Stress generation theory in couples with depression: A latent profile analysis

Profiles of depressive symptoms

Although there is a wealth of literature on the study of depressive symptoms, particularly on the association between depressive symptoms and relationship quality or satisfaction (e.g., Beach et al., 2003; Morgan, Durtschi, & Kimmes, 2018; Morgan, Love et al., 2018; Proulx et al., 2007; Whitton & Kuryluk, 2012), researchers argue for a more nuanced examination of depressive symptoms (Huq et al., 2016; Mullarkey et al., 2019; Pollitt et al., 2017). This is because depressive symptoms can be experienced differently and symptoms vary for each individual. For example, the American Psychiatric Association (2013) has specified that major depression can vary by level of severity: mild, moderate, and severe. Due to this, researchers have expanded their study of depressive symptoms to test the full range of depressive symptoms (none to severe) by identifying types, or profiles, of depressive symptoms among various samples of people (e.g., Hybels et al., 2009; Li et al., 2005). Generally, studies have identified two to four profiles, on average, of depressive symptoms among adults (Li et al., 2005). Together, this literature examines profiles at the individual level, but depressive symptoms is inherently relational illness (Mackinnon et al., 2012), so this means that dyadic profiles can categorize the depression experience at the couple level. Although many studies examine both partners’ depressive symptoms simultaneously (e.g., Whisman, 2013), we were unable to identify studies that examined dyadic profiles of depressive symptoms—as well as in conjunction with conflict. This is an important gap in the literature because from a stress generation perspective, couples can vary by degree of depressive symptoms as well as by conflict, yet research has not been able to test this. Thus, we aim to continue this trend of studying profiles of depressive symptoms but fill this gap in the literature by examining the various profiles of conflict and depressive symptoms among romantic couples.

Furthermore, in larger population samples, individual profiles were comprised of the vast majority of people in a no or mild level depressive symptoms with a small portion of the sample in the severe level profile of depressive symptoms. In other words, larger population samples had, on average, lower levels of depressive symptoms. To address this, some researchers argue to test differing levels of depressive symptoms within a sample (Segrin & Dillard, 1992). Previous studies using the Pairfam—which we use for this study—also showed samples of lower levels of depressive symptoms (Morgan, Durtschi, & Kimmes, 2018; Morgan, Love et al., 2018), thus we aim to test profiles as way to examine potentially varying levels of depressive symptoms among a general population of low depressive symptoms. Potentially, significant findings suggest that depressive symptoms do not need to reach clinical levels in order to have an affect (Kouros & Cummings, 2010).

Conflict and depressive symptoms

Given that people experience different levels of depressive symptoms, couples also experience different levels of conflict. For example, not all couples have intense, frequent arguments. It is possible that couples experience different levels of both conflict and depressive symptoms. From this, we aim to examine profiles of both couples based upon each partner’s depressive symptoms and perceptions of conflict. This not only adheres to a stress generation perspective but also adds new insights into the relationship between depressive symptoms and conflict among romantic couples. Despite the sizable literature on profiles of depressive symptoms, we were not able to identify studies that had identified profiles on both depressive symptoms and conflict among romantic couples, or on conflict alone. Some studies, however, compare low and high conflict couples (e.g., Whisman et al., 2015), but they have yet to study the varying levels of conflict among couples. Although we were unable to identify studies that examined profiles of depressive symptoms and conflict, and profiles of conflict among romantic couples, there is a wealth of literature on depressive symptoms and conflict in couples that provides context to our investigation.

Broadly, conflict among romantic couples experiencing depressive symptoms has been studied extensively in the literature—albeit by similar constructs, including relationship distress and relationship discord (e.g., Atkins et al., 2009; Beach et al., 1998; Coyne et al., 2002; Whisman, 2001, 2013). Conflicts are inevitable within any romantic relationship, but couples seem to argue more when one or both partners experience depressive symptoms (Atkins et al., 2009; Dennis & Ross, 2006). Particularly, depressed partners are known to be more irritable, fatigued, feel worthless, and experience changes in sleep and appetite. Considering these symptoms, it makes sense that these depressed partners may struggle to communicate with their partner and feel in emotional turmoil (Sharabi et al., 2016), which can make them more vulnerable to conflicts.

