Why Do Even Satisfied Newlyweds Eventually Go on To Divorce?

Abstract

Although divorce typically follows an extended period of unhappiness that begins early in marriage, some couples who are very happy throughout the first several years of marriage will also go on to divorce. This study aims to identify risk factors early in marriage that distinguish initially satisfied couples who eventually divorce from those who remain married. We identified 136 couples reporting stably high levels of relationship satisfaction in the first 4 years of marriage. We compared the couples who went on to divorce by the 10-year follow-up with the couples who remained married, on initial measures of commitment, observed communication, stress, and personality. Divorcing couples displayed more negative communication, emotion, and social support as newlyweds compared to couples who did not divorce. No significant differences were found in the other domains, in relationship satisfaction, or in positive behaviors. Overall, results indicate that even couples who are very successful at navigating the early years of marriage can be vulnerable to later dissolution if their interpersonal exchanges are poorly regulated. We speculate that, paradoxically, the many strengths possessed by these couples may mask their potent interpersonal liabilities, posing challenges for educational interventions designed to help these couples.

Although divorce commonly occurs in the wake of low levels of satisfaction (e.g., Karney & Bradbury, 1997), highly satisfying marriages are also vulnerable to dissolution. Using two waves of survey data, Amato and Hohmann-Marriott (2007) found that approximately half of the individuals who went on to divorce in the next 6 years reported relatively high levels of marital happiness prior to the divorce and low projected likelihood of divorce. These “low-distress” divorcing couples had been married an average of 6.8 years at the first assessment, and the marital happiness reported by the husbands in these relationships was higher than that of husbands in marriages that remained intact. Although happiness may have declined at some point in the 6 years between waves, these findings nonetheless raise the question of how couples who have high levels of satisfaction after nearly seven years of marriage are still at risk of divorce within a few years. Which factors foreshadow deterioration in these relationships, and are these factors discernible early in the marriage? The present research aims to address these questions, with the goal of clarifying particularly potent antecedents of relationship distress, identifying couples at elevated risk for dissolution, and specifying key targets for educational and preventive interventions.

Drawing on social exchange theory (Levinger, 1965, 1976), Amato and Hohmann-Marriott (2007) speculated that spouses in low-distress marriages end their relationships because they have low levels of commitment and thus fewer barriers to leaving the marriage. Commitment was not measured directly in this study, but was inferred from evidence that these dissolving couples tended to be younger, more likely to have divorced parents, and hold relatively liberal and optimistic views about divorce as compared to couples who remained married. At the same time, Amato and Hohmann-Marriott observed that these characteristics could also increase couples’ levels of conflict, raising the possibility that poor communication contributed to the demise of these relationships – even while corresponding reports of satisfaction were relatively high.

Existing research lends tentative support to both predictions. Lower levels of relationship commitment are associated with lower levels of relationship satisfaction cross-sectionally (Stanley, Markman, & Whitton, 2002), and predict subsequent relationship dissatisfaction (e.g., Sacher & Fine, 1996), instability (e.g., Rhoades, Stanley, & Markman, 2010; see Le & Agnew, 2003 for review), and divorce (e.g., Impett, Beals, & Peplau, 2001). These effects are likely due to the fact that individuals with lower levels of commitment are characterized by higher levels of thinking about relationship alternatives and feeling trapped in the relationship (Stanley et al., 2002), and are more vulnerable and reactive to negative information about their partners (Arriaga, Slaughterbeck, Capezza, & Hmurovic, 2007). Further, following social exchange theory, these couples are likely to have fewer obstacles to leaving their current relationship in order to pursue alternatives, even if their level of rewards within the relationship was relatively high (i.e., high marital happiness; Amato & Hohmann-Marriott, 2007).

Diverse theoretical accounts acknowledge communication as an important factor in couples’ relationships, both as a means of building intimacy and support that maintain relationships (e.g., Reis & Patrick, 1996) and as a method of resolving relationship conflicts (Heyman, 2001). Guided by social exchange theory (Thibaut & Kelley, 1959), research in this area has argued that happy marriages can be distinguished from unhappy marriages by the ratio of positives to negatives in the relationship, and observational studies indicate that distressed couples are indeed more likely to engage in more negative communication behaviors (e.g., complaining, criticizing, blaming, denying responsibility) and fewer positive communication behaviors (e.g., agreeing, laughing, using humor, smiling) than nondistressed couples (see Bradbury & Karney, 2010 for review). Couples with higher levels of premarital negative communication have also been shown to have lower levels of marital adjustment over the first five years of marriage (Markman, Rhoades, Stanley, Ragan, & Whitton, 2010). Observed negative communication among engaged couples (Clements, Stanley, & Markman, 2004) and newlyweds (Gottman, Coan, Carrere, & Swanson, 1998) is also predictive of later divorce, although these findings have not been entirely consistent or robust (cf. Kim, Capaldi, & Crosby, 2007; Markman et al., 2010).

