This study examines two overlapping longitudinal samples of U.S. couples with children, covering a period of 15 years after the first child’s birth. The first sample extended from the pregnancy with a first child until that child was 5.5 years old; the second from ages 4.5 to 14.5. Growth curve analyses revealed that marital satisfaction declined over 15 years for both husbands and wives. Attachment security measured in the second sample was associated with greater marital satisfaction, but did not buffer against declines in marital satisfaction over time. Husbands’ lower initial level of marital satisfaction measured around the first child’s transition to school was the only significant predictor of marital dissolution. The discussion emphasizes theoretical and practical implications of these findings. Many married individuals experience significant changes in their lives after they become parents, including identity changes, shifting roles in the marriage and outside the family, and changes in the relationship with their own parents. How do couple relationships fare over time after partners become parents, and what are the factors that predict the long-term marital success of these couples? Over the past 50 years, a number of researchers have proposed that marital satisfaction peaks around the time of the wedding and tends to decline from that point on (e.g., Burgess & Wallin, 1953; Vailliant & Vailliant, 1993). Nonetheless, some recent evidence suggests that when children leave home couples experience an increase in their marital satisfaction (Gorchoff, John, & Helson, 2008). The transition to parenthood is a particularly important milestone event in a marriage that provides excitement and joy, but is also often related to distress in the individual parents. Thus, the period following this transition may be a critical time for determining the health and longevity of the marital relationship. Many of the early studies of marital satisfaction relied on cross-sectional designs (e.g., Blood & Wolfe, 1960; Dentler & Pineo, 1960) and therefore provided limited information about how marriage unfolds over time. In the past decade, there has been a significant increase in the number of published longitudinal studies of marriage (see Berscheid, 1994 and Gottman & Notarius, 2002 for reviews). In spite of the growing longitudinal literature in this field, many studies suffer from methodological problems, such as failing to distinguish between childless couples and parents, and failing to analyze husbands’ and wives’ data separately (see Karney & Bradbury, 1995 for a review). Moreover, according to Karney and Bradbury (1995), over 60% of studies follow marriages for 5 years or less. Consequently, many studies of the processes that promote or impede relationship satisfaction and stability over time fail to cover a substantial part of the duration of a marriage. Other longitudinal studies of marriage that have covered longer time periods either do not measure marital satisfaction at all (e.g., Orbuch, Veroff, Hassan, & Horrocks , 2002), measure marital satisfaction at the end of the study and not throughout (Kelly & Conley, 1987), or measure marital satisfaction at infrequent intervals that do not enable a high-resolution analysis of the early years of marriage (e.g., Vaillant & Vaillant, 1993). In the present research, we examined the changes that marital relationships undergo over time, starting from the birth of the first child. To do so, we combined the data from two demographically comparable longitudinal samples spanning 6 and 10 years respectively. We collected data from the first sample from the time the parents made the transition to parenthood until their first child had made the transition to elementary school at 5.5 years of age. We collected data from the second sample between the time the couples’ first child was 4.5, prior to the transition to kindergarten, until he or she was 14.5 years old and had made the transition to high school. The two samples overlapped at ages 4.5 and 5.5, thus enabling us to cover a relatively long period of time, as well as to compare the two samples over the same stage of parenthood. Our research had three specific goals. First, we examined changes in the levels of men’s and women’s marital satisfaction over the course of 15 years following the transition to parenthood. Second, we tested whether attachment security with regard to the couple relationship (measured only in the second sample) was prospectively related to both the level of marital satisfaction and the rate of change in marital satisfaction. Third, we tested whether marital satisfaction and attachment security were significant and meaningful predictors of relationship dissolution in the second sample. The birth of a first child presents a significant challenge for married couples, as their relationship undergoes a transition from a dyadic unit to a family of three or more. This transition may affect the family system in many different ways, both positive and negative. On the positive side, parents often experience a sense of gratification and joy over having a new baby. On the negative side, they may also experience exhaustion, lack of time for themselves, and more disagreement over issues pertaining to care of the baby and the division of family labor (e.g., Belsky & Pensky, 1988; Cowan & Cowan, 2000; Twenge, Campbell, & Foster, 2003). These strains and difficulties may affect the quality of their relationship as a couple adversely. One of the earliest findings in the marital satisfaction literature is that partners’ satisfaction tends to be high around the time of the wedding, after which it begins a slow but steady decline (Burgess & Wallin, 1953; see Gottman & Notarius, 2002 and Karney & Bradbury, 1995 for reviews of subsequent research). The birth of the first child is not the only factor responsible for the decline in marital satisfaction. It is possible that some of the decline in marital satisfaction is a function of time and erosion in the relationship that may characterize childless couples as well (MacDermid, Huston, & McHale, 1990). Nevertheless, the period following childbirth is a time that merits special attention because the transition seems to introduce additional stress and strife into the couple relationship, which may accelerate the decline in marital satisfaction (e.g., Belsky & Kelly, 1994). Indeed, a recent meta-analysis reveals that although childless couples experience a decline in marital satisfaction over time, parents are significantly less satisfied than non-parents are, and number of children is reliably related to marital dissatisfaction (Twenge et al., 2003). Since the pioneering study of LeMasters (1957), research has consistently shown that the transition to parenthood poses a serious challenge if not a crisis for marriage (Belsky & Pensky, 1988; Cowan & Cowan, 1995; Cowan & Cowan, 1988; Twenge et al., 2003). Given the high rates of divorce in contemporary marriages (Schoen & Canudas-Romo, 2006), it seems imperative that we understand the key risks and buffers to marital stability. Much of the research to date has been concerned