.00 (0.00)
Quality
11. Good Relationship	0.85 (0.01)	0.75 (0.01)	−0.10 (0.02)**	0.03 (0.01)**	0.04 (0.01)**	0.01 (0.01)
12. Received Support	0.27 (0.02)	0.36 (0.02)	0.09 (0.02)**	−0.01 (0.01)	−0.01 (0.01)	0.00 (0.02)
13. Gave Support	0.33 (0.02)	0.39 (0.02)	0.05 (0.03)*	−0.03 (0.01)*	0.00 (0.01)	0.03 (0.02)
14. Approve Relationship	0.93 (0.02)	0.84 (0.02)	−0.09 (0.02)**	−0.01 (0.01)	0.03 (0.01)*	0.04 (0.02)*

Note: All intercepts significantly different from 0. For slopes:

^*p < .05,

^**p < .01.

Intercepts.

Compared to their husbands, wives began marriage with significantly more female and fewer employed network members, as well as more spouse’s family and spouse’s friends, but fewer own friends. Thus, by the beginning of their marriage, wives seemed to have invested more in their husbands’ networks than vice versa. Of the network quality measures, all four differed between spouses. Compared to husbands, wives gave and received more support to and from their network members at the beginning of the marriage, but had fewer good relationships with their network, and perceived less relationship approval.

Network change.

Table 1 also shows the linear rates of change for each network characteristic for husbands and wives and the test of the difference between husbands’ and wives’ slopes. Over 18 months, six of the 10 network composition dimensions changed significantly for husbands, while three changed significantly for wives. Many of the changes reflected increases in connection and access to social capital. For example, husbands and wives both experienced a significant increase in the proportion of married people and spouse family, as well as in the proportion of employed network members. Wives also experienced a significant increase in the proportion of financially stable network members across the 18 months. Regarding network quality, husbands and wives both experienced increases in the proportion of good relationships, husbands experienced a decline in the proportion of their network to whom they gave support, and wives experienced an increase in their perception of relationship approval. Husbands also experienced a significant increase in the proportion of divorced network members and a significant decline in the proportion of own friends and spouse friends. The decline in own friends and spouse friends occurred in conjunction with an increase in the proportion of spouse family, indicating that, for husbands, integrating in-laws within the network comes at a cost of losing relationships with friends. In contrast, the proportions female, parents of children under 18, own family, and received support did not change significantly for husbands or wives over 18 months. The stability of own family further clarifies the process of network integration: Although husbands discard friends to accommodate in-laws, spouses hold on to their own family relationships. The difference between husbands’ and wives’ slopes was significant on only one dimension: Wives experienced a significant increase in their receipt of network approval while husbands reported a (nonsignificant) decrease. This may be a consequence of wives beginning their marriage with significantly lower perceptions of network approval than husbands.

Comparing Premarital Parents and Non-parents

Controlling for premarital cohabitation and relationship duration, network composition differed in several ways for premarital parents and non-parents (see Table 2). Compared to those who began their marriages without children, husbands and wives who had children started the marriage with significantly more parents of children under 18 and fewer spouse friends in their networks. Husbands who were premarital parents also had more divorced network members than non-parent husbands. Additionally, wives who were premarital parents had more of their own family in their network than did non-parent wives, perhaps an indication that wives with children are relying on their own family to help with childcare. Wives who were premarital parents also had fewer employed network members than those who were premarital non-parents. Network quality differed on only one dimension: Husbands and wives with a child began the marriage with fewer good relationships. The significant differences in the networks of parent and non-parent spouses at baseline did not generally extend to network change. Only one significant effect was found: Non-parent husbands experienced a significantly steeper decline in spouse friends over the first 18 months of marriage than husbands with premarital children.

Table 2.

