The Development of Marital Tension: Implications for Divorce among Married Couples

Abstract

Marriages are often characterized by their positive and negative features in terms of whether they elicit feelings of satisfaction and happiness or conflict and negativity. Although research has examined the development of marital happiness, less is known about the development of negativity among married couples. We examined how marital tension (i.e., feelings of tension, resentment, irritation) develops within couples over time and whether marital tension has unique implications for divorce. Specifically, we examined marital tension among husbands and wives within the same couples from the first to the sixteenth year of marriage, as well as links between marital tension and divorce. Participants included 355 couples assessed in years 1, 2, 3, 4, 7 and 16 of marriage. Multilevel models revealed that wives reported greater marital tension than husbands. Marital tension increased over time among both husbands and wives, with a greater increase among husbands. Couples were more likely to divorce when wives reported higher marital tension, a greater increase in marital tension, and greater cumulative marital tension. Findings are consistent with the emergent distress model of marriage, but indicate that despite the greater increases in marital tension among husbands, wives’ increased marital tension over the course of marriage is more consistently associated with divorce.

Marriages are often characterized by their positive and negative features in terms of whether they elicit feelings of satisfaction and happiness or conflict and negativity (Bookwala, 2016; Fincham & Linfield, 1997). It is well established in the marital literature that couples experience changes in their marital happiness over time (Birditt, Hope, Brown, & Orbuch, 2012; James, 2015; Lavner & Bradbury, 2010). There is much less known about the development of negative aspects of the marital relationship. Longitudinal studies of married individuals have been inconclusive, showing evidence of both stability over time (Kamp Dush & Taylor, 2012) as well as increased negativity (Birditt, Jackey, & Antonucci, 2009). The research to date has focused on married individuals rather than married couples. Thus, little is known about how negativity varies between husbands and wives and whether its development varies between individuals in the same marital relationship.

It is important to understand the development of negative aspects of the marital tie within married couples as negative aspects of relationships may not exhibit the same patterns of development as marital happiness. In fact, there is reason to believe that negative aspects of relationships may be more important for marital longevity (i.e., length of marriage) than happiness. The present study examined marital tension over 16 years and whether the development of marital tension for husbands and wives differentially predicted marital longevity.

Marital Tension

Positive and negative aspects of relationships are not simply opposite ends of the same continuum. People can have simultaneously positive and negative feelings for one another (Fingerman, Hay, & Birditt, 2004; Rauer, Williams, & Jensen, 2015; Rauer & Volling, 2013). Negative aspects of relationships also tend to be more highly associated with marital longevity as well as with psychological and physical well-being than do the positive aspects of relationships (Karney & Bradbury, 1995; Kurdek, 1995; Orbuch, Veroff, Hassan, & Horrocks, 2002; Umberson, Williams, Powers, Liu, & Needham, 2006). Specifically, marital tension captures overall feelings of negativity in marriage. Marital tension includes negative emotions such as feeling tense, resentful, and irritated about the marital relationship. These emotions occur as a result of irritations, disappointment, disagreements, and conflict in marriage. It is a broader construct than marital conflict, which is defined as overt arguments. Marital tension is also distinct from the particular types of conflict strategies that couples use. For example, one couple may yell and scream at one another during a marital disagreement, whereas another may calmly discuss problems yet both couples could experience high levels of marital tension.

Models of Marital Tension over Time

Developmental models of marital quality vary in terms of whether they predict stability, declines or improvements in marital quality over time. We extend these models and apply them in our examination of the development of marital tension. Drawing from the enduring dynamics, emergent distress, disillusionment, and accommodation models of marital development, we describe how marital tension may develop and change from the newlywed to the middle years of marriage. We consider each perspective and their predictions concerning the course of marriage and divorce below.

Enduring dynamics model: Stability over time

The enduring dynamics model (Caughlin, Huston, & Houts, 2000) suggests that marital quality is relatively stable over time. The model posits that certain interpersonal characteristics and patterns established during courtship are maintained throughout the course of marriage. The presumption is that partners are aware of each other’s positive and negative qualities, including shortcomings, prior to marriage and that many couples marry despite being aware of their incompatibilities (Huston, Caughlin, Houts, Smith, & George, 2001). Thus, according to this model, marital tension will be stable over time.

Emergent distress and disillusionment models: Increased tensions over time

Unlike the enduring dynamics model, the emergent distress and the disillusionment models posit that interpersonal characteristics and patterns among newlyweds are not predictive of marital outcomes (Gottman, Coan, Carrere, & Swanson, 1998; Huston & Houts, 1998). Rather, most newlyweds begin their marriage very happy and view their partner as having predominantly positive qualities with very few negative ones, but then either experience increasing emergent distress or otherwise become disillusioned with their partner over time.

The emergent distress model presumes that the ways in which couples resolve their differences are the key factors in determining marital success versus failure. Couples who do not create a constructive problem-solving context may express more negativity. After some time, couples experience a rise in negativity that corrodes their relationship (Bradbury, Cohan, & Karney, 1998). The increase in marital conflict and negative behavior contributes to marital dissatisfaction and divorce.