with identifying factors that moderate the quality of marriage using predominantly cross sectional and short term longitudinal studies (i.e., over the course of 1 to 2 years). Although clearly important, research has often overlooked the more fundamental question of tracing change in marital satisfaction and quality over longer durations of time (Karney & Bradbury, 1997). Moreover, studies that have tracked marital satisfaction over time have frequently employed data-analytic strategies that examine differences between the average levels of marital satisfaction at two or more time points. Studies have less often examined whether different individuals follow unique change trajectories, or specific factors that might predict such individual differences in change (Karney & Bradbury, 1997). In recent years, longitudinal studies have employed statistical procedures such as growth curve models that offer a more nuanced assessment of change (e.g., Bryk & Raudenbush, 1987; Singer & Willett, 2003). In the current study we took advantage of these procedures to examine change in marriage over a longer period of time with more frequent assessments than any other studies have covered thus far, starting from the first child’s birth and extending into adolescence. In addition to tracing the course of marital satisfaction over time, we were also interested in understanding factors that influence individual trajectories of change. Because the construct of attachment security taps the extent to which one feels comfortable and stable in close relationships (Mikulincer, Florian, Cowan, & Cowan, 2002; Shaver & Hazan, 1993), we chose to focus on attachment security as one factor that might influence marital trajectories. Most studies examining the relationship between marital satisfaction and attachment orientations have employed cross-sectional designs, and research has only sporadically attempted to examine the role of attachment security in longitudinal studies of marriage. Moreover, in spite of the vast literature on attachment and marriage, there is surprisingly little data on the role of attachment security in marital dissolution (for an exception see Crowell, Treboux, & Brockmeyer, 2009). In the current study, we were interested in examining links between parents’ attachment security and their satisfaction with marriage over a substantial number of years of marriage. Bowlby conceptualized attachment theory (1973) as a psycho-evolutionary system that guides social behavior “from the cradle to the grave” (Bowlby 1979, p. 129) and functions to maintain an optimal level of proximity to a significant other. As such, the attachment theoretical framework is particularly appropriate for the study of adult romantic relationships and for the examination of how these relationships unfold over time. Bowlby considered a person’s attachment history to influence marriage, the primary adult relationship (Bowlby, 1979). Research following this prediction revealed that, compared to adults with insecure attachment styles, individuals characterized as securely attached held more positive beliefs about romantic love and believed that romantic love can be sustained over time (Hazan & Shaver, 1987). Securely attached persons also hold more positive relationship expectations (Collins, 1996; Collins & Read, 1990), and enjoy greater relationship satisfaction (e.g., Brennan & Shaver, 1995; Collins & Read, 1990; Feeney, 1994; Feeney, Noller & Callan, 1994; Fuller & Fincham, 1995). One of the major limitations of these studies is that most measured attachment security and marital quality at the same time, a study design that cannot establish whether attachment security has a long-term impact on marital satisfaction and stability. Moreover, many of these studies examined primarily college student samples (see Bartholomew, 1994 and Noller & Feeney, 1994 for a full critique). A few studies have examined attachment orientations in the context of close relationships using longitudinal study designs. In one study of dating relationships, Kirkpatrick and Hazan (1994) found that in a 4-year period, individuals with a secure attachment style had more stable and committed relationships than those with insecure attachment styles. A 31-year longitudinal study (Klohnen & Bera, 1998) revealed similar results. Securely attached women, classified on the basis of scales created from the Adjective Check List (Klohnen & John, 1998), at age 52 had been more committed to getting married when they were 21, were more likely to be married at age 27, and were more likely to stay married and report higher relationship satisfaction at age 52. In the only study to our knowledge that examined the influence of attachment security on change in marital satisfaction, Simpson and Rholes (2002) followed pre- to post-natal changes in marital satisfaction among first time parents. This study revealed that low levels of perceived spousal support among women characterized as ambivalent were associated with significant declines in marital satisfaction for both the women and their husbands (Simpson & Rholes, 2002). Although this study examined attachment orientations and marital satisfaction around the transition to parenthood, it covered a period of only 7 months, which may not be enough to reveal the influence that attachment security exerts on marital satisfaction over time. The current study examines the influence of attachment security, measured only in the second sample, on the level of marital satisfaction and on changes in marital satisfaction over a period of 10 years starting from the couple’s first child’s transition to kindergarten. One possibility is that attachment security makes a relatively constant contribution to marital satisfaction throughout the course of marriage, but over time, secure individuals decline just as much as insecure individuals do. Another possibility is that attachment security buffers against declines in marital satisfaction, such that the differences between secure and insecure individuals become larger over time. The marital satisfaction literature suggests that one of the critical periods for the marital relationship is at midlife, when most people have young teenage children (e.g., Steinberg & Silverberg, 1987). In cross-sectional research, this period in the family life cycle is associated with the lowest point in marital satisfaction during the relationship (e.g., Orbuch, House, Mero, & Webster, 1996; White & Booth, 1991). Nonetheless, marital satisfaction does not necessarily correspond to marital stability. Some marriages are very stable even when couples express low levels of satisfaction and experience much discord (e.g., Rands, Levinger, & Mellinger, 1981). The fact that the transition to parenthood is often a time of declining marital satisfaction, but fairly high marital stability is another example of the independence of marital quality and stability (e.g., Cowan & Cowan, 2000; Karney & Bradbury, 1995). The longitudinal research on marriage has indicated that the higher the initial level of commitment or satisfaction, the more likely the couple will stay together (Burgess & Wallin, 