Given evidence for both commitment and communication as risk factors for later divorce, our goal is to test these alternate explanations for why low-distress marriages dissolve. Distinguishing between these two possibilities has important implications for understanding how relationships change and deteriorate, and for identifying appropriate targets for intervention. For example, if low-distress newlyweds who eventually divorce are characterized by low levels of commitment, spouses in these marriages may have adequate skills for relationship maintenance but insufficient inclination to deploy them, particularly when difficulties or attractive alternatives arise. Interventions targeting these concerns would need to emphasize relationship maintenance and enrichment, commitment, and the importance of stable partnerships for child development. On the other hand, if these divorces are primarily a consequence of deficient interactional processes, it would suggest that, despite adequate levels of commitment, these spouses are not resolving difficulties well, or are ineffective in communicating support and appreciation, or are interacting in ways that are insufficiently rewarding. Quality of couple communication was not examined by Amato and Hohmann-Marriott (2007), though low-distress divorcing spouses did report relatively infrequent arguments and relatively frequent interaction with their partner. Because infrequent arguments and frequent interaction may or may not be adaptive in marriage, observational data are used in the present study to assess couple communication directly. Evidence in support of this latter possibility would, in turn, raise questions about how couples manage to maintain relatively satisfied relationships for some time in the face of unrewarding behavioral exchanges, and would call for interventions focused on helping couples impose and regularly maintain ground rules for safe and nonthreatening communication.

We also expand beyond this primary focus on commitment and communication to consider two other causes of dissolution implicated in the larger literature on relationship functioning. First, longstanding intrapersonal models emphasize how spouses’ enduring characteristics and personality traits affect spouses’ reactions to each other and influence the course of marriage (e.g., Burgess & Cottrell, 1939; Terman, 1948). Indeed, traits such as neuroticism or low-self esteem, or a global propensity to be short-tempered, are known to increase risk for relationship distress and dissolution (e.g., Kelly & Conley, 1987; Kinnunen & Pulkkinen, 2003; Murray, Holmes, Griffin, Bellavia, & Rose, 2001; see Karney & Bradbury, 1995 for review). Considered in the context of low-distress newlyweds who eventually divorce, we can speculate that these relatively stable partner characteristics may operate in the background early in marriage, only to grow more salient and maladaptive as spouses settle into familiar roles and routines (e.g., Huston, Caughlin, Houts, Smith, & George, 2001). Partners’ evaluations of marriage also might be compromised by stress outside the marriage, either in the form of acute events or ongoing chronic demands (see Story & Bradbury, 2004). High levels of external stress constrain adaptive functioning in marriage by limiting partners’ ability to separate variability in their day-to-day relationship experiences from their overall levels of satisfaction (Neff & Karney, 2009), contributing to lower levels of relationship satisfaction cross-sectionally (Tesser & Beach, 1998) and over time (e.g., Bodenmann, 1997). Relatively satisfied newlyweds who go on to divorce may be those who experience more stress early in their marriages, which, in time, might heighten perceptions of relationship problems and promote harmful attributions about those problems, thus eroding positive perceptions of the partner and relationship (Neff & Karney, 2004). As with poor interactional processes, these explanations point to low levels of rewards as antecedents of distress, but raise the possibility that these can be driven by either intrapersonal or external factors.

Our review of the literature thus suggests that low-distress couples who go on to divorce may differ from low-distress couples who do not divorce in four specific ways: (1) lower levels of commitment; (2) poorer observed communication (higher negative and/or lower positive); (3) more maladaptive personality characteristics; and (4) higher levels of stress. To distinguish among these alternative explanations for why low-distress marriages ultimately end in divorce, we used repeated self-report assessments of relationship quality to identify a group of newlyweds with stable and high levels of relationship satisfaction. Within this group we (a) identify subgroups of couples who either remain married or divorce in the first ten years of marriage and (b) distinguish between these two groups using self-report data (assessing relationship commitment, personality traits, stress, and verbal aggression) and observational data (assessing communication quality during problem-solving and social support conversations). Following Amato and Hohmann-Marriott (2007), we also test whether demographic characteristics such as age, education, income, ethnicity, cohabitation prior to marriage, parental divorce, and having children during the first four years of marriage distinguish among divorcers and non-divorcers.

Method

Participants

Using criteria detailed below, the 136 couples studied here come from a sample of 172 newlywed couples identified from marriage licenses filed in Los Angeles County between May 1993 and January 1994. Marriage licenses were screened to identify couples who were married for the first time, had been married less than 6 months, were between the ages of 18 years and 35 years, and had a minimum of 10 years of education. Couples who met the criteria were sent a letter describing the project and requesting that they return a postcard if they were interested in participating. Of the 3,606 letters that were sent, 637 couples (17.8%) returned the postcard (a comparable response rate to the 18% reported by Kurdek, 1991, in a similar study), 41 letters were undeliverable (1.1%) and 2,928 couples (81.2%) did not respond. Compared with nonresponders, responders were more likely to cohabitate premaritally (43% vs. 35%, effect size r = .11), were in school longer (15.2 years vs. 14.6 years, effect size r = .18 for husbands; 15.4 years vs. 14.5 years, effect size r = .29 for wives), were older (26.6 years vs. 26.2 years, effect size r = .07, wives only), and were in higher status jobs (effect size r = .20 for husbands; effect size r = .18 for wives)¹. Interested couples were interviewed by telephone to insure that they met all inclusion criteria including the additional criteria that they had no children, were not currently expecting a child, could read and speak English, were living together, and had no plans to leave the Los Angeles area. Eligible couples were invited to participate in the project, and the first 172 who met the screening criteria and kept their scheduled laboratory appointment were included in the sample. Nearly all initial laboratory sessions took place within the first 6 months of marriage.