The disillusionment model also predicts increased marital tension but due to a different developmental mechanism. According to the disillusionment model, couples are overly happy in the early years of marriage due to idealized notions of one another and high unrealistic expectations (Huston & Houts, 1998). Couples may present themselves more positively in the early years. As couples get to know one another better and are less able to maintain a happy façade, conflict and marital tension increase over time. Couples may also have unrealistically high expectations at the beginning of the relationship, which leads to disappointment and frustration over time when faced with the stresses of daily life. Thus, according to these models, marital tension will increase with time.

The accommodation model: Decreased or curvilinear marital tension over time

Similar to the emergent distress and disillusionment models, the accommodation model presumes that initial marital patterns do not necessarily determine marital outcomes (Rusbult, Verette, Whitney, Slovik, & Lipkus, 1991). The accommodation model differs, however, in that it presumes newlyweds can experience and accommodate to early marital problems. These initial marital problems may cause difficulties and disappointments, but partners learn to adapt their expectations and accommodate to each other, which allows for a better marital relationship (Huston & Houts, 1998; Rusbult, Yovetich, & Verette, 1996). By learning to overcome marital problems, couples are able to sustain a happy marriage. According to the accommodation model, marital tension will decrease over time or increase briefly followed by a decrease (curvilinear).

Longitudinal Patterns of Negative Dimensions of Marriage

Researchers examining the longitudinal course of marriage have examined married individuals rather than couples and vary in terms of how they assess negative aspects of marriage. In addition, most studies examined overt conflict, which is a distinct construct from general underlying marital tension. Longitudinal studies of self-reported conflict revealed inconsistent findings with regard to the average change over time among married individuals. Some studies have found stability in self-reported conflict among married individuals over time (Johnson, Amoloza, & Booth, 1992; McGonagle, Kessler, & Schilling, 1992). Kamp Dush and Taylor (2012) examined individuals’ self-reports of marital conflict over 20 years (“How often do you disagree with your spouse?”). They found three marital conflict trajectories – low, medium, and high – that were mostly stable over time. In contrast, other studies of married individuals found increased self-reported, overt marital disagreements over time (Hatch & Bulcroft, 2004). James (2015) examined the development of marital conflict among wives aged 35 and older, such as arguing with husbands over chores and responsibilities, children, money, showing affection, religion, leisure, drinking, other women, his relatives, and her relatives. He identified three trajectories including a low, moderate, and high group, which varied significantly in their initial levels of conflict. All groups showed increased conflict over the first 10 years of marriage followed by declines.

Few studies examined marital tension indicators or made the distinction between overt arguments and underlying marital tension. Birditt et al. (2009) found that negative marital quality (i.e., the extent to which spouse gets on nerves and makes too many demands) increased over time. Similarly, Umberson et al. (2006) examined disagreements and marital tension over time and found increases over 8 years of marriage among married individuals. Whereas studies reported inconsistent findings on whether overt disagreement and conflict are stable or increase over time, the limited studies on marital tension-related concepts suggest that marital tension increases over time. In addition, the studies presented analyzed data from individuals who were married. We argue that this is equivalent to examining unrelated mothers and children to study the development of the mother-child attachment relationship. It is particularly important to examine husbands and wives within couples because research suggests that wives report lower marital satisfaction, greater marital tension and are more distressed by relationship problems (Almeida & Kessler, 1998). They are also more affected by their husbands’ stress, health problems, and opinions of the relationship than the reverse (Kiecolt-Glaser et al., 2005; Strawbridge, Wallhagen, & Shema, 2007; Yorgason, Roper, Sandberg, & Berg, 2012). However, studies have yet to consider whether there are variations in the development of marital tension between husbands and wives within couples.

Marital Tension and Divorce

Greater negativity (e.g., more hostility, destructive conflict strategies), in the first years of marriage predicts increased marital instability and divorce in studies examining the newlywed years through middle age (Birditt, Brown, Orbuch, & McIlvane, 2010; Matthews, Wickrama, & Conger, 1996; Stanley, Markman, & Whitton, 2002). Nevertheless, we know little about the effects of the development of marital tension on the risk of divorce. Changes in marital tension may have greater implications for marital longevity than positive aspects of marriage, such as happiness and satisfaction, because negative aspects of relationships tend to be more highly associated with well-being than positive aspects of relationships (Kiecolt-Glaser & Newton, 2001).

Theories of marital development can be used to make differential predictions regarding the development of marital tension and divorce. The enduring dynamics model suggests that high levels of marital tension in the first year of marriage will predict increased likelihood of divorce (Huston et al., 2001; Markman, Rhoades, Stanley, Ragan, & Whitton, 2010). In contrast, the emergent distress and disillusionment models suggest that increasing marital tension over time predicts increased likelihood of divorce, whereas the accommodation model predicts that the inability to reduce marital tensions by accommodating to their spouse would predict greater likelihood of divorce. These models suggest that the development of marital tension may have distinct effects on marital longevity beyond the level of marital tension in the first year of marriage. Based on these different models of marriage, marital tension needs to be examined in several different ways as marital tension in the newlywed years may be important in setting the tone of the marriage, and yet changes in marital tension may be more indicative of relationship quality and longevity. This study uses unique longitudinal couple data to examine the importance of marital tension for marital longevity (i.e., the length of marriage). We assessed different mechanisms of dissolution with regards to marital tension: (1) tension in the first year of marriage, (2) average marital tension over time, (3) marital tension in the most recent wave, (4) change in marital tension from the previous wave, and (5) cumulative marital tension. Each of the five mechanisms of dissolution focused on a different explanatory contribution of tension to the longevity of marriage.