1953; Clements, Stanley, & Markman, 2004; Levinger, Senn, & Jorgensen, 1970). A 4-year longitudinal study supports these findings by examining whether initial high levels of marital satisfaction lead to disillusionment and less satisfaction over time, or whether initially high levels of marital satisfaction predict higher levels of marital satisfaction 4 years later (Karney & Bradbury, 1997). The findings supported the latter hypothesis and indicated that couples who were more satisfied at the beginning of the study were more likely to be together 4 years later. The current study examined this trend further in order to determine whether first-time parents’ marital satisfaction measured early on is a stronger predictor of marital stability than is marital satisfaction measured closer to the time of divorce. In addition, the current study examined the influence of attachment security measured early in the relationship on marital stability over time. In spite of the vast literature on attachment orientations and close relationships, there is a paucity of research on the contribution of attachment security to marital dissolution. The current research attempts to fill this gap and examine whether individuals with secure models of attachment experience more stable couple relationships. The current study followed two samples of married couples with children to examine changes in marital satisfaction over time, beginning with the transition to parenthood and extending across the first child’s transition to elementary school (Cohort 1) and then from the transition to elementary school to high school in mid-adolescence (Cohort 2). It also sought to examine prospectively the effect of the parents’ attachment styles, measured only in Cohort 2, on their marital satisfaction from the time the child makes transitions to elementary school and into adolescence. We attempted to replicate and extend some of the trends proposed in the literature, such as the decline in marital satisfaction over time, as well as to provide some new insights into how each partner’s attachment security relates to long-term marital satisfaction. In addition, we examined how husbands’ and wives’ marital satisfaction and attachment security relates to the risk of marital dissolution. We examined change in marital satisfaction across both cohorts; however, because only Cohort 2 participants completed the attachment measure, we examined the contribution of attachment security to marital satisfaction and marital stability only on Cohort 2 couples. Based on our review of the literature, we formulated and tested the following hypotheses: Hypothesis 1: Marital satisfaction will decline over time across the combined samples for both husbands and wives. Hypothesis 2: In Cohort 2, attachment security will be related to greater levels of marital satisfaction over time, and will buffer against declines in satisfaction for both husbands and wives. Hypothesis 3: Higher levels of marital satisfaction around the first child’s transition to school (Cohort 2) will predict future marital stability. Hypothesis 4: Higher levels of attachment security measured around the first child’s transition to school (Cohort 2) will predict future marital stability. We collected data for this study from two longitudinal samples of married couples with children. Participants in Cohort 1 were couples whom we followed from their first pregnancy until their child was 5.5 years old and had made the transition to elementary school; participants in Cohort 2 were couples whose first children were 4.5 years old and about to make the transition to elementary school and were followed until the children were 14.5 years old and had made the transition to high school. To test Hypothesis 1, we combined data from the two cohorts to create a 15-year continuum of marriage starting just before the birth of the first child – which was, on average about 4 years after marriage (see Figure 1 for times of measurement and overlap between the cohorts). We described the advantages and details of this method of linking shorter-term longitudinal studies into a single study spanning a longer period of development in the results section (see also Raudenbush & Chan, 1992). Analyses for Hypotheses 2, 3 and 4 focus on the Cohort 2 sample (the parents of 4.5–14.5 year olds). Cohort 1 included 81 couples first seen in the years 1979–1982 in the last trimester of pregnancy with a first child and then followed longitudinally when their children were 6 months, 1.5 years, 3.5 years, and 5.5 years of age. These couples were part of a larger longitudinal study, the Becoming a Family Project (BAF), that focused on family formation and its relationship to marital and child development (see Cowan & Cowan, 2000). Cohort 2 included a new set of 96 couples first seen in 1990–1992 when their oldest children were in pre-kindergarten (M=4.5 years), and followed longitudinally when their children were 5.5, 6.5, 9.5, and 14.5 years of age. These couples were also part of a larger study, the Schoolchildren and Their Families Project (SAF), which focused on the family factors in children’s transition to school (see Cowan, Cowan, Ablow, Johnson, & Measelle, 2005). We selected couples from both cohorts for the present study if the partners provided marital satisfaction information on the Locke-Wallace Marital Adjustment Test (MAT; Locke & Wallace, 1959) on at least two occasions of measurement1. Of the 81 Cohort 1 couples, 1 (1%) divorced by the time the children were 1.5, 9 divorced (11%) by the time the children were 3.5, and 15 divorced (19%) by the time the children were 5.5. In Cohort 2, 2 (2%) of the original 96 families divorced by the time the children were 6.5, 7 (7%) by the time they were 9.5, and 17 (18%) by the time the children were 14.5. As discussed in the results section, we conducted all growth curve analyses of marital satisfaction with and without the divorced couples included. We reported details of how we handled divorces for survival analyses in the results section. We recruited couples for both cohorts through doctors’ offices, day care centers, preschools, and public service announcements in the media throughout the larger San Francisco Bay Area of California, in the United States. We paid couples for participation in this research In each of the two studies, research assistants informed couples that the investigators were interested in how different aspects of family life are associated with parents’ coping with the challenges of being parents and partners and children’s intellectual, social, and emotional development. Couples who we ultimately enrolled in the study did not differ significantly from responders who declined participation on a number of measures of adaptation (see Cowan, Cowan, & Heming, 2005). We described characteristics of the participants at the first time point in Table 1. To simplify this description, we present data separately for the two cohorts. As can be seen in Table 1, Cohort 1 participants were similar to Cohort 2 participants on several demographic characteristics. Both samples consisted of a largely middle-class population