Husbands averaged 27.6 years of age (SD = 3.9) and 15.6 years of education (SD = 2.2), with a median income between $21,000 and $30,000. Sixty-seven percent were Caucasian, 15% were Latino-Chicano, 13% were Asian American–Pacific Islanders, and 4% were African-American. Wives averaged 26.0 years of age (SD = 3.4) and 16.2 years of education (SD = 2.0), with a median income between $11,000 and $20,000. Sixty-one percent were Caucasian, 16% were Latina-Chicana, 15% were Asian American–Pacific Islanders, and 5% were African-American. These data are consistent with the racial breakdown of Los Angeles County in the 1990 census.

Procedures

Couples participated in a 3-hour lab session within 6 months of their wedding, in which they completed questionnaires, interaction tasks, and interviews (Time 1). After 6 months, spouses completed and returned questionnaires via mail (Time 2), and 6 months later participated in a second laboratory session (Time 3). Assessments at Times 4–8 were conducted via mail at 6-month intervals. Relationship status (intact versus divorced or permanently separated) was assessed in telephone calls prior to the regular assessments and was again assessed after 10 years of marriage. Couples were paid $75 for lab sessions and $25 at each follow-up. As detailed below, marital satisfaction was assessed at Times 1–8, commitment and stress were assessed at Times 1–3, and personality traits and communication were assessed at Time 1.

Measures

Marital satisfaction.

Marital satisfaction was assessed with the widely used 15-item Marital Adjustment Test (MAT; Locke & Wallace, 1959). The MAT assesses global evaluations of the marriage, marital cohesion, degree of agreement in various domains, and retrospections on whether partners would marry their spouse again. The MAT was scored as originally suggested (without weighting), thus scores range from 2 to 158, with higher scores indicating greater satisfaction and scores below 100 indicative of marital distress. The MAT discriminates between non-distressed spouses and spouses with documented marital problems and yields a split-half reliability estimate of .90 (Locke & Wallace, 1959).

Commitment.

Relationship commitment was assessed with 12 items from the Commitment Inventory (Stanley & Markman, 1992). Spouses rated items assessing relationship commitment (e.g., “I want my marriage to stay strong no matter what rough times we encounter”; “I want to have a strong identity as a couple with my partner”) on a 7-point scale, with higher scores indicating higher commitment. Average coefficient alpha across Times 1–3 was 0.83 for husbands and 0.73 for wives.

Personality traits.

Three traits shown to increase risk for relationship distress (see Karney & Bradbury, 1995) were assessed: neuroticism, assessed with the 23-item Neuroticism scale of the Eysenck Personality Questionnaire (Eysenck & Eysenck, 1978; sample items: “Are you a worrier?” “Does your mood go up and down often?”); spouses’ general tendency to be angry, without specific reference to the marriage, assessed with the 25-item Multidimensional Anger Inventory (Siegel, 1986; e.g., “It is easy to make me angry” “I am secretly quite critical of others”); and self-esteem, assessed with the 10-item Rosenberg Self-Esteem Scale (Rosenberg, 1965; e.g., “On the whole I am satisfied with myself ” “At times I think I’m no good at all,” reverse scored). Coefficient alpha ranged from .83 to .87 across measures in the present sample, with very similar values for men and women.

Acute stress.

Spouses completed a 192-item version of the Life Experience Survey (LES; Sarason, Johnson, & Siegel, 1978), supplemented with events from other checklists that would be appropriate for young couples. For events occurring in the past 6 months, spouses were asked to indicate the impact it had on their lives (−3 = extremely negative, +3 = extremely positive). Total acute stress was calculated for each spouse as a sum of the negative impact ratings; events that might be a consequence of marital distress (e.g., relationship with spouse worsened a lot) were excluded.

Chronic stress.

Following Hammen et al. (1987), at Times 1 and 3, spouses were interviewed about the quality of nine life domains over the prior 6 months: the marriage, family relationships, in-laws, friends, school and/or work, finances, own health, and spouse’s health. For each domain, interviewers probed for concrete indicators of the ongoing stressors that the spouse may have been experiencing. After describing each domain, spouses rated their experiences in that domain on a 9-point scale (1 = very positive circumstances; 9 = very stressful circumstances). At Time 2, only the ratings were completed, via mail. Ratings from the eight nonmarital domains were averaged to form a score indicating the overall level of nonmarital chronic stress experienced by each spouse. To assess the validity of spouses’ self-ratings of chronic stress, the interviewers were also asked to make ratings of the chronic stress experienced by spouses in each domain using the same scales that the spouses used; these mean ratings were significantly correlated with husbands’ and wives’ ratings of their own stress at Time 1 (for husbands, r = .54, p < .01; for wives, r = .65, p < .01) and Time 3 (for husbands, r = .51, p < .01; for wives, r = .64, p < .01).