Tension in the first year of marriage and average marital tension over time addresses the enduring dynamics model, which suggests that couples have enduring proclivities to tension or irritations that are destructive to marriage long-term. Marital tension in the most recent wave, change in marital tension from the previous wave, and cumulative marital tension scores were used to test the emergent distress and disillusionment models, which suggest that rising tensions over the course of a relationship may be responsible for divorce, as well as the accommodation model, which suggests that the inability to decrease tension would predict whether couples divorce.

Finally, there may be distinct effects of husband and wife perceptions of marital tension over time on marital longevity. Wives reports of marital tension may be more highly predictive of divorce as wives are more likely to initiate divorce than are husbands (Amato, & Previti, 2003). Further, wives are more likely to use destructive conflict strategies than are husbands which is predictive of divorce (Birditt et al., 2010).

Present Study

This study moved beyond the present literature by examining marital tension among married couples over 16 years of marriage and the implications of the development of marital tension for marital longevity. We addressed the following hypotheses:

H1) Because wives often report greater negative marital quality than husbands (Almeida & Kessler, 1998: Birditt et al., 2015) we predicted that wives would report greater marital tension than husbands.

H2) Marital tension would increase among both husbands and wives. We also predicted that there would be a greater increase in marital tension among wives than among husbands because wives tend to report greater marital tension.

H3) Consistent with the emergent distress and disillusionment models, increasing marital tension would predict marital longevity to a greater degree than marital tension in the first year of marriage.

H4) Because wives are more likely to initiate divorce and are more likely to use destructive conflict strategies in response to conflict (Amato, & Previti, 2003; Birditt et al., 2010), we predicted that wives’ reports of marital tension would be more predictive of marital longevity than would husbands’ reports of marital tension.

The present study also controlled for several sociodemographic characteristics which have been identified as predictors of marital quality and divorce, including race (Broman, 2005; Orbuch, Bauermeister, Brown, & McKinley, 2013; Orbuch & Brown, 2007), socioeconomic status (Conger & Conger, 2002; Karney & Bradbury, 2005), number of months cohabited, whether they have premarital children, and how old they were when they married (Timmer & Orbuch, 2001; Veroff, Douvan & Hatchett, 1995). Models predicting divorce also controlled for marital happiness in order to examine whether marital tension has distinct effects on marital longevity beyond marital happiness.

Method

Participants

Participants were from the Early Years of Marriage Project (EYM) which began in 1986. Couples were re-interviewed in Years 2, 3, 4, 7, and 16 (see Table 1 for sample description). The original sample of couples was selected from those who applied for a marriage license in Wayne County, Michigan from April through June, 1986. The Wave 1 sample included 373 (174 White American and 199 Black American) newlywed couples in their first marriage. On average, husbands were age 27 (SD = 3.95) and wives were age 24 (SD = 3.76). Couples were interviewed between the first four to nine months of marriage. In order to assess how the EYM sample compared to the national population, we compared the EYM sample to the General Social Survey (GSS) nationally representative data from 1980–1994. There were no differences between the EYM sample and the GSS first-marriage sample by race, income, education, likelihood of cohabitation, parental status, and employment status (Orbuch et al., 2002).

Table 1.

Descriptive Statistics on Marital Tension and Sociodemographics (n = 710)

		Year 1		Year 2		Year 3		Year 4		Year 7		Year 16
		H	W	H	W	H	W	H	W	H	W	H	W
Marital Tension	M	2.37	2.52	2.26	2.46	2.39	2.58	2.17	2.44	2.51	2.66	2.52	2.47
	SD	.76	.74	.79	.78	.70	.70	.76	.78	.61	.65	.73	.74
	n	353	353	320	320	236	236	226	226	157	157	153	153
Paired t-test	t	3.40**		3.77***		3.79***		4.38***		2.54*		−.77
Sociodemographics
Age at Marriage	M	25.98	23.87
	SD	3.95	3.76
Black	%	51.8
Household Income	M	31395
	SD	16930
Education (Years)	M	13.15	13.14
	SD	1.92	1.89
Cohabited (Months)	M	10.69
	SD	19.20
Premarital Child(ren)	%	34.1	37.7

Note. H = Husband, W = Wife. Cohabited = months of cohabitation prior to marriage,

^**p < .01

^***p < .001

Participants completed face-to-face interviews in their homes with race-matched interviewers in Years 1, 3, 7 and 16. Spouses were interviewed separately and then together as a couple. Brief telephone interviews were conducted in Years 2 and 4. The current study utilizes data from all time points: Years 1, 2, 3, 4, 7 and 16. Divorce rates increased as the study progressed from 5% in Year 2 to 41% (of the total number of the original 746 individuals) divorcing by Year 16 (see Appendix).

Attrition

The response rate varied across waves with an average of 80% of the original sample participating (range 70%–93%; calculated by dividing the total number of husbands or wives interviewed by the number eligible to participate). Attrition was due to a series of factors including refusal to participate, death, and loss of contact (See Appendix). A total of 320 married individuals (130 Black, 190 White) completed Year 16 interviews (43% of the original sample and 73% of the non divorced sample). These response rates are consistent with other longitudinal studies, such as the National Survey of Families and Households, which reports a 23% attrition rate for Blacks and a 15% rate for Whites from 1987 to 1994. Supplementary Table 1 includes the attrition and divorce rates by wave.