in their respective time eras. In Cohort 1, the Becoming a Family sample, 85% self-identified as European American (Caucasian); 4% as Asian American, 4% as Latino, and 7% as African American. In Cohort 2, the Schoolchildren and their families sample, 84% self-identified as European American (Caucasian); 6% as Asian American, 3% as Latino, and 7% as African American. In both samples, 41–43% had female children. The three major differences between the samples are in age, length of marriage, and income, with Cohort 2 being about 8 years older, married 4 years longer, and with more than twice the income of Cohort 1. Because Cohort 2 includes couples that have older children than Cohort 1, it is reasonable that this sample will be several years older with lengthier marriages on average. In addition, we recruited Cohort 2 11 years after Cohort 1. When adjusting the average income for inflation during this period, and considering the age differences between the samples, the gap in average income between the samples is not as dramatic as it initially appears. Sample Descriptions at the First Time Point of Measurement
In both samples, once couples agreed to participate, there was a relatively small attrition rate, with 89% of the original Cohort 1 couples and 96% of the original Cohort 2 couples having data on at least two measurement occasions. Missing data analyses based on levels of marital satisfaction at Time 1 revealed no significant differences in either sample between those with missing data at later time points and those with complete data. Each mother and father made a visit to our research laboratory at each time of data collection, during which we asked them to complete an interview and a set of questionnaires. Couples completed additional questionnaires at home and mailed back to the researchers. We paid families for participation. At each session, we obtained informed consent. In the initial introduction and throughout both sessions, we assured confidentiality to all participants. To assess each spouse’s level of marital satisfaction, both partners in both samples completed the Short Marital Adjustment Test (Locke & Wallace, 1959). The Locke-Wallace is a 15-item, self-report multiple-choice inventory of marital adjustment derived from the fundamental items of six marital adjustment tests used prior to 1959. Items assess marital quality, and amount of agreement about issues affecting the relationship, such as finances, recreation, affection, friends, and conflict resolution (e.g., “do you confide in your mate?”). Participants answered items on a 6-point scale ranging from (1) always agree to (6) always disagree (e.g., “Handling family finances”). Other items are forced-choice questions (e.g., “In leisure time do you usually prefer: (a) to be on the go; (b) to stay at home.”). Scoring includes arbitrary weights for each of the items, and the marital satisfaction score constitutes a sum of participants’ responses multiplied by the weight assigned to each response. Possible scores range from 2 to 158, with higher scores indicating greater marital satisfaction. Internal consistency across both cohorts, at each time of measurement, and for both husbands and wives using Cronbach’s alpha ranged from .72 to .80. We measured attachment security with regard to the couple relationship only in Cohort 2 at the second assessment (Time 2) when children were approximately 5.5 years old. To examine husbands' and wives' sense of attachment security in their relationship as a couple, we asked each partner in Cohort 2 to complete a 17-item attachment questionnaire based on Bartholomew and Horowitz's (1991) attachment inventory, that participants answered on an 8-point scale, with responses ranging from extremely uncharacteristic (1) to extremely characteristic (8). We conducted principal components analyses of both the husbands' and wives' responses. Scree plots showed a large first component and a much smaller second component. Inspection of item loadings indicated that the first unrotated component represented a continuum of overall security vs. insecurity. To form an attachment security scale, we retained items that met the following conditions: (a) the item must have a loading greater than .30 on the first unrotated component, and (b) cross-loadings on the second component must be smaller than on the first component. We calculated the relationship security subscale as the mean of 12 items that met these conditions. These included items about aspects of the relationship, such as emotional closeness to the partner (e.g., "it is easy for me to be emotionally close to my partner"), and dependence ("I am comfortable depending on my partner"). We chose to use this measure because at the time we conducted this study, this continuous measure of attachment constituted a significant improvement over the categorical Hazan and Shaver (1987) measure. In the current study, we calculated a separate score for each spouse, with higher numbers reflecting a stronger sense of attachment security with the partner. Cronbach's alpha showed acceptable internal consistency among the items for both husbands, .87, and wives, .88. Table 2 shows means and standard deviations for all substantive variables for each gender in each cohort at each time point of measurement. Figure 2 plots the marital satisfaction means from Table 2 as a function of time since the first child’s birth separately for each cohort and for husbands and wives. Means and Standard Deviations of Study Variables
Correlations among indices of marital satisfaction over time ranged from 0.25 to 0.62 for Cohort 1 husbands, 0.35 to 0.73 for Cohort 2 husbands, 0.45 to 0.66 for Cohort 1 wives, 0.53 to 0.80 for Cohort 2 wives, 0.20 to 0.71 between Cohort 1 husbands and wives, and 0.37 to 0.70 between Cohort 2 husbands and wives. These associations indicate that self-reported marital satisfaction of husbands and wives in both cohorts was consistent over time. Moreover, these associations indicate across both samples and over time that individuals who felt satisfied with their marriage tended to have more satisfied spouses. In Cohort 2, correlations between attachment at T2 (child’s age = 5.5) and marital satisfaction at various times ranged from 0.43 to 0.62 for husbands, and 0.33 to 0.56 for wives. These associations indicate that attachment security was associated with marital stability both concurrently and over time, such that attachment security measured at one time point was associated with marital satisfaction measured at later time points. Analyses first considered whether husbands’ marital satisfaction declined from the time of first pregnancy through the time the oldest child was 14.5 years old. We conducted this analysis by treating the two-cohort design as an accelerated longitudinal study, and fitting growth curve models using the SAS PROC MIXED procedure (see Littell, Henry, & Ammerman, 1998). Previous research and simulation studies have demonstrated that accelerated longitudinal designs can approximate growth curves adequately, and, as such, represent an effective method for analyzing developmental data (Bell, 