Verbal aggression.

Verbal aggression (e.g., insulting, threatening, saying something to spite the other) in the past year was assessed using the six-item Verbal Aggression subscale of the Conflict Tactics Scale (Straus, 1979). Each item was rated on a 3-point scale (0 = never, 1 = once, and 2 = twice or more). Reliability was adequate for husbands and wives (alpha = .71 and .75, respectively).

Problem-solving skills.

In Time 1 lab sessions, each spouse identified a source of tension in the relationship that he or she would discuss with the partner for 10 minutes. The Kategoriensystem für Partnerschaftliche Interaktion (KPI; Hahlweg et al., 1984) was used to code the skills that spouses displayed in these interactions. The KPI consists of 27 codes derived from the research literature on communication skills training and behavioral marital therapy (see Hahlweg et al., 1984). It is used to code each speaking turn in an interaction and distinguishes reliably between distressed and non-distressed couples (e.g., Hahlweg et al., 1984). On the basis of factor analyses (e.g., Hills & Johnson, 2000), the skill codes in the present study were reduced to a positive skills variable (the sum of direct expression of feelings, direct expression of wishes and needs, direct expression of attitudes and opinions, constructive solution, compromise, paraphrasing, interested questions, feedback, understanding the partner, direct agreement, accepting responsibility, agreeing, problem description, relevant questions, setting the course of the conversation and clarification, and theme suggestion) and a negative skills variable (the sum of specific critique, devaluation of partner, justification of own behavior, denial of responsibility, demands, apparent suggestions for solution, disagreement, yes–but sentences, declining remarks, and blocking off). Composite variables were calculated by summing the scores on each index across the husband and wife problem topics and, given interdependence between partners, by summing across partners (after first standardizing within sex); these scores represent couple-level variables. Interobserver agreement was adequate: ICC = .84 (husbands’ positive skills), .62 (husbands’ negative skills), .90 (wives’ positive skills), and .75 (wives’ negative skills), all p < .01.

Problem-solving affect.

Displays of specific emotions during the problem-solving discussions were coded using the Specific Affect Coding System (Gottman, 1994), which instructs coders to emphasize facial expressions, posture, gestures, and voice tone and pitch when coding affect; verbal content was insufficient by itself for coding a specific affect. Discussions were divided into 5-second units to allow for the possibility that multiple emotions could occur in a speaking turn. Each 5-second unit was coded as displaying one of five negative affects (anger, contempt, whining, sadness, anxiety), one of three positive affects (humor, affection, interest), or neutral affect. We dropped whining, sadness, and anxiety because reliabilities or frequencies were too low. Following Johnson’s (2002) factor analysis, we reduced the remaining codes to a positive composite (the sum of humor, affection, and interest) and a negative composite (the sum of anger and contempt). We created total scores on the indices by summing across the two conversations and, given high correlations between partners (r’s = 0.76 for positive affect and 0.62 for negative affect), summing across partners (after first standardizing within sex); the scores therefore represent couple-level variables. Interobserver reliability was adequate: intraclass correlation (ICC) = .83 (husbands’ positive affect), .66 (husbands’ negative affect), .68 (wives’ positive affect), and .91 (wives’ negative affect), all p < .01.

Support behavior.

Spouses engaged in two 10-minute conversations, in each of which one randomly selected partner was instructed to “talk about something you would like to change about yourself” that was not a source of tension in the marriage. The other partner was instructed to respond as he or she normally would if this topic came up between them, and the supportiveness of this partner’s behaviors was coded with the Social Support Interaction Coding System (SSICS; Pasch, Harris, Sullivan, & Bradbury, 2004). Trained observers assigned a code for each speaking turn for the spouse who had chosen the topic (the “helpee”) and the spouse who was responding (the “helper”). Each speaking turn by the helpers was rated as either positive or negative (e.g., criticizes or blames spouse; minimizes or maximizes the scope of the problem), and positive behavior was further delineated as positive emotional (e.g., statement that reassures consoles, or provides genuine encouragement to spouse; conveys that helpee is loved and cared for), positive instrumental (e.g., makes specific suggestions; gives helpful advice or access to information regarding the problem), or positive other (e.g., general analysis of the problem; see Pasch & Bradbury, 1998). A summary positive helper code was created to simplify analyses by summing the three positive codes. As with the other behavioral variables, we summed scores across partners (after first standardizing within sex); the scores therefore represent couple-level variables. Intraclass correlations indicate adequate interobserver reliability (.80 and .86 for helpers’ negative and positive support, respectively).