Romantic couples can experience both depressive symptoms and conflict simultaneously (Whisman, 2001). Partnered individuals with higher levels of conflict report higher depressive symptoms than those with less conflict (Whisman et al., 2015). In a clinical study of couples with depression, conflict in the relationship was shown to lower their chances of remission 12 weeks later (Denton et al., 2010). It has been shown that conflict is linked with increases in depressive symptoms (Choi & Marks, 2008) and depressive symptoms are linked with increases in relationship distress (Beach & O’Leary, 1993; Davila et al., 1997). These associations, however, seem to be only occasionally supported in the cross-sectional literature (Denton et al., 2003) and have been mediated by adult attachment orientations (Marchand, 2004) and other factors (see Whisman, 2013). Longitudinal studies, however, suggest that relationship discord is linked with greater depressive symptoms over time (Beach et al., 2003; Whisman, 2013). Together, this literature supports that depressive symptoms and conflict co-occur in romantic relationships.

There are a number of important factors to consider when examining conflict. This includes the types of topics that couples argue about and the couples’ ability to resolve conflict in a healthy way, without the use of verbal aggression (Choi & Mark, 2008; Sharabi et al., 2016). Couples with depressive symptoms may argue to the point to being verbally aggressive, meaning that they yell, scream, and verbally berate their partner (Sharabi et al., 2016). Additionally, verbal aggression has been linked with greater depressive symptoms (Lawrence et al., 2009; Du Rocher Schudlich et al., 2011). Considering this, we aim to examine profiles of conflict generally, common topics of conflict, and verbal aggression.

Depressive symptoms and conflict have been studied across various types of samples including clinical trials (e.g., Atkins et al., 2009; Beach & O’Leary, 1993), community samples (Marchand, 2004), large populations (Uebelacker & Whisman, 2006), and women with postpartum depression (Dennis & Ross, 2006). Although most studies were able to test both partners, these samples were smaller than 400 couples (e.g., N = 166 couples; Beach et al., 2003). The few studies that did use large samples, however, were only able to test partnered individuals and not both partners (e.g., N = 2,538 individuals; Uebelacker & Whisman, 2006). Outside of the clinical trials (e.g., Atkins et al., 2009), a handful of studies were longitudinal and generally at two time points (e.g., Beach et al., 2003). From this, there is a need for not only more studies of larger samples but also large samples of couples that are tested over time. We aim to address this need by testing depressive symptoms and conflict among couples with a large dyadic sample at three time points.

Depressive symptoms and relationship instability

Before couples separate and divorce, partners often perceive relationship problems and think about separation and divorce. This is more than just dissatisfaction in the relationship, as partners are distressed to the point of perceiving their relationship is in trouble, hence relationship instability. This is problematic because the more instable partners feel about the relationship puts them at risk for separating or divorce (Müller & Castiglioni, 2015). There is support that depressive symptoms are linked with greater relationship instability (Sharabi et al., 2016). Additionally, greater couple conflict is linked with more perceived relationship instability (Johnson et al., 2018). Furthermore, there is a sizable literature that supports that relationship uncertainty is associated with greater depressive symptoms in couples (Knobloch & Delaney, 2012; Knobloch et al., 2011; Knoblochet et al., 2016). Relationship uncertainty is similar to relationship instability in that partners are uncertain about the status of their relationship. However, relationship instability could be argued to be more severe than relationship uncertainty as these couples are not just lacking confidence in their relationship but potentially considering the end of the relationship. From this literature, there is a need for more studies on relationship instability and depressive symptoms, but it has not been clarified if this varies be the level of depressive symptoms and conflict. Thus, we aim to examine if the profiles of depressive symptoms and conflict differ in terms of relationship instability.