1953; Duncan, Duncan, & Hops, 1996; Nesselroade & Baltes, 1979). In the accelerated longitudinal modeling technique, shorter longitudinal segments from temporally overlapping cohorts are “linked” in order to determine the existence of a single underlying growth function. Thus, for purposes of the present investigation, we combined Cohorts 1 and 2 to form one dataset covering the span of pregnancy through the time the oldest child was 14.5 years old. We then fit Growth models for husbands’ marital satisfaction at each assessment point to these data using SAS PROC MIXED with maximum likelihood estimation (MLE). MLE yields less biased estimates of variance components than alternative procedures (Pampel, 2000). At Level 1 of this model, we modeled husbands’ scores on the marital satisfaction scale at each time point (e.g., last trimester of pregnancy, child age 6 months, 1.5 years, 3.5 years, 4.5 years, and 5.5 years, 6.5 years, 9.5 years, and 14.5 years) as a function of an intercept factor and a linear slope representing time centered around the time-point shared by both cohorts (i.e., child age = 5.5), and measurement error: MARSAT = β0 + β1(TIME) + r At Level 2, we modeled both the intercept and slope as a function of a dummy code representing the sample (Cohort 1=−1 and Cohort 2=1) and a random effect (i.e., a Level-2 variance component): β0=γ00+γ01(SAMPLE)+u0β1=γ10+γ11(SAMPLE)+u1 We specified an otherwise identical model with wives’ marital satisfaction as the Level-1 dependent variable. We initially fit all models with couples who ultimately divorced included. For these couples, we modeled marital satisfaction scores at time points after divorce as missing data. We then refit models with couples who ultimately divorced dropped from the models. As there were no meaningful differences between models with and without divorced partners included, we present only the initial models with all couples included below. As shown in Table 3 and Figure 2, marital satisfaction declined significantly by approximately 2 points per year for both husbands (γ10 = −2.04, p ≤.01) and wives (γ10 = −1.84, p ≤ .01). Figure 2 illustrates how marital satisfaction declined from a level of about 120 points around the transition to parenthood (an average score for this time period according to previous research, e.g., Shapiro, Gottman, & Carrère, 2000) to a level of about 95 (a below average score) 15 years later. Level-2 effects of sample on slope revealed that the rate of decline was consistent across Cohorts 1 and 2 for wives (γ11 = 0.43, p > .05), but not for husbands (γ11 = .078, p ≤ .01). Husbands showed a faster rate of decline in Cohort 1, which covered the period from pregnancy to kindergarten, than in Cohort 2, which covered the period from pre-kindergarten to grade 9 in high school, although husbands’ marital satisfaction continued to decline in Cohort 2. Because of the limited overlap between the two samples, it was not possible to test whether this change in the rate of decline reflected a common quadratic trajectory shared by both cohorts, or a cohort difference in the rate of decline. Growth Curve Results for Marital Satisfaction Over Time (Cohorts 1 and 2 Combined)
The random effects showed significant variation in both the intercepts and the linear slopes for husbands (u0 = 226.51, p ≤ .01; u1 = 1.34, p ≤ .05) and wives (u0 = 352.10, p ≤ .01; u1 = 1.58, p ≤ .05). In other words, although marital satisfaction declined on average, there were significant individual differences in initial levels and in the rate of change. The inclusion of a subject-level predictor, such as attachment security could potentially explain this variation (as addressed in hypothesis 2). Analyses also tested growth models separately within each cohort to ensure that the results observed were not simply an artifact of the accelerated longitudinal methodology. Linear growth models fit the data well for both husbands and wives in each cohort, and all showed a significant negative linear slope. We limited analyses of the effects of attachment security with respect to the couple relationship on marital satisfaction trajectories to the N = 78 couples in Cohort 2 who completed the measure of attachment. We modeled growth curve analyses that predicted husbands’ marital satisfaction intercepts and slopes in Cohort 2 from husbands’ and wives’ attachment security in the relationship using SAS PROC MIXED with maximum likelihood estimation (MLE). In Level 1 of this model, we modeled husbands’ scores on the marital satisfaction scale at each time point as a function of an intercept factor and a linear slope representing time centered as in the previous analyses (child’s age minus 5.5), and measurement error: MARSAT = β0 + β1(TIME) + r At Level 2, we modeled both the intercept and slope as a function of the husband’s own attachment security, the partner’s (i.e., wife’s) attachment security, and a random effect: β0=γ00+γ01(OWNATT)+γ02(PARTATT)+u0β1=γ10+γ11(OWNATT)+γ12(PARTATT)+u1 We specified an otherwise identical model with wives’ marital satisfaction as the Level-1 dependent variable. As shown in Table 4, higher levels of a person’s own attachment security in the relationship were significantly associated with higher initial levels of marital satisfaction in both the model with husbands’ satisfaction (γ01 = 7.87, p ≤ .01) and the model with wives’ satisfaction as the dependent variable (γ02 = 10.65, p ≤ .01). Further, having a partner with a higher level of attachment security was associated with an additional boost in attachment security level for both husbands (γ02 = 5.05, p ≤ .01) and wives (γ01 = 4.53, p < .05). Attachment Predicting Wives’ and Husbands’ Marital Satisfaction Over Time (Cohort 2)
Analyses next examined whether the rate of decline in marital satisfaction over time depends on husbands’ or wives’ attachment security with respect to their relationship. Husbands’ rate of decline in marital satisfaction over time did not depend on their own (γ11 = −0.07, p > .05) or their wives’ (γ12 = −0.17, p > .05) initial level of attachment security in their relationship as a couple. Similarly, wives’ rate of decline in marital satisfaction over time did not depend on their own (γ12 = 0.10, p > .05) or their husbands’ (γ11 = −0.42, p > .05) security of attachment in the relationship. Thus security of attachment with regard to the couple relationship, implying that a partner expects the other to function as a secure base in times of stress, is associated with initial levels of marital satisfaction but does not appear to protect against the rate of decline in marital satisfaction over time for either husbands or wives. Analyses next tested whether either partner’s recent reports of marital satisfaction or their initial reports of marital satisfaction were related to future divorce. This analysis used a Cox proportional hazards model as a basis for performing linear regression analyses of censored survival data. We created two variables that represented (a) divorce status, and (b) time of divorce. Couples known to be divorced received a divorce score of 1 and a time score equal to the age of their child at