The analytical strategy allows for missing data. Thus, all participants who had at least one wave of marital tension and marital longevity data were included. This resulted in a total of 710 participants in 355 couples. We examined whether a series of variables assessed in Year 1, including age, race, household income, education, number of months cohabited, premarital children, marital happiness and marital tension, predicted the number of waves of marital tension completed with a linear regression in which all variables were entered together as predictors of the number of waves of these data completed. For both wives and husbands, those who were White (women: b = .76, p < .001; men: b = .59, p < .01), had more education (women: b = .21, p < .001; men: b = .17, p < .001), had higher income (women: b = .68, p < .05; men: b = .95, p < .01), and had greater marital happiness in wave 1 (women: b = .39, p < .05; men: b = .46, p < .01) completed more waves.

Measures

Marital tension

In Years 1, 2, 3, 4, 7, and 16, husbands and wives completed two items regarding marital tension: 1) how frequently in the past month they felt irritated or resentful about things their spouses did or did not do; and 2) how frequently they felt tense from fighting, arguing, or disagreeing with their spouses from 1 (never), 2 (rarely), 3 (sometimes) to 4 (often). Items were averaged to create a score in which higher scores reflected greater marital tension. Wives and husbands were calculated separately. Due to the two-item scales, Spearman-Brown coefficients were calculated to estimate reliability (Cortina, 1993). The Spearman-Brown coefficients were an average of .71 for wives (range = .66 – .77) and .65 for husbands (range = .55 – .71). Although these coefficients are low, these numbers are typical and expected when using two item scales (Birditt et al., 2009; Umberson et al., 2006).

Models assessing whether marital tension predicted marital longevity examined different variables representing different mechanisms of dissolution with regards to marital tension. Year 1 marital tension reflected the level of marital tension in the first year of marriage. The rest of the variables were time-varying. Average marital tension reflected a moving average given the particular wave in which it was calculated. For example, the average marital tension in Year 4 reflected the average of years 1, 2, 3, and 4, average marital tension in Year 7 was the average of Years 1, 2, 3, 4, and 7. Level of marital tension in the most recent wave was the level of marital tension in the most recent wave of data collection (i.e., this score reflects the most current data available for the specific year; e.g., wave 4 data were used in year 4). Difference in marital tension level from the previous wave indicated the change in the level of marital tension between the most recent wave and the previous wave. Cumulative marital tension reflected the sum of marital tension until the present wave.

Gender

Wives were coded as 1 and husbands were coded as −1.

Time

Time was coded as the year of marriage that these data were collected, which included Years 1, 2, 3, 4, 7, and 16.

Divorce and marital longevity

Participants were coded as 1 if they were divorced and 0 if they were still married to their original partner in each year of the study. Marital longevity was defined as the number of years the couple was married.

Covariates

Sociodemographic factors

A series of sociodemographic characteristics were assessed in the first wave. Age in Year 1 of marriage was defined as a participant’s age in 1986. Race was coded as 1 (Black American) and −1 (White American). For household income in 1986, participants were asked to indicate, based on a series of income ranges, what all members of their household made together in 1986 before taxes, including everything from salaries to dividends and interest. The possible income ranges were narrowly defined at the lower income ranges (e.g., none or less than $2,999, $3,000–$4,999) to broader ranges at the highest income ranges (e.g., $60,000–$74,999, $75,000 and over). The midpoints of these categories were used as the scores to approximate a continuous variable and the log was taken to reduce skew. Participants reported the highest grade in school or year of college attained in 1986 from 0 to 17+. The months of cohabitation prior to marriage variable was the number of months, if any, participants stated they lived with their partner prior to their marriage in 1986 (range 0 to 108). If couples did not cohabit, they were given a 0. For presence of premarital children, participants were given a value of 1 if they had children prior to marriage and a value of −1 if they did not. We also considered employment status (working = 1; not working = −1) and number of children in the household younger than age 18 in each wave.

Marital happiness

In Years 1, 2, 3, 4, 7 and 16, husbands and wives completed two marital happiness items: how happy would you describe your marriage (1 [not too happy] to 4 [very happy]); and how satisfied are you with your marriage (1 [very dissatisfied] to 4 [very satisfied]). Items were averaged to create a marital happiness score in which higher scores reflected greater marital happiness (Crohan & Veroff, 1989; Veroff et al., 1995). Wives and husbands scores were calculated separately. Spearman Brown coefficients ranged from .76 to .87 for wives and from .69 to .79 for husbands. We included marital happiness in the models predicting divorce and created scores for marital happiness that were parallel to marital tension mechanisms of dissolution. Thus, we created scores that represented marital happiness in Year 1, average marital happiness over time, most recent level of marital happiness, change in happiness from the previous wave, and cumulative marital happiness. The marital happiness scores were included as covariates in the respective marital tension model (e.g., marital happiness in Year 1 as a covariate in the model examining marital tension in Year 1) to show that the effects of marital tension on divorce are above and beyond positive marital quality.