Data reduction

To facilitate comparisons, we standardized scores on the above measures, within sex. As we had no specific predictions for the three personality measures, and in view of high correlations among them (mean r = 0.40 for husbands and 0.53 for wives), we created a negative personality index by summing participants’ standard scores on each of the measures (after first reverse coding self-esteem so that a positive z-score represented low self-esteem), which we again standardized within sex. Additionally, given considerable stability in the commitment, acute stress, and chronic stress measures over time (cross-time correlations ranged from .31 to .70, median correlation = .54, all p < .01) and to reduce the number of overall comparisons, we created average measures of commitment, acute stress, and chronic stress across Times 1–3. The pattern of results described below did not change when these averages were used instead of individual time points.

Selecting stably satisfied newlyweds

We used semiparametric group-based mixed modeling (Nagin, 1999) to identify newlyweds with stable, satisfied trajectories over the first four years of marriage. This group-based approach assumes the population consists of a number of groups with different trajectories and seeks to identify them. Models were estimated using SAS Proc Traj (Jones, Nagin, & Roeder, 2001). This procedure accommodates missing data (missing data are assumed to be missing at random), and does not require equal assessment intervals or identical assessment procedures across participants; thus, all participants were included. Trajectories were estimated separately for husbands and wives, and models were estimated with intercept, linear, and quadratic coefficients, which we removed when analyses indicated they were not significant for particular groups.

We determined the number of groups that best fit the data by evaluating models with more groups and evaluating fit using the Bayesian Information Criterion (BIC), with greater (less negative) values indicating better fit. Analyses indicated that a five-group model provided optimal fit for husbands and wives (for details of these analyses, see Lavner & Bradbury, 2010), revealing three groups of spouses with very high levels of satisfaction and small, if any, declines over four years [mean across the eight time points was 126.66 (SD = 11.74) for husbands and 128.73 (SD = 11.78) for wives], and two groups characterized by low initial satisfaction scores and/or rapid linear declines in satisfaction. Overall, the groups could be distinguished by their intercepts and slopes, and by a host of initial characteristics (i.e., personality traits, interactional patterns, and external stress; see Lavner & Bradbury, 2010). Only couples in which both partners were in the first three groups (the “satisfied” groups; N = 136) were included in the analyses conducted here (see Table 1 for intercepts and slopes of these groups). As expected, the two groups characterized by low initial satisfaction scores and/or rapid linear declines in satisfaction had high rates of divorce (ten-year divorce rates ranged between 40–60%) and had high initial levels of maladaptive personality traits, negative interaction patterns, and stress, and low levels of positivity (see Lavner & Bradbury, 2010, for details).

Table 1.

Four-year satisfaction intercepts and slopes for the satisfied groups

Group	n	Intercept	Linear Slope
Husbands
Satisfied group 1	22	143.91	--
Satisfied group 2	70	135.22	−0.44
Satisfied group 3	44	117.20	−0.11
Wives
Satisfied group 1	40	142.76	--
Satisfied group 2	67	133.52	−0.22
Satisfied group 3	29	122.34	−0.29

Note: All parameter estimates significant at p < .01.

Results

Correlations among variables at Time 1 show reasonable independence in the four main sets of variables used to compare intact and dissolved marriages (see Table 2). Although there was some overlap among verbal aggression, the personality composite, and the stress measures (median r ~ .25), the commitment and communication variables were unrelated.

Table 2.

Within-sex correlations between commitment, communication, personality, and stress variables among satisfied spouses (N = 272 spouses)

	1	2	3	4	5	6	7	8	9	10	11
Husbands
1. Commitment	—
2. Positive affect	0.08	—
3. Positive skills	0.09	0.02	—
4. Positive support	0.11	0.15	0.01	—
5. Negative affect	0.08	−0.18*	−0.39**	−0.03	—
6. Negative skills	0.01	−0.28**	−0.38**	−0.14	0.59**	—
7. Negative support	−0.08	−0.17*	−0.23**	−0.35**	0.49**	0.43**	—
8. Verbal aggression	0.08	−0.11	−0.03	−0.03	0.22**	0.26**	0.20*	—
9. Personality composite	−0.20*	−0.08	−0.06	−0.01	0.09	0.19*	0.18*	0.30**	—
10. Acute stress	−0.16	−0.07	0.02	−0.01	−0.02	0.09	−0.01	0.27**	0.29**	—
11. Chronic stress	−0.25**	−0.02	−0.15	−0.04	0.08	0.08	0.03	0.17*	0.29**	0.33**	—
Wives
1. Commitment	—
2. Positive affect	0.17*	—
3. Positive skills	0.13	—	—
4. Positive support	0.11	—	—	—
5. Negative affect	0.04	—	—	—	—
6. Negative skills	−0.02	—	—	—	—	—
7. Negative support	−0.03	—	—	—	—	—	—
8. Verbal aggression	−0.16	−0.23**	0.02	−0.08	0.24**	0.28**	0.17*	—
9. Personality composite	−0.27**	−0.14	−0.10	−0.07	0.04	0.13	0.15	0.34**	—
10. Acute stress	−0.04	−0.06	−0.09	−0.01	0.10	0.12	0.11	0.25**	0.22**	—
11. Chronic stress	−0.24**	0.00	0.01	0.01	0.07	−0.01	0.01	0.24**	0.43**	0.20*	—

^*p < .05.