Additional variables for consideration

Considering this literature at large, there are number of important controls to examine when testing depressive symptoms and conflict among romantic couples. First, depressive symptoms and conflict seem to be more problematic as people age, or among older samples (Whisman, 2013). Second, married couples tend to have a higher conflict and depressive symptoms when compared to cohabitating couples (Uebelacker & Whisman, 2006). Third, the longer men were in a problematic relationship, the more vulnerable they were to depressive symptoms compared to men who were in their relationship for a shorter period of time (Kouros et al., 2008). Fourth, couples with more education and income tend to have less depressive symptoms. Lastly, relationship satisfaction has been shown to be an important factor when considering depressive symptoms among couples (Morgan, Durtschi, & Kimmes, 2018; Katz et al., 1999).

Present study

From this literature and stress generation theory, we aim to test profiles of couples with depressive symptoms and conflict behaviors with relationship instability 1 year later. To do this, we ask the following research questions:

RQ1: What are the profiles of couples’ depressive symptoms and conflict behaviors?

Sample

At Wave 2, 60.3% of couples were from Cohort 1 (1971–1973), 37.8% from Cohort 2 (1981–1983), and 1.9% from Cohort 3 (1991–1993). On average, these couples had been together for 9.82 years (standard deviation [SD] = 6.03), 68% of couples had at least one child, and had an average household income of €2,950.47 (SD = €1,433.36). The majority of couples were married (67.7%), 21.5% were cohabitating, 7.5% were together, but currently living a part, and 3.2% were together, but separated or divorced. Males in these relationships were, on average, 35 years old (SD = 6.35) while 40.9% of males reported having a university-level education. Females in these relationships were, on average, 32 years old (SD = 5.84) while 31.9% of females reported having a university-level education. Overall, these couples reported high levels of relationship satisfaction (8.28 for males, 8.26 for females) and had mild levels of depressive symptoms (1.58 for males, 1.66 for females).

Measures

Depressive symptoms

Ten items from the State-Trait Depression Scale (STDS; Spaderna et al., 2002) assessed overall mood. Although this German measure assessed overall mood, it has been highly correlated with the Beck Depression Inventory (Beck & Steer, 1987) and demonstrated consistency with the STDS English version in English-speaking samples (Krohne et al., 2002). These 10 items ranged from “I am depressed” and “I am sad” to “I feel good” and “I feel secure.” Partners responded to these items from 1 (not at all) to 5 (all the time). Five items were reverse coded, then all 10 items were averaged together, so that higher scores indicated more depressive symptoms. Both partners were asked to identify their biological sex (i.e., male, female). Thus, to create variables for males and females, these averaged variables were separated by biological sex in order to create males’ and females’ depressive symptoms. Each of these variables demonstrated acceptable reliability in males’ depressive symptoms at Wave 2 (α = .90) and Wave 3 (α = .87) as well as females’ depressive symptoms at Wave 2 (α = .90) and Wave 3 (α = .89; see Table 1).

OPEN IN VIEWER

Table 1.

Descriptives and sample characteristics (N = 1,598).

Characteristic/variable	M or %	SD	Range	α
W2 females’ depressive symptoms	1.66	0.47	1–5	.90
W2 males’ depressive symptoms	1.58	0.41	1–5	.90
W2 females’ conflict	2.55	0.64	1–5	.77
W2 males’ conflict	2.46	0.62	1–5	.74
W2 household monthly income (euros)	2,902.31	1,170.23	199–7,000	—
W2 relationship duration	9.82	6.03	0.08–33.58	—
W2 females’ age	32.07	5.84	18–57	—
W2 males’ age	34.85	6.35	18–69	—
W2 relationship satisfaction	8.26	1.94	1–10	—
W2 relationship satisfaction	8.29	1.85	1–10	—
W2 females’ university-level education (ISCED)	40.90%	—	0, 1	—
W2 males’ university-level education (ISCED)	31.90%	—	0, 1	—
W3 females’ depressive symptoms	1.66	0.48	1–5	.89
W3 males’ depressive symptoms	1.59	0.43	1–5	.87
W3 females’ conflict	2.57	0.66	1–5	.76
W3 males’ conflict	2.47	0.65	1–5	.80
W3 females’ verbal aggression	1.89	0.94	1–5	.75
W3 males’ verbal aggression	1.66	0.80	1–5	.74
W3 females’ conflict: Leisure time	2.06	1.01	1–5	—
W3 males’ conflict: Leisure time	2.15	0.99	1–5	—
W3 females’ conflict: Household labor	2.28	1.03	1–5	—
W3 males’ conflict: Household labor	2.25	1.09	1–5	—
W3 females’ conflict: Finances	2.07	1.02	1–5	—
W3 males’ conflict: Finances	2.04	1.01	1–5	—
W3 females’ conflict: Career	1.92	0.99	1–5	—
W3 males’ conflict: Career	2.02	1.01	1–5	—
W3 females’ conflict: Relationship	2.08	1.02	1–5	—
W3 males’ conflict: Relationship	2.02	0.98	1–5	—
W4 females’ relationship instability	0.35	0.78	0–3	.81
W4 males’ relationship instability	0.45	0.89	0–3	.76