Mixed Model Intercepts and Slopes Assessing Effect of Premarital Parenthood on Network Variables

Measure (Proportion of Network)	Test of Parent/Non-Parent Husbands Intercept Difference (b (SE))	Test of Parent/Non-Parent Wives Intercept Difference (b (SE))	Test of Husband/Wife Intercept Effect Difference (b (SE))	Test of Parent/Non-Parent Husbands Slope Difference (b (SE))	Test of Parent/Non-Parent Wives Slope Difference (b (SE))	Test of Husband/Wife Slope Effect Difference (b (SE))
Composition
1. Female	−0.02 (0.02)	0.01 (0.02)	0.03 (0.02)	0.00 (0.01)	0.00 (0.01)	0.00 (0.01)
2. Married	0.00 (0.03)	0.00 (0.02)	0.00 (0.03)	−0.01 (0.01)	−0.02 (0.01)	−0.01 (0.02)
3. Divorced	0.05 (0.02)**	0.02 (0.02)	−0.02 (0.02)	−0.01 (0.01)	0.00 (0.01)	0.01 (0.01)
4. Raising Children	0.09 (0.02)**	0.10 (0.02)**	0.01 (0.03)	−0.02 (0.01)	−0.02 (0.01)	0.00 (0.02)
5. Employed	−0.03 (0.02)	−0.06 (0.02)**	−0.04 (0.02)	−0.01 (0.01)	0.01 (0.01)	0.02 (0.01)
6. Good Finances	−0.02 (0.03)	0.00 (0.03)	0.02 (0.04)	−0.01 (0.02)	−0.03 (0.02)	−0.02 (0.02)
7. Own Family	0.04 (0.03)	0.09 (0.03)**	0.04 (0.04)	0.02 (0.02)	−0.01 (0.01)	−0.02 (0.02)
8. Spouse Family	−0.01 (0.02)	−0.01 (0.02)	0.00 (0.02)	−0.01 (0.01)	0.00 (0.01)	0.01 (0.01)
9. Own Friend	−0.04 (0.03)	−0.02 (0.03)	0.02 (0.04)	0.02 (0.02)	0.01 (0.02)	−0.01 (0.02)
10. Spouse Friend	−0.02 (0.01)**	−0.04 (0.01)**	−0.01 (0.01)	0.01 (0.00)*	0.01 (0.01)	0.00 (0.01)
Quality
11. Good Relationship	−0.05 (0.03)*	−0.07 (0.03)**	−0.02 (0.04)	0.00 (0.02)	0.00 (0.02)	0.00 (0.02)
12. Received Support	−0.05 (0.03)	−0.04 (0.03)	0.01 (0.04)	0.03 (0.02)	0.01 (0.02)	−0.02 (0.03)
13. Gave Support	−0.03 (0.04)	−0.02 (0.03)	0.01 (0.05)	0.02 (0.02)	0.02 (0.02)	−0.01 (0.03)
14. Approve Relationship	−0.06 (0.03)	−0.06 (0.03)	0.00 (0.04)	0.00 (0.02)	0.00 (0.02)	0.00 (0.03)

Note:

^*p < .05,

^**p < .01.

Comparing Cohabitators and Non-cohabitors

Despite expectations that the networks of cohabitors and non-cohabitors would differ significantly in many regards, they generally did not. For husbands and wives, no network measures differed significantly between cohabitors and non-cohabitors at baseline. The network change analyses revealed only one significant difference: Husbands who cohabited experienced a decrease in the proportion of parents with children under 18 in the network, but husbands who did not cohabit experienced an increase. Table A2 in the online supplement details all findings.

Effects of Relationship Duration

Despite expectations that longer relationships would be associated with greater network integration and more social capital available to couples, this largely was not the case. Although husbands in longer relationships did have more spouse family at the start of the marriage and wives in longer relationships also tended to have more good relationships in the network, relationship duration had only one association with network change: Husbands in longer relationships tended to have a less positive change in spouse family. Table A3 in the online supplement shows all relationship duration results.

Discussion

What happens to couples’ social networks after they get married? Does marriage tie spouses more closely to their shared relationships (Milardo, 1982), do they withdraw from friends (Johnson & Leslie, 1982), or do their most important relationships generally persist across time (Antonucci & Akiyama, 1987; Cherlin, 2004)? To date, a lack of detailed longitudinal data on spouses’ networks has left unclear which network properties are characterized by change or stability during the early years of marriage. Addressing this gap, the current investigation drew upon multiple waves of network interviews to offer, for the first time, a detailed description of change and stability in spouses’ networks over the newlywed period.