^**p < .01.

Notes: Commitment, acute stress, and chronic stress represent averages across Times 1–3; all other variables assessed at Time 1. The personality composite was formed by normalizing and combining individuals’ scores on neuroticism, trait anger, and self-esteem. Communication variables are computed at the dyadic level, so the correlations between these variables are the same for husbands and wives and are thus reported only once, for husbands.

We began by identifying the percentage of initially satisfied spouses who remained married at ten years. Fifteen percent of couples (n = 21) divorced over ten years, with three of these divorces occurring before the fourth year of marriage². Rates of divorce varied between the three groups (ranging from 9% to 21%), but not significantly so (p > .10). Overall, then, consistent with Amato and Hohmann-Marriott’s (2007) finding, divorce occurred even among spouses who were highly satisfied early in their marriages. To ensure that dissolution was not simply a reflection of low satisfaction, we compared overall levels of satisfaction between divorcers and non-divorcers. No differences were found for husbands or for wives, and follow-up analyses comparing satisfaction at each time point found only a few small differences that followed no clear pattern; 14 of the 16 comparisons were nonsignificant, p > .05. Thus, for men and women, low-distress newlyweds who either remained married or divorced at ten-year follow-up did not differ consistently in their satisfaction over the first four years of their marriages.

We then turned to demographic differences between the couples who remained married and those who divorced. Consistent with Amato and Hohmann-Marriott (2007), satisfied couples who went on to divorce tended to be younger: the mean ages at Time 1 for husbands and wives who divorced were 25.48 (SD = 3.31) and 24.90 (SD = 3.06), respectively, compared to mean ages of 27.74 (SD = 3.69) for husbands and 26.43 (SD = 3.06) for wives who remained married, p = .01 for husbands and p = .06 for wives. These couples were also characterized by husbands with lower incomes (t = 2.95, p < .01) and by a greater proportion of husbands whose parents divorced [35% parental divorce in the eventual divorce group, versus 14% in the non-divorce group, χ²(1, N = 135) = 5.36, p = .03]. No differences were found, however, with regard to husbands’ or wives’ ethnicity or levels of education; wives’ incomes or rates of parental divorce; or in rates of premarital cohabitation (all p > .10). Lastly, we examined whether couples differed in whether they became parents during the first four years of marriage or in the timing of children’s births. Of the fifty-two couples (38%) that became parents over the first four years of marriage, divorcers and non-divorcers did not differ in whether they became parents, or in the timing of their children’s births (all p > .10).

We then examined differences between couples who remained married and those who divorced in the four domains of interest. No differences were found between groups for husbands’ or wives’ average commitment over three assessments (approximately the first one-and-a-half years of marriage; Table 3 shows standardized group means, t values, and Cohen’s d for all variables). Next we examined differences in interpersonal behaviors. No differences were found in observed joint positive affect, skills, or support. However, observed negative affect, skills, and support, along with husbands’ self-reported verbal aggression, all distinguished couples who did and did not divorce. In all cases, these forms of negative interaction were higher for couples who went on to divorce³. Personality, average acute stress, and average chronic stress did not reliably differentiate between groups (p > .05, see Table 3), though a marginally significant difference was found such that satisfied newlywed couples who went on to divorce were characterized by wives with more negative personalities and higher average acute stress than couples who remained married.

Table 3.

Initial and across time differences between couples who do and do not go on to divorce (N = 136 couples)

	Married (n = 115)			Divorced (n = 21)				Cohen’s
	n	M	SD	n	M	SD	t	d
Commitment
Average commitment (H)	115	0.20	0.87	21	0.31	0.64	−0.53	−0.13
Average commitment (W)	115	0.19	0.80	21	0.33	0.78	−0.73	−0.17
Interactional processes
T1 positive affect (J)	115	0.18	1.07	21	−0.10	1.01	1.11	0.27
T1 positive skills (J)	115	0.10	1.03	21	−0.17	0.79	1.14	0.27
T1 positive support (J)	115	0.01	0.91	21	0.07	1.31	−0.21	−0.05
T1 negative affect (J)	115	−0.21	0.83	21	0.21	1.26	−1.97*	−0.47
T1 negative skills (J)	115	−0.26	0.78	21	0.36	1.03	−3.18**	−0.76
T1 negative support (J)	115	−0.23	0.70	21	0.28	1.03	−2.16*	−0.52
T1 verbal aggression (H)	114	−0.21	0.90	20	0.43	1.16	−2.82**	−0.69
T1 verbal aggression (W)	114	−0.23	0.92	19	0.01	0.96	−1.07	−0.27
Personality
Negative composite (H)	107	−0.12	0.97	20	0.03	0.94	−0.62	−0.15
Negative composite (W)	112	−0.21	0.93	18	0.26	1.26	−1.89+	−0.48
Stress
Average acute stress (H)	115	−0.04	1.05	21	−0.04	0.72	0.02	0.00
Average acute stress (W)	115	−0.17	0.98	21	0.28	1.15	−1.87+	−0.45
Average chronic stress (H)	115	−0.07	1.03	21	−0.21	0.67	0.60	0.14
Average chronic stress (W)	115	−0.20	0.94	21	−0.09	0.86	−0.50	−0.12

⁺p < .10.