Analytic Approach

To identify how marital tension changed over time, we conducted growth curve analysis using multilevel models with MIXED in SPSS. Multilevel models are ideally suited for dyadic and longitudinal data. These models account for correlated errors due to interdependency between dyad members and repeated measures of individuals over time (Kenny, Kashy, & Cook, 2006). These data can also be unbalanced with data from only one dyad member or data for fewer than the total number of time points. The models included the recommended two levels for longitudinal dyadic data, in which the lower level represents variability due to within person repeated measures for husbands and wives and the upper level represents between couple variability across husbands and wives (Bolger & Laurenceau, 2013). Separate intercepts and slopes were estimated for husbands and wives. The intercept and slope of time were allowed to vary between and within couples and all possible correlations were estimated. The model also allowed for correlated errors between husband and wives within a given wave using a heterogeneous compound symmetry error structure (CSH).

To examine whether marital tension trajectories varied by gender, we estimated a multilevel model in which marital tension was the outcome and gender and time were the predictors. Gender was entered as a factor variable and the ‘noint’ command was used to obtain separate intercepts and slopes for husbands and wives. Time was centered on the first wave, such that Year 1 received a score of 0. Models that included the quadratic effect of time were not significant and so time was estimated as a linear slope. Covariates included: age; premarital children; years of education; months of cohabitation; income; and race. We tested gender differences in intercepts and slopes of time with additional test statements in the MIXED command. Example lower level equations for husband and wife for dyad i at time t are provided below:

$${\text{Marital Tension}}_{\mathrm{H}\mathit{\text{ti}}}=b_{\mathit{0}\mathrm{H}i}+b_{\mathit{1}\mathrm{H}i}{\mathrm{T}}_{\mathit{\text{ti}}}+b_{\mathit{2}\mathrm{H}i}{(\text{Covariates})}_{\mathit{\text{ti}}}+e_{\mathrm{H}\mathit{\text{ti}}}$$

$${\text{Marital Tension}}_{\mathrm{W}\mathit{\text{ti}}}=b_{\mathit{0}\mathrm{W}i}+b_{\mathit{1}\mathrm{W}i}{\mathrm{T}}_{\mathit{\text{ti}}}+b_{\mathit{2}\mathrm{W}i}{(\text{Covariates})}_{\mathit{\text{ti}}}+e_{\mathrm{W}\mathit{\text{ti}}}$$

where b₀ is the intercept and b₁T is the slope for time. Since time is scaled at zero for the initial data collection, b₀ is the estimate of the husband or wife i’s initial marital tension score at year 1 of marriage. A positive slope represents an increase in marital tension over time and a negative slope represents a decrease in marital tension over time. The error represents husband or wife i’s marital tension score at time t that is not predicted by time or the other covariates. Upper level equations for husband are provided as an example:

$$b_{\mathit{0}\mathrm{H}i}=a_0+{\mathrm{\mu }}_{0i}$$

$$b_{\mathit{1}\mathrm{H}i}=c_0+{\mathrm{\mu }}_{1i}$$

where a₀ represents the average intercept and c₀ represents the average slope between husbands. µ_0i and µ_1i represent how much husbands vary in their tension scores at year 1 and how much husbands vary in their rate of change in tension over time.

Next we examined whether marital tension predicted marital longevity to test different models of marital dissolution. To do this analysis, discrete time survival models using the clog-log function were conducted using Stata 14 (Pevalin & Karen, 2009). This type of modeling was appropriate given that data collection times were uneven and exact time of divorce (i.e., in terms of the month of divorce) was unknown. This type of modeling also allowed for the use of time-varying predictors. To conduct these models data were structured in person-time data format. Each couple had multiple data records and each record represented a wave of the study up to and including the wave in which divorce occurred or the observation was censored (due to drop-out). Couples who remained married throughout the entire study had six data records, whereas couples who were divorced in year 4, for example, had only four data records. Each record contained a value for the binary dependent variable for divorce, which had a value of 1 if divorce had occurred since the last data collection wave or a value of 0 if the couple was still married or the couple was censored. In these models, time was measured with six binary indicator variables (w1-w6), one for each wave of the study, as recommended by Heeringa et. al (2010). Each person-time record contained all time indicator variables, but only the time indicator corresponding to the specific wave had a value of 1, with all other indicators having a value of 0. This method reduced bias of the regression parameters by adjusting for variability in size of time units given the gaps in data collection for the last two waves.

A series of models was estimated to examine links between husband and wife reports of marital tension and divorce. We tested each of the mechanisms of dissolution outlined above in five separate models with separate tension predictors for the husband and wife. Model 1 tested tension at wave 1, Model 2 tested average tension, Model 3 tested tension in the most recent wave, Model 4 tested change in tension from the previous wave, and Model 5 tested cumulative tension. Models 1 and 2 addressed the enduring dynamics model and models 3 through 5 tested the emergent distress/disillusionment model of marital dissolution. An example equation is provided below.

$$\text{log}[-\text{log}(1-h_{c,t})]={\alpha }_t+B_{\mathit{1}}{x}_{1,t}+\dots +B_p{x}_{p,t}$$

where h_c,t refers to the hazard of divorce at discrete time t for couple c, α_t = 0 (no intercept), and B₁x_1,t…B_px_p,m represents the function of the regression parameters.

The sociodemographic covariates from Wave 1 (age in Year 1 of marriage, household income in 1986, highest level of education, months of cohabitation prior to marriage, presence of premarital children) were considered time-invariant. All models controlled for marital happiness which was calculated in the same ways as the marital tension variables.