the time of divorce. We treated couples not known to be divorced as censored data, and received a divorce score of 0 and a time score equal to their child’s age at the last time we assessed them. We then predicted divorce status by a time-invariant covariate representing initial marital satisfaction, and a time-varying covariate representing a person’s marital satisfaction as measured at the most recent time point prior to the divorce. As shown in the top of Table 5, husbands’ initial marital satisfaction, but not most recent marital satisfaction, was related to marital survival (χ2 = 5.18, p ≤ .05), even when controlling for wives’ initial marital satisfaction and both partners’ ongoing marital satisfaction. By contrast, neither wives’ initial nor most recent marital satisfaction was significantly related to divorce status. In other words, the best predictor of divorce is whether men were dissatisfied in their marriages around their first child’s transition to school – approximately 8 years after marriage. Marital satisfaction and attachment security as predictors of divorce (Cohort 2)
We examined whether attachment security would relate to marital survival in a model paralleling that described for marital satisfaction above. We found no such effect. That is, attachment security of wives and husbands with respect to their relationship, as assessed at their first child’s transition to kindergarten after about 8 years into marriage, was not significantly related to the survival of the marriage over a 10- year period (see Table 5)2. The current research followed married couples over time, starting from their transition to parenthood and continuing until the first child had made the transition to high school at 15 years of age. Both husbands and wives experienced steady declines in marital satisfaction over this time period. Individuals who felt more secure with the partner at Time 1 in Cohort 2 consistently reported higher levels of marital satisfaction. Conversely, individuals who felt less secure with the partner were less satisfied in their marriages even before the children entered elementary school, and perhaps because of this, the declining levels of marital satisfaction that occurred over the childhood and early adolescent years affected them most. Husbands’ marital satisfaction around the first child’s transition to school was the best predictor of divorce by the time the child was 15 years old. Overall, our findings on the developmental course of marital satisfaction confirm and substantially extend the previous literature indicating that marital satisfaction declines significantly over time. The current findings support this trend in a longitudinal sample that extends beyond the time frames usually investigated in longitudinal studies of marriage. Our findings suggest that by the time the first child is 15, parents’ marital satisfaction has declined on average by almost one standard deviation. Thus, although the decline in marital satisfaction may seem moderate in shorter longitudinal samples, over time this steady decline amounts to a significant change for the average couple, and a very large change for some couples, which may explain why middle-aged couples with teenage children often suffer from marital dissatisfaction and a high rate of divorce (Gottman & Levenson, 2000). Our findings also indicate that the rate of decline in marital satisfaction does not seem to taper off over time. Thus, at least across the first 15 years of parenthood, which in these two samples represents about 19 years of marriage, partners appear to experience a consistent rate of decline in satisfaction with their relationship as a couple. Some research suggests that this trend eventually reverses, and couples experience an improvement in their relationship as they age and children leave the home (Gorchoff et al., 2008). Alternatively, our survival analyses suggest that marriages that survive to old age are likely to have started out stronger than marriages that dissolved along the way. Thus, it is possible that the increase in marital satisfaction found among older couples in some cross-sectional research (e.g., Burr, 1970; Levenson, Carstensen, & Gottman, 1993; Rollins & Feldman, 1970) reflects a survival effect rather than a global trend: perhaps it is the stronger marriages that endure. Future research might consider whether marital satisfaction continues to decline at a similar rate beyond the 15 years examined in the current research. Our results support previous research that has indicated that marital satisfaction is significantly related to attachment orientations (e.g., Fuller & Fincham, 1995; Hazan & Shaver, 1987). Moreover, these new results suggest that security of attachment in relation to the spouse is associated with not only one's own but also one’s partner’s marital satisfaction. Thus, individuals who feel more securely attached to their spouses, feel that their marriage is more satisfying and have spouses who report feeling more satisfied. Marital satisfaction is comprised of many different factors, such as mutual interests, mutual values, sexual satisfaction, and communication styles (e.g., Fowers & Olson, 1989; Gottman, 1999). It seems that the sense of security one feels in a relationship is a component of relationship satisfaction, probably because one of the most basic functions of close relationships is to provide a stable and reliable sense of protection and safety in a changing and threatening world (Mikulincer, Florian, & Hirschberger, 2003). It is noteworthy that although there is conceptual overlap between the constructs of attachment security and marital satisfaction (see Mikulincer, Florian, Cowan et al., 2002), the correlations between attachment security and marital satisfaction in our samples was moderate and in keeping with the relationship between these variables in previous research. Some longitudinal research suggests that attachment security covaries with marital satisfaction over time (Crowell, Treboux, & Waters, 2002; Davila, Karney, & Bradbury, 1999). But, these studies examined young married couples without children. Moreover, in the Davila and colleagues’ (1999) study there was an inverse relationship between marital satisfaction and attachment security such that couples became more secure but less satisfied over the early years of marriage. The results of our research indicate that over the course of 10 years, attachment security is associated with more satisfying marriages, and that the two constructs, though related, are not redundant. One of the important questions concerning the relationship between marital satisfaction and attachment is whether attachment to the spouse predicts who generally tends to be more satisfied (i.e., level of marital satisfaction), or whether it also influences the extent to which marital satisfaction changes over time (e.g., rate of change in marital satisfaction). The current study indicates that individuals who have more secure relationship cognitions consistently report more satisfaction in their marriages relative to individuals who have a lower sense of attachment to their spouses. Yet, we