Newlyweds’ Social Networks

At the outset of marriage, husbands and wives described relatively separate networks, with roughly 70% of each spouse’s network consisting of their own family and friends, and less than 20% consisting of each other’s family and friends. Wives included significantly more of their spouse’s family and friends in their own networks than husbands, indicating wives’ greater involvement with their husbands’ network than vice versa. This gender difference is consistent with prior research indicating that wives, typically socialized to be more relational than husbands (Parks & Barta, 2018), take on the majority of caregiving responsibilities for parents and in-laws (Lee et al., 2020). In this context, it makes sense that wives were more involved with their in-laws than husbands were, and that wives gave more support to and received more support from their networks at marriage than husbands did. What is surprising, however, is that despite the seemingly greater effort they put into these relationships, wives did not have better relationships with their network than husbands. On the contrary, wives perceived worse relationships with their network and lower rates of approval for their marriage than husbands. One possible source of these differences is that, because women are more often socialized into a role of providing emotional support and caregiving, they may be required to interact with network members who they do not like and who do not like them (Barbee et al., 1993). Thus, wives’ networks may be more critical of the marriage, and more withholding of approval, than husbands’ networks.

How Social Networks Change and Remain Stable After Marriage

When we observed how each of 14 network dimensions changed over the subsequent 18 months after their wedding, we found evidence to support each of the three competing perspectives on network change, depending on the specific dimensions in question.

What changes?

In line with predictions, husbands’ and wives’ networks included increasing proportions of their spouse’s family over the first 18 months of marriage, reflecting the role of marriage as an integrator of spouses’ social lives (Cox et al., 1997; Kalmijn, 2003; Milardo, 1982). The wedding itself may encourage spouses to interact with each other’s family (Chesser, 1980) and subsequent family gatherings may strengthen these connections. The characteristics of spouses’ connections also changed in ways that reflected more prosperity and stability: Over time spouses included more married people in their networks, husbands included more employed people, and wives included more financially stable people. Given the transition to marriage generally occurs during a period when individuals are transitioning from emerging adulthood into more stable roles (Robinson, 2015), findings are in line with our prediction that the early years of marriage would be accompanied by a greater focus on ties that are experiencing similar transitions. This change may reflect a choice, whereby spouses seek contacts who are married and stably employed in lieu of those whose lives are more in flux, or it could reflect age-related development, such that friends and family at a similar life stage as the couple are experiencing similar transitions at the same time. The fact that husbands reported increases in the proportion of divorced people in their networks supports the latter possibility, as spouses are likely not seeking out divorced people. Evaluating whether spouses are dropping and adding network members or whether the same network members are changing their marital status would clarify these issues. Finally, and again consistent with predictions (Cox et al., 1997), husbands and wives reported increases in the quality of their relationships with the network, and wives reported an increase in the proportion of the network who approved of the marriage. Some scholars (e.g., Sprecher & Felmlee, 2000) suggest this change may occur because network members wait to give full approval until they know that the relationship is committed, with marriage being perhaps the most important sign of commitment. An alternative is that newlyweds may be engaging in a process similar to that described by socioemotional selectivity theory (Carstensen, 1992), whereby, even as early as young adulthood, people are motivated to spend more time with rewarding connections and less with network members with whom they experience conflict. Marriage may kickstart that process, either because spouses have less time or because they are motivated to invest in people who support their marriage.

Even as many characteristics of spouses’ networks became stronger or more rewarding over the newlywed period, results also revealed some support for the dyadic withdrawal hypothesis (Johnson & Leslie, 1982; Williamson et al., 2013). Husbands (but not wives) experienced a significant decline in the proportion of own and spouse friends in the network, an apparent tradeoff with the increase in the proportion of their spouse’s family. Research on gender differences in friendships reveals that men’s friendships typically center around shared activities which, given the time constraints of marriage, may help explain why husbands drop friends from their network but wives do not (Aukett et al., 1988). Family relationships also play a bigger role in the well-being of men (Cable et al., 2013), so husbands may be more motivated to maintain bonds with family, even at the expense of their friendships.