Finally, post hoc models with interactions between husband and wife marital tension (e.g., husband Time 1 tension X wife Time 1 tension) were estimated. We tested significant interactions with simple slope tests using the Preacher online utility for probing interactions in multilevel models (Preacher et. al, 2006).

Results

In this section, we present findings from the examination of marital tension over 16 years. First, the descriptive statistics for marital tension are presented. Next we present how marital tension changes over time among husbands and wives. Finally, we consider whether the development of marital tension predicts divorce.

Descriptives

Table 1 describes marital tension from Years 1 to 16. The average marital tension ranged from 2.17 to 2.66 (on a 4-point scale) over 16 years, indicating that, on average, spouses reported “rarely to sometimes” experiencing marital tensions (Table 1). We examined whether there were gender differences in each wave with paired sample t-tests presented in Table 1. The findings indicated that wives reported significantly greater marital tension than did husbands in every year with the exception of Year 16 in which there was no significant difference by gender. Table 2 presents the correlations of all the study variables.

Table 2.

Correlations Between All Variables of Interest

Variable	1	2	3	4	5	6	7	8	9	10	11	12
1. Age W
2. Age H	.64**
3. Education W	.38**	.18**
4. Education H	.28**	.19**	.52**
5. Child(ren) W	.04	.05	−.33**	−.26**
6. Child(ren) H	−.03	.07	−.31**	−.28**	.83**
7. Black	.14*	.18**	.04	−.10	.32**	.31**
8. Household Income	.08	.06	.15**	.04	−.08	−.08	−.01
9. Cohabited (months)	.32**	.24**	−.13*	−.17**	.29**	.30**	.21**	−.05
10. Marital Tension W	−.13*	−.07	−.15**	−.13*	.11*	.13*	−.05	−.00	.07
11. Marital Tension H	−.06	−.02	−.09	−.01	.12*	.08	.06	−.04	−.03	.37**
12. Marital Hap W	.12*	−.00	.19**	.19**	−.24**	−.27**	−.25**	−.01	−.10	−.39**	−.29**
13. Marital Hap H	.07	.01	.16**	.18**	−.21**	−.18**	−.22**	−.06	−.08	−.32**	−.44**	.49**

Note. H = Husband, W = Wife, Child(ren) = premarital children, Cohabited = months of cohabitation prior to marriage, Hap = happiness.

^*p < 0.05

^**p < 0.01

An unconditional multilevel model predicting tension was estimated first to examine the amount of variance in tension between and within couples over time. A total of 28.0% (z = 9.6, p < .001) of the variance was between husbands in their intercepts and 20.5% (z = 8.1, p < .001) of the variance was between wives in their intercepts. 0.1% (z = 3.6, p < .001) and 0.06% (z = 1.5, p >.05) of the variance was between husbands and wives in their slopes, respectively. A total of 23.0% (z = 21.6, p <.001) and 28.3% (z = 21.0, p <.001) of the variance was within husband and wife error variance, respectively.

Marital Tension over Time within Couples

The marital tension trajectories by gender model (Table 3) showed that wives reported greater initial tension than husbands in Year 1 of marriage and that tension increased for both husbands and wives. The test of whether the intercept varied between husbands and wives showed that the wives reported significantly higher tension in Year 1 (b = .21, SE = .04, p < .001). The test of whether the time effect varied between husbands and wives revealed that there was a greater increase in tension over time among husbands than among wives (b = −.01, SE = .004, p < .01; Figure 1). Thus, consistent with our hypothesis, wives reported greater marital tension than husbands and marital tension increased among both husbands and wives. However, contrary to our prediction, the increase in tension over time was greater among husbands than wives. The developmental pattern of increased tension is consistent with the emergent distress and disillusionment models of marriage, which suggests that couples typically begin their marriages as content and view their partner as having predominantly positive qualities with very few negative ones, but then either experience increasing emergent distress or otherwise become disillusioned with their partner over time.

Table 3.

Multilevel Models Examining Marital Tension as a Function of Time and Gender

	b (SE)
Husband intercept	2.32 (.04)***
Wife intercept	2.53 (.03)***
Husband slope	.02 (.00)***
Wife slope	.01 (.00)*
Covariates
Age	.00 (.01)
Premarital Child(ren)	.03 (.03)
Education (in years)	−.02 (.01)
Cohabit (in months)	.00 (.00)
Income	−.21 (.10)*
Black	.02 (.03)
Variance Estimates
Between husband intercept	.3096 (.03)***
Between wife intercept	.2172 (.03)***
Between husband slope	.0012 (.00)**
Between wife slope	.0007 (.00)*
Within husband residual	.2835 (.01)***
Within wife residual	.3442 (.02)***
−2 log likelihood(−2LL)	5867.2
Change in −2LL	11.1**

Note. b coefficients represent the unstandardized estimates. Change in −2LL reflects change from model with only covariates (−2LL = 5878.3).

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

Figure 1.

Wife and husband trajectories of tension

Divorce as a Function of Marital Tension

We estimated a series of discrete time survival models to test whether marital tension predicted divorce using variables representing different mechanisms of dissolution (Table 4). First, we addressed the enduring dynamics model, which suggests that couples have enduring proclivities to tension or conflict that are destructive to marriage. Model 1, examining links between the level of marital tension in Year 1 and divorce, showed that there was no association between husband or wife reports of tension in the first year of marriage and divorce. Model 2 tested the time-varying average level of tension and revealed that average level of tension was not associated with marital longevity for either husbands or wives. Thus, according to the tested models, there was no evidence of enduring dynamics as a mechanism of dissolution.