found that attachment security did not affect the rate of decline in marital satisfaction. Nonetheless, attachment security does appear to have long-term consequences for each partner’s level of marital satisfaction when their children are preschoolers, because it may prevent the slide in marital satisfaction over the next 10 years from reaching the level of distress typical of couples who enter marital therapy (e.g., Johnson, 2004). A large body of literature suggests that one of the central functions of attachment security is to serve as a buffer against stress (Mikulincer & Florian, 1998). According to Bowlby (1988), the attachment system functions as an inner resource during encounters with stressful events, in which seeking proximity to others, or relying on internalized representations of attachment figures may alleviate distress. Experimental studies support this notion and have indicated that adults characterized as securely attached respond with reduced distress to various laboratory-induced stressful situations such as imagining separation from close others (Mikulincer, Florian, Birnbaum, & Malishkevich, 2002), and field studies have indicated that attachment security is associated with better coping with real life distress, such as having a child with a chronic illness (Berant, Mikulincer, & Florian, 2001). It seems that attachment security is a psychological resource that also enables individuals to cope more successfully with the challenges of marital life, and those who enjoy a stronger sense of attachment security also have more satisfying long-term marriages. This stress-buffering explanation receives further support from a recent study indicating that similarity in personality has a negative effect on long-term marital satisfaction (Shiota & Levenson, 2007). Thus, it is probably not similarity in attachment profiles that is contributing to marital satisfaction, but the stress inoculating properties of attachment security. Nevertheless, the current findings also indicate that attachment security does not provide a full protective shield against the longitudinal vicissitudes of couple and family life, and that even spouses who are securely attached to one another experience declines in their marital satisfaction over time. In the current study, we sought to examine the power of initial levels of marital satisfaction to predict eventual divorce, and whether change in marital satisfaction predicts divorce. The findings of the current research are in keeping with previous research (Karney & Bradbury, 1997) and indicate that levels of marital satisfaction, measured about 9 years into marriage around the first child’s transition to school, predicted which couples divorced, but later measures of marital satisfaction did not. Nevertheless, the current findings differed from Karney and Bradbury’s (1997) by indicating that only husbands’ initial level of marital satisfaction was a significant predictor of marital stability. A recent study of a large community sample of British parents corroborates our findings and indicates that the strongest predictor of relationship breakup over a period of 8 years was fathers’ initial dyadic adjustment (Stevenson-Hinde, Curley, Chicot, & Jóhannsson, 2007). Similarly, a study on the transition to parenthood found that husbands’ caregiving served as a significant buffer against new mothers’ depression and relationship dissatisfaction (Feeney, Alexander, Noller, & Hohaus, 2003). Overall, these findings attest to the significant and unique role of the father in the family system. The literature on interpersonal perception in romantic relationships has also indicated that for both men and women, a tendency to perceive the partner in a positive light is associated with relationship satisfaction, but only men’s perceptions predict relationship stability (Murray, Holmes, & Griffin, 1996; Srivastava, McGonigal, Richards, Butler, & Gross, 2006). The results of the current study suggest that high levels of marital satisfaction for husbands buffered against the effects of marital distress on marital dissolution, possibly because husbands who are high in marital satisfaction hold more positive illusions about their spouses. Recent research supports this possibility by indicating that husbands hold more positive illusions about marriage than wives do (Lin & Raghubir, 2005). The results of the current research, as well as previous findings (e.g., Karney & Bradbury, 1997; Stevenson-Hinde et al., 2007), imply that early marital dissatisfaction may potentially lead to divorce, not because of an acute problem in the relationship at a specific time point, but rather because of the cumulative, ongoing burden of marital dissatisfaction that becomes increasingly difficult to tolerate over time. Future research should focus more directly on the development of marital dissatisfaction early on in the marriage and its development towards eventual divorce. The current study represents one of the few attempts to examine the influence of attachment security on marital dissolution. Although our findings revealed a long-term effect of attachment security on marital satisfaction, and a relationship between marital satisfaction and marital dissolution, we did not find a significant effect of attachment security on marital dissolution. This finding attests to the independence of attachment security and marital satisfaction and reveals the extent to which attachment security measured early in a relationship influences a marriage over time. One possible reason for the failure of early measures of attachment security to predict marital stability is that attachment security is not constant and, according to some research, changes over time. But, unlike marital satisfaction that changes in a predictable direction, research has found that attachment security to a spouse may increase over time (Davila, Karney & Bradbury, 1999; Hazan & Hutt, 1990), decrease under adverse circumstances (Cozzarelli, Karafa, Collins, & Tagler, 2003; Davila & Cobb, 2003), or change in an unpredictable direction (Baldwin & Fehr, 1995; Davila, Burge & Hammen, 1997; Scharfe & Bartholomew, 1994). Future longitudinal research should examine the development of attachment security over time and determine whether different trajectories of change predict different marital outcomes. The fact that husbands’ initial levels of marital satisfaction, and not later measures of marital satisfaction, were significant predictors of divorce in the current sample, has important implications for intervention. First, these findings suggest that by measuring husbands’ marital satisfaction early on in the relationship, clinicians may be better able to predict which couples are at higher risk for divorce. Second, most of the literature on parenthood has traditionally focused on the psychological state of the mother. Only in the past decade and a half has research directed more attention at the role of the father in the family system (e.g., Cohn, Cowan, Cowan, & Pearson, 1992; Feeney et al., 2003; Feldman, 2000; Katz & Gottman, 1996). The current findings support the position held by proponents