^*p < .05.

^**p < .01.

Notes: H = husbands; W = wives; J = joint. Group means represent z-scores normed by sex. Commitment, acute stress, and chronic stress are averaged across Times 1–3.

Lastly, we conducted a series of binary logistic regressions in which we examined each of the negative behavioral variables in tandem with the three other domains of interest (separately) to ensure that the effects held (e.g., that negative affect was still significant when controlling for personality; that negative support was significant when controlling for commitment; that negative skills was significant when controlling for acute and chronic stress). In seven of out of nine cases, the negative behavioral variables remained significant (p < .05), and were marginally significant (p < .10) in the other two cases (negative affect controlling for commitment and negative affect controlling for acute and chronic stress). Thus, the results identified above appear robust.

Discussion

Amato and Hohmann-Marriott’s (2007) discovery that stably happy couples who went on to divorce in the next several years were younger, more likely to have divorced parents, and hold relatively liberal and optimistic views about divorce, led them to speculate that spouses dissolving their relationship had lower levels of initial commitment than their counterparts who remained married. Using three assessments of commitment for husbands and for wives that extended into the second year of marriage, we tested this explanation and found no evidence that low initial commitment characterized low-distress newlyweds who went on to dissolve their marriages, despite replicating their finding that spouses who divorced tended to be younger and had higher rates of parental divorce (among husbands). Three other factors hypothesized to promote relationship distress – personality traits reflecting hostility and negative affectivity, acute life events, and chronic stress – also failed to consistently distinguish between low-distress newlyweds who did and did not end their marriages at ten-year follow-up, as did self-reported relationship satisfaction over the first four years of marriage.

These groups of couples did differ, however, on another factor mentioned by Amato and Hohmann-Marriott and widely acknowledged as a powerful force in relationships: low-distress marriages that eventuated in divorce were characterized by the display of more anger and contempt and by more negative skills (e.g., disagreement, blame, invalidation) during laboratory-based discussions of important relationship difficulties. Moreover, when discussing a personal issue that their partner wanted to change, newlyweds in low-distress relationships that eventually ended displayed support that was more negative (e.g., expressing inappropriate pessimism, discouraging expression of feelings, insisting that partners resolve it on their own) compared to otherwise similar couples who remained married. Low-distress husbands who went on to divorce also reported being more verbally aggressive early in their relationship than those who remained married. Interestingly, these levels of negativity resembled those found in the less satisfied groups (see Lavner & Bradbury, 2010), suggesting that the negative interaction patterns of these satisfied couples who eventually went on to divorce are more behaviorally similar to the low marital satisfaction groups than the high marital satisfaction groups to which they belong. Although this pattern of results highlights the power of various forms of interpersonal communication to disrupt relatively satisfied marriages, it does not extend to positive behaviors, whether assessed during problem-solving (e.g., compromising, directly expressing feelings, exhibiting humor or affection) or social-support (e.g., reassuring the spouse) conversations.

Before turning to the implications of these results, it is important to note that our sample of couples is small and not nationally representative. The intensive longitudinal data on relationship satisfaction did permit us to identify a group of couples that consistently and repeatedly indicated their high levels of happiness with the relationship over the first four years of marriage, and the observational data that proved useful here are often not possible in larger nationally-representative surveys. Nevertheless, our sample size provided limited power to detect small effects, it is likely that our sampling approach limited the range of the variables that we studied (e.g., levels of chronic and acute stress), and our focus on childless newlyweds in their first marriages severely limits the conclusions we can draw about older couples, previously married couples, gay and lesbian couples, and couples with children before marriage. It is also possible that additional differences would have emerged had the couples been studied later in marriage (as was the case in Amato and Hohmann-Marriott’s study), or for longer durations of time; all of the dynamic predictors included here were assessed in the first 18 months after marriage, raising the possibility that additional differences may have emerged later on [e.g., the accumulation of life stress interacting with couples’ poor communication skills (Karney & Bradbury, 1995)]. Thus, to fully understand why these satisfied newlyweds went on to divorce, we need a more complete longitudinal picture of the other factors in their lives, though we note that any differences that do emerge in this case might be better conceptualized as contributing factors rather than the early antecedents of relationship satisfaction. Lastly, we caution that the trajectory of relationship satisfaction was unknown for couples between years four and ten. This six-year gap was similar to Amato and Hohmann-Marriott’s (2007) sample, but leaves unanswered questions about whether these satisfied newlyweds remained satisfied until shortly before their divorce, or whether they began to experience a gradual decline in the trajectory of their relationship satisfaction sometime after year 4. Additional research is needed to address this limitation of our work, with recognition of the possibility that newly emerging predictors (e.g., life stress, depression, extramarital affairs) might be a reflection of communication deficits already in place.