Table 4.

Discrete Time Event History Analysis Predicting Marital Longevity as a Function of Marital Tension

	Model 1	Model 2	Model 3	Model 4	Model 5
	Wave 1	Average	Most Recent Wave	Change	Cumulative
	HR (SE)	HR (SE)	HR (SE)	HR (SE)	HR (SE)
Husband Tension	1.05(.13)	.97(.17)	.98(.14)	1.00(.13)	.99(.06)
Wife Tension	1.14(.14)	1.12(.21)	1.38(.19) *	1.30(.15)*	1.12(.06)*

Note. HR = Hazard Ratio, which represents the conditional probability that divorce will occur at time t given that it has not occurred before time t. Models controlled for marital happiness (specific to the tension variable), age in Year 1 of marriage, household income in 1986, highest level of education, months of cohabitation prior to marriage, presence of premarital children. In Model 5, the cumulative marital happiness scores were too highly correlated to be included for both the husband and the wife. Thus the cumulative happiness score represented the average for the husband and the wife.

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

The next set of models tested the emergent distress and disillusionment models, which suggest that rising tensions may be responsible for divorce and the accommodation model which suggests that the inability to decrease tension would predict divorce. Model 3 indicated that wives who reported greater tension in the most recent wave showed an increased likelihood of divorce whereas, there was no association between most recent level of tension and divorce for husbands. Model 4, examining the change in tension from the previous wave, found that couples were more likely to divorce when wives showed a greater increase in tension from the previous wave whereas, there was no association between change in tension and divorce for husbands. Model 5 showed that greater wife cumulative tension was associated with an increased likelihood of divorce but there was no association between cumulative tension and marital longevity for husbands. Consistent with the emergent distress and disillusionment models and our hypotheses, these models indicate that greater recent tension, and increasing tension, predicted divorce. These models provide evidence that emergent distress or disillusionment may be the mechanisms accounting for marital dissolution. In addition, the findings support our prediction that wives’ reports of tension would be more consistently predictive of divorce than husbands’ reports of tension.

Post hoc tests

All models were estimated again testing the statistical interaction between the husband and wife tension scores predicting divorce. There was a significant interaction between husbands’ and wives’ cumulative tension predicting divorce (Figure 2). A graph of the interaction and tests of the simple slopes revealed that greater cumulative tension among wives was more highly associated with divorce when husbands had low cumulative tension (b = .24, p <.01). Wives’ cumulative tension was not associated with divorce when husbands had high cumulative tension (b = .08, p >.05).

Figure 2.

The association between wife cumulative tension and divorce by husband cumulative tension

Because employment status and number of children in the household may influence the level of tension, we re-estimated all of the models controlling for time, varying employment status of the husband and the wife, as well as the number of children in the household younger than age 18. The same developmental trajectories of tension, as well as the same links between tension and divorce emerged.

Discussion

The purpose of this study was to capitalize on unique longitudinal couple data to examine how marital tension develops over the first 16 years of marriage and to determine whether tension has implications for marital longevity beyond marital happiness. Previous studies have focused on married individuals rather than husbands and wives in couples, and on marital conflict or disagreements rather than marital tension more broadly. This study showed that tension increased over time among husbands and wives with greater increases among husbands. Furthermore, consideration of possible mechanisms of dissolution suggests that it was not tension in the first year or the average level of tension that predicts divorce but rather the recent higher levels and increased tension over time among wives that predicted an increased likelihood of divorce. Thus, the findings support the emergent distress and disillusionment models rather that the enduring dynamics models of marital dissolution. Results indicate that both husbands and wives show evidence of emergent distress or disillusionment, but that emergent distress among wives is more deleterious to marriage than tension among husbands.

The Development of Marital Tension

Capitalizing on these unique longitudinal data, we first considered the development of marital tension among husbands and wives and whether there were changes in the levels of tension over time. Husbands and wives began marriage with moderate levels of tension and showed differential change over time with husbands showing a greater increase in tension over time than wives. Indeed, wives reported greater marital tension than husbands in the early years of marriage but this difference decreased over time until there was no gender difference. This finding is fascinating given the research demonstrating that wives tend to report lower marital satisfaction, greater marital tension, and greater negative marital quality than husbands (Birditt et al., 2010; Kiecolt-Glaser & Newton, 2001). It appears that husbands ‘catch up’ to wives with regard to marital tensions over the course of marriage. Previous studies have often examined married individuals rather than married couples and thus have missed these differential developments in marital tension over time.

Previous longitudinal work on trajectories of marital disagreements among married individuals showed stability over time (Kamp Dush & Taylor, 2012) or curvilinear change over time (James, 2015). In contrast, research on negative marital quality demonstrated that marital tension increases over time (Birditt et al., 2009). The present study examined marital tension and found that it increases over time but that the increase varied in degree with husbands showing a greater increase than wives. Thus, the development of tension varies between husbands and wives within couples.