Note. SD = standard deviation.

Data analysis plan

We conducted the analysis plan in several steps. First, we coded and created our variables using SPSS 25. Second, prior to testing our research questions, we conducted preliminary analyses using both SPSS 25 and Mplus 8.0 (Muthén & Muthén, 1998–2012). Particularly, we tested for measurement differences between anchor and partners and determine an estimator to address missing data as well as non-normality. Third, to test our first research question, we used latent profile analyses in Mplus. Latent profile analyses are finite mixture models that test the heterogeneity within a sample and organize profiles by grouping people together with similar responses. In following common latent profile analysis (LPA; McCutcheon, 1987) procedures, we tested a range of profiles from 1 to 5. For each model, the controls at Wave 2 predicted the profiles which were comprised of males’ and females’ depressive symptoms, topics of conflict, verbal aggression, and conflict at Wave 3. Because latent profile models use all of the variables in the model to calculate the model, we also had our profiles predict males’ and females’ relationship instability at Wave 4. Profiles were evaluated by common indicators (Nylund et al., 2007): lower Akaike’s information criterion (AIC), Bayesian information criterion (BIC), adjusted BIC (ABIC), a significant Lo–Mendell–Rubin likelihood ratio test (LMR-RT), and a significant bootstrap likelihood ratio test (BLRT). Lastly, logistic regressions were then used for the profiles to predict males’ and females’ relationship instability at Wave 4. To provide a more in-depth analysis of relationship instability, we will run additional tests of the profiles predicting males’ and females’ three individual items of relationship instability at Wave 4 (i.e., males’ and females’ reports of relationship is in trouble, considering separation/divorce, and have suggested separation/divorce).

Preliminary analyses

Determining profiles

We tested the LPA model from one profile to five profiles that provided information based our criteria (see Table 2). Most noticeably, the two- and three-profiles had significant LMR-RT, meaning that they were a significant improvement upon the previous profile. Although the four-profile did not have a significant LMR-RT, it did have a significant BLRT and lower AIC, BIC, and ABIC values. The five-profile was similar to the four-profile model except it had a profile of 8% of the sample, which questions the reliably of this group because it could be an outlier group rather a distinct group. A previous simulation study found BLRT and BIC to be robust indicators in comparing profiles in mixture models (Nylund et al., 2007). Based on this, we used the four-profile model because it had had lower BIC values than the two- and three-profiles as well as a significant BLRT.

OPEN IN VIEWER

Table 2.

Descriptive of indicators for a number of profiles for the latent profile models for couples (N = 1,598).