As efforts are made to examine causes of relationship dissolution among a more diverse set of low-distress couples over time, the present findings do highlight negative exchanges as one potentially consequential contributor. Perhaps the most important implication of these findings is not that low-distress couples with mismanaged and poorly regulated negative communication are at elevated risk for eventual dissolution, but that couples characterized by this form of communication can also report high levels of satisfaction for a long span of time. How can it be that spouses who appear to be functioning so well in one respect are interacting in ways that eventually compromise their relationship, to the point where they decide to end it?

One possibility is that these spouses’ reports of satisfaction are invalid or otherwise disconnected from their experiences in the relationship. Indeed, the fact that satisfaction in the first four years of marriage does not discriminate between those low-distress couples who do and do not divorce lends support to this possibility. We cannot test this idea directly, and future studies are needed to explore the conditions under which specific relationship experiences engage and affect judgments of satisfaction (McNulty & Karney, 2001). Another possibility is that some low-distress couples are so globally satisfied with their relationship that they engage in relatively high levels of negative communication, confident but ultimately mistaken in the belief that their relationship can withstand these experiences.

A third possibility, which may well make the second possibility more likely, is that low-distress couples who eventually divorce have many assets and relatively few liabilities, thus limiting or slowing the rate at which these liabilities exert their effect. That is, our data suggest that although these couples resemble less satisfied couples with regard to their negative communication, they are largely indistinguishable from highly satisfied, enduring couples in their personality traits, chronic and acute stress, and positive communication behaviors, all of which can stabilize and benefit a marriage. In the absence of high levels of negative traits and stress, low-distress spouses may be able to avoid, ‘compartmentalize,’ or rationalize the negative exchanges that do occur in their relationship. These factors might mask or buffer any effects of poor communication on satisfaction until the potent liability that the low-distress divorcing couples do possess – their relatively high level of negative communication across different interactional contexts – eventually proves to be detrimental. How this process unfolds is an open question: negative communication may prove problematic on its own, perhaps because these couples are contending with more severe relationship issues (e.g., Sanford, 2003), or in tandem with contextual factors such as increases in stress that further impair couples’ adaptive capacities (e.g., Story & Repetti, 2006; Neff & Karney, 2009) and ultimately render them vulnerable to dissolution. We note that there was suggestive evidence that wives’ negative personalities and acute stress also distinguished couples who did and did not go on to divorce: perhaps these characteristics, coupled with the negative communication, ultimately served to overwhelm these couples’ capacity to cope, either because the women themselves were more likely to initiate divorce (e.g., Amato & Previti, 2003) or because their husbands respond negatively to their stress (e.g., Neff & Karney, 2005). Future research should further investigate the reason for this gendered effect.

These findings raise important implications for couples’ relationship education (CRE) programs designed to enhance marital functioning. Negative communication has long been emphasized as a target of change in preventive interventions for couples (see Halford, 2011), and the present findings underscore this focus. Indeed, we find it significant that a one-time observational assessment of couple communication in an artificial setting outperforms reported commitment, personality, and stress in discriminating intact and dissolving marriages. At the same time, our findings suggest that the effects of negative communication may emerge only gradually, particularly when spouses’ personalities and circumstances remain sanguine; the negative interactional processes that will eventually erode the relationship exist alongside a host of other relationship-sustaining influences (e.g., low levels of chronic and acute stress, high levels of commitment) that might offset the adverse effects of poor communication. Thus, it may only be several years into marriage – when additional stress emerges, or when fundamental disagreements about life values boil over – that negative communication exerts its impact. If valid, this view suggests that any communication changes produced by CRE might need to be maintained for several years post-treatment in order to buffer couples. At present, however, few studies test the maintenance of CRE effects on couple satisfaction or communication for more than one year (for meta-analyses, see Hawkins, Blanchard, Baldwin, & Fawcett, 2008; Blanchard, Hawkins, Baldwin, & Fawcett, 2009), and the evidence that does exist suggests that low-risk couples such as these may lose the benefits gained from CRE around this time (Halford, Sanders, & Behrens, 2001). Fully addressing these low-distressed couples’ risks may thus require a shift in the dosage and timing of CRE toward less-intense, longer-duration programs so that the timing of intervention more closely aligns with when couples need to implement their new skills (Bradbury & Lavner, 2011). Taken together, the present findings reaffirm the value of targeting negative communication in preventive interventions, while also qualifying expectations regarding whether and when changes in negative communication will have beneficial effects on couple outcomes.

In conclusion, the data reported here are among the first to our knowledge to corroborate the surprising finding that some stably satisfied couples are vulnerable to eventual dissolution. The present results suggest that negative skills and emotions exchanged during problem-solving and socially supportive conversations are a leading explanation for this effect, whereas commitment to the relationship failed to distinguish between intact and dissolving couples. We note that this potent liability in interpersonal communication appears to be embedded in relationships with many assets, which may mask its effects in the short-term and necessitate longer-term intervention efforts in order to keep these vulnerable relationships healthy and strong.