All of these findings are consistent with the emergent distress and the disillusionment models of marital development. The emergent distress model suggests that couples who are not able to manage their differences experience a rise in negativity (Bradbury et al., 1998) while the disillusionment model suggests that couples start out with idealized notions that fade over time. The findings are inconsistent with the enduring dynamics and accommodation models of marriage, which suggest either that couples are stable over time or that couples experience declines in tension as they resolve difficulties.

We considered several possible explanations for why husbands reported greater increased tension over time and showed even greater evidence of emergent distress and disillusionment than wives. It is possible that as newlyweds, wives have more realistic expectations of marriage than husbands. Alternatively, husbands may have more idealistic expectations than wives and tension may increase when those expectations are not met. It is also possible that this finding is consistent with the demand/withdraw pattern in which husbands withdraw and wives demand (Heavey, Layne, & Christensen, 1993), but only initially. Husbands may initially withdraw but become increasingly engaged over time; thus ‘catching up’ to the degree of tension reported by wives. The increased tension among husbands (but not among wives) could also be a sign of growth. For example, greater tension could indicate greater investment or commitment in the relationship, i.e. a willingness to engage in difficult dialogues to work out difficulties.

Implications of Tensions for Marital Longevity

Interestingly, the present study showed that it is not how couples start out in their first year of marriage, or their enduring level of tension (average) that determines their fate, as the enduring dynamics model would suggest. There was no association between tension in the first year of marriage or average level of tension and divorce. However, the findings indicated that couples were more likely to divorce when wives reported higher levels of tension in the most recent wave, a greater increase in tension from the previous wave and greater cumulative tension. These findings regarding the mechanisms of dissolution are consistent with our hypothesis and the emergent distress and disillusionment models of marriage, which suggest that a rise in negativity predicts greater likelihood of divorce. Interestingly, wives’ tension was more consistently associated with divorce even though husbands showed greater increased tension over time. Wives are more likely to initiate divorce and experience greater tension; thus, wives may be more frustrated by their attempts to resolve tension (Amato & Previti, 2003). Indeed, wives’ cumulative tension was associated with increased likelihood of divorce when husbands reported low tension, but not when they reported high tension. These findings indicate that high levels of tension may mean different things to husbands and wives. Husbands and wives who have different opinions about the quality of their relationship may be particularly at risk. Low cumulative tension among husbands (but not wives) may reflect a lack of investment in the relationship. Further, husbands with low tension may be unwilling to change, or believe it is unnecessary to change behaviors that are frustrating to their wives.

Together, these findings suggest that when predicting divorce, it is essential to consider the perspectives of both the husband and the wife, as well as the change in marital tension over time. This study showed it is important to consider both the husband and the wife because tension increased more among husbands than among wives, and increased tension among wives (and not increased tension among husbands) predicted divorce. This study also showed that it is important to consider change over time because it was the change in tension and not the level in Year 1, the overall average (i.e., the absolute level of tension) that predicted divorce. The findings contribute to the literature by examining marital tension beyond one point in time. It is also important to note that marital tension had implications for divorce beyond marital happiness since the above analyses controlled for marital happiness. With the unique opportunity of examining both husband and wife, and over the course of marriage from newlywed years onwards, the findings call for models of marital development to consider husbands’ and wives’ differential development and mechanisms of dissolution.

Limitations and Directions for Future Research

Although this study has many unique strengths, there are several limitations and areas for future research to consider. The marital tension measure included only two items, which resulted in somewhat low but acceptable reliability. A more comprehensive measure of marital tension would provide better reliability, as well as a broader understanding of marital tension dynamics over time. Next, these data do not include consecutive years of data from Years 4 to 16 and marital tension trajectories may have revealed different patterns if there were data for all years. It would be fascinating to have more fine grained data to examine possible links between the mechanisms of dissolution and the overall development of tension, the specific timing of the increased tension, and how they all interact with on another. In addition, marital tension trajectories may change after Year 16. The dynamics of marriages differ after children leave home, couples retire, and people develop age-related chronic illnesses (Levenson, Carstensen, & Gottman, 1993). Future research should examine marital tension over a longer time period to look beyond early and middle stages of marriage and into later adulthood as well as the specific role changes and events that may trigger increases in marital tension. It would be beneficial to understand how stressors that are external to the marriage, e.g., job stress and life events, affect marital tension over time. Future work should also consider how husband and wife perceptions of marital tension predict other aspects of marriage including marital happiness, as well as psychological and physical health. These models could be estimated with actor partner interdependence models, which allow for a more dyadic understanding of the implications of marital tension for marriage.

Overall, this study indicates that marital tension increases and varies within couples over time. In addition, the development of marital tension has important implications for marital longevity beyond marital happiness. Further research is needed to understand the factors that influence the development of marital tension and divorce. Marital tension may be expressed in ways that have yet to be examined in the literature. The present findings make it clear that it is critical to view marriage from a developmental perspective in order to more fully understand its evolution. In sum, we believe that the current research findings represent important initial insights into understanding the insidious effects of marital tension on the dynamic, evolving marital relationship.

Appendix

Attrition across Years 1 through 16 of the Early Years of Marriage Study

	Married and Completed Survey	Divorced	Refused/Deceased/ Dropped Out/Unknown
Year 1	746	--	--
Year 2	682	40	24
Year 3	559	96	91
Year 4	477	111	158
Year 7	341	167	238
Year 16	320	305	121