What stays the same?

Amidst all of these changes, both husbands and wives remained in steady contact with their own family members, whose proportion within spouses’ networks did not change significantly over time. These results offer longitudinal support for a key tenet of the convoy model, which suggests that a core group of close network members tends to stick with people throughout life, withstanding even major life changes like marriage (Antonucci & Akiyama, 1987). Wives, in addition, did not experience a significant change in the proportion of their spouse’s friends they included in their networks. Spouses’ friends made up less than 5% of husbands’ and wives’ networks throughout the first 18 months of marriage, suggesting that the move toward integrated social networks operates more strongly on family than friends.

Moderators of Change in Spouses’ Social Networks

If the changes observed in spouses’ networks after marriage are associated with signs of commitment that accompany marriage (like living together), rather than with the mere act of marriage itself, then we should have observed different changes in spouses’ networks depending on spouses’ experiences of cohabitation, parenthood, and relationship duration prior to marriage. With rare exceptions, none of these premarital experiences moderated how couples’ networks changed across the first 18 months of marriage. Thus, despite the increasing prevalence of non-marital intimacy, such as cohabitation, in the United States (Curtin & Sutton, 2020), marriage itself continues to shape spouses’ social relationships in unique ways. One explanation for the unique role of marriage in shaping spouses’ social networks is that the wedding itself may bring spouses’ families and friends together in a way that premarital relationships do not. The wedding could be the first time that in-laws meet each other, signaling that marriages do not just bring together spouses in a lifelong commitment but that they merge networks as well (Cherlin, 2004; Chesser, 1980). An alternative possibility is that, prior to marriage, spouses may resist putting in the effort to incorporate each other’s family and friends, and network members may be hesitant to invest time, given the elevated possibility of breakup before marriage. Getting married may be a critical signal to the network that the couple is no longer temporary.

Strengths and Limitations

Our confidence in these results is enhanced by a number of strengths in the methods and design of this study. First, this is one of only a handful of studies of couples’ networks (e.g., Bost et al., 2002) to utilize a longitudinal design of more than two waves, allowing for greater sensitivity to detect change (Ployhart & MacKenzie Jr, 2015). Second, an extensive interview allowed for a detailed look at partners’ networks, circumventing biases associated with global network evaluations and mischaracterization of the network based on only a handful of global properties. Third, the sample was diverse in racial identity and SES. Couples in traditionally underrepresented groups in psychological research, such as racial minorities and those living with lower incomes, may be more affected by their ability to mobilize support from their network compared to White, affluent couples who have the resources to take on the challenges of marriage relatively independently (Huurre et al., 2007). Thus, these analyses offer insights into the social lives of more vulnerable newlywed couples.

Notwithstanding these strengths, the study was limited by a few factors. First, while the sample was diverse it was also entirely composed of mixed-gender first marriages, so the findings cannot be generalized to other relationship types. Network integration may be slower in remarriages, for example, as in-laws may be more cautious to get close to a spouse for fear of another separation. Older newlyweds may also experience less network integration, as older individuals become increasingly cautious about who they include in their social networks as they age (Carstensen, 2006). The development of social networks after marriage may also look different for same-sex couples, as same-sex partners tend to perceive less approval for their relationships from family and place more importance on their friends’ feelings about the relationship than mixed-gender partners (Blair & Pukall, 2015). Second, descriptions of change in couples’ networks may differ over longer periods. Research over longer timespans can assess whether the changes that we report here are truly linear or, perhaps more likely, flatten out over time. Third, these results cannot speak to whether spouses consciously act upon their networks or whether couples’ networks are changing due to age-related development.

Conclusion

Existing models offer competing predictions about whether couples’ social networks become closer, grow distant, or remain stable after marriage. By identifying partial support for each of these predictions in the same sample, these results highlight the need for comprehensive models that can accommodate and explain the range of changes that occur to different dimensions of couples’ social networks. Further longitudinal, multidimensional data from diverse populations can provide the necessary foundation for an integrated perspective to describe these changes, accommodate the possibility that network change differs in various populations, and capture the psychological mechanisms that may motivate spouses to spend time with certain people and not with others.