of ecological or family system models that have long argued that the psychological well being of the father as spouse and as parent should not be overlooked (Belsky, 1984; Cowan & Cowan, 2000). Directing intervention efforts at fathers may bolster couples at risk against growing marital unhappiness and divorce. This is not to say that wives should be ignored, as both husbands and wives experience marked declines in their relationship satisfaction, but rather that interventions aimed at strengthening family relationships would be well advised to include fathers. Most longitudinal research is limited in nature due to the practical difficulties of maintaining participants' cooperation over long periods of time, and due to increasing attrition expected in longitudinal research over time. In the current research, we employed a unique method of combining two temporally overlapping and demographically comparable cohorts that together covered a period of 15 years of marriage once parenthood had begun. Although ideally one would follow a single sample over the entire period, this possibility is often unrealistic when the question of interest requires data that span many years. Because the course of marriage often extends beyond the professional life of most researchers (Spanier & Lewis, 1980), the accelerated longitudinal design used in the current research may enable researchers to obtain a fuller picture of marital life that until now was not seen as a realistic possibility. This type of research would allow us to improve our understanding of important crossroads in a marriage such as the transition to parenthood, or the “empty nest” after children leave home. Nevertheless, the accelerated longitudinal design employed in the current research has several notable limitations. First, because we recruited our samples in different decades, they might have had a somewhat different experience as parents and partners. Second, it is unlikely that two longitudinal samples will be perfectly matched on all of the relevant variables. In the current research, there are some significant differences between cohorts such as in their level of income. Third, the samples employed in this research are not representative samples, and this limits our ability to generalize our findings. The results of our research indicating a linear decline in marital satisfaction across both samples over 15 years of marriage underscores the power of this method in spite of its limitations. It is noteworthy that although the findings of the current study are in keeping with previous research, and indicate that marital satisfaction tends to decline after the birth of the first child, it would be premature to conclude that such change characterizes all marriages of couples with young children. In fact, findings from the random effects model indicate significant variation in intercept and slopes for both husbands and wives. This variance suggests that although marital satisfaction declines on average, there are meaningful differences between couples in how satisfied they tend to be in their marriages, and in their subsequent rate of decline in marital satisfaction. Thus, the decline in marital satisfaction is not a predetermined fate. Rather, it seems that future research should consider additional variables as moderators of change in marital satisfaction after the transition to parenthood (e.g., paternal involvement, maternal depression and anxiety), to better understand which couples are at highest risk, and to identify the psychological resources that promote marital happiness even during stressful transitions. Finally, one should note that self-report methodologies cannot account for other aspects of marital quality that couples reveal in their behavior, language, and physiological arousal. Furthermore, despite the fact that the current study has identified changes in important variables in a marriage over time, we cannot draw any conclusions about the causal relationships between these variables. In spite of the above limitations, the current research clearly demonstrates that becoming a parent is one of the most challenging tasks men and women cope with during their lifetime. Our findings illustrate the price first-time parents pay in the form of declining marital satisfaction over time, but also identify protective factors and areas that may benefit from clinical intervention. The current research has only touched upon such protective factors and the picture remains incomplete. The field of relationship research needs additional research to understand the factors that contribute to a resilient and happy long-term marriage. The studies described here were funded by RO1-NIMH RO1-31109 to the last two authors. 1We asked participants to complete the MAT at each time point of measurement. Because some of the couples did not participate at all time points, we considered participation in at least two points of measurement as the minimal criteria for inclusion in the analyses. In the Becoming a Family (BAF) sample, 78 (96%) of the couples had data on at least 3 time points (more than half of the possible time points). Of these couples 28 (35%) had complete data on all time points. Nineteen (23%) couples had gaps or "recaptures" (dropped-out and then returned to the study) in their data (15 of these couples did not have data on the first measurement). In the Schoolchildren and their Families (SAF) sample, 88 (92%) of the couples had data on at least 3 time points. Of these couples 59 (62%) had complete data on all time points. Twenty-three (24%) of the couples had gaps in their data. 2We reran the survival model with both attachment security and marital satisfaction as predictors to examine whether we would obtain different results when including both attachment security and marital satisfaction in the same model. Because we measured attachment security only at time 2, this analysis was conducted on the 85 couples that made it as far as time 2. The results of this analysis reaffirmed our findings and indicated that only husbands’ initial marital satisfaction was a significant predictor of divorce (Coefficient = −0.04, SE = 0.02, χ2 = 4.23 *), even after including husband and wives’ attachment security. None of the other variables significantly predicted divorce, Husband’s attachment security: Coefficient = 0.13, SE = 0.45, χ2 = 0.08 (ns); Wife’s attachment security: Coefficient = 0.45, SE = 0.40, χ2= 1.31 (ns); Wife’s marital satisfaction: Coefficient = 0.01, SE = 0.02, χ2 = 0.26 (ns). Gilad Hirschberger, Department of Psychology, Bar-Ilan University, Berkeley. Sanjay Srivastava, Department of Psychology, University of Oregon, Berkeley. Penny Marsh, Department of Psychology, University of Washington, Berkeley. Carolyn Pape Cowan, Department of Psychology and Institute of Human Development, University of California, Berkeley. Philip A. Cowan, Department of Psychology and Institute of Human Development, University of California, Berkeley.
|
At what point of the lifespan is marital satisfaction usually the lowest for most married couples )?
Postagens relacionadas
direito autoral © 2024 mesadeestudo Inc.