Profiles	Con	LL	AIC	BIC	ABIC	LMR-RT	BLRT	Ent	C1%	C2%	C3%	C4%	C5%
1	Y	−63,569.27	127,258.55	127,581.14	127,390.53	—	—	1.00	1.00	—	—	—	—
2	Y	−22,786.04	45,704.08	46,041.28	45,831.64	4,128.49**	−24,859.36**	0.89	0.64	.36	—	—	—
3	Y	−22,268.48	44,732.95	45,233.64	44,922.35	1,030.59**	−22,786.04**	0.85	0.36	.45	.19	—	—
4	Y	−22,043.58	44,347.16	45,011.34	44,598.40	447.83	−22,268.48**	0.81	0.37	.23	.21	.17	—
5	Y	−21,870.96	44,065.92	44,893.59	44,379.01	334.53	−22,038.96**	0.81	0.23	.32	.20	.15	.08

Note. Con = convergence; LL = log likelihood; AIC = Akaike’s information criterion; BIC = Bayesian information criterion; ABIC = sample-size adjusted BIC; LMR-RT = Lo–Mendell–Rubin adjusted likelihood ratio test; Ent = entropy, C1% = percentage of sample in Profile 1 and so forth.

*p < .05. **p < .01 Bold values determine the use of the 4-profile model for the subsequent analyses..

Four profiles

The four-profile model revealed four distinct profiles that ranged from low to moderate levels of conflict and depressive symptoms (see Figure 1 and Table 3). We found the congruent low conflict profile—that comprised 37.40% of the sample—where both males and females reported almost none to minimal levels of conflict, verbal aggression, depressive symptoms, and almost never argued about the five topics of conflict. This profile is congruent in that both partners (males and females) reported similarly low levels. Next, the congruent moderate conflict profile (17.10%) had both males and females reporting mild depressive symptoms and moderate levels of conflict and conflict behaviors. Similar to the congruent low conflict group, both partners were generally congruent in that they reported a similar degree of conflict and depressive symptoms. The incongruent female moderate conflict profile (23.90%) had females reporting mild depressive symptoms and moderate levels of conflict, verbal aggression, and perceiving more arguments about topics of conflict, while their partners reported lower depressive symptoms, conflict, verbal aggression, and perceived less arguments about topics of conflict. Lastly, the incongruent male moderate conflict profile (21.60%) was opposite from the incongruent female moderate conflict profile, in that males reported mild depressive symptoms and moderate conflict, verbal aggression, and perceived more arguments on conflictual topics, while their partners reported lower depressive symptoms, conflict, verbal aggression, and perceived less arguments on conflictual topics.

OPEN IN VIEWER

Figure 1.

Profiles of partnered males’ and females’ depressive symptoms, conflict types, and conflict behaviors. Note. For an example of interpretation, the congruent low conflict profile shows females and males who both endorsed low levels of depressive symptoms, conflict, verbal aggression, and topics of conflict.

OPEN IN VIEWER

Table 3.

Descriptive statistics for profile groups.

	Congruent low conflict (N = 463, 37.40%)	Incongruent female moderate conflict (N = 288, 23.90%)	Incongruent male moderate conflict (N = 261, 21.60%)	Congruent moderate conflict (N = 211, 17.10%)
	M (SD)	M (SD)	M (SD)	M (SD)
F. Depressive symptoms	1.47 (.42)	1.87 (.42)	1.54 (.42)	1.95 (.42)
F. Conflict	2.12 (.48)	2.86 (.24)	2.49 (.48)	3.30 (.48)
F. Verbal aggression	1.34 (.73)	2.2 (.73)	1.71 (.73)	3.00 (.73)
F. Conflict: Leisure time	1.6 (.92)	2.35 (.92)	2.08 (.92)	2.72 (.92)
F. Conflict: Household labors	1.71 (.94)	2.76 (.94)	2.17 (.94)	3.22 (.95)
F. Conflict: Career	1.59 (.90)	2.34 (.90)	2.01 (.90)	2.94 (.90)
F. Conflict: Finances	1.50 (.92)	2.29 (.91)	1.91 (.91)	2.35 (.93)
F. Conflict: Relationship	1.48 (.89)	2.49 (.84)	1.92 (.84)	3.06 (.84)
M. Depressive symptoms	1.39 (.37)	1.51 (.38)	1.75 (.38)	1.90 (.38)
M. Conflict	2.02 (.47)	2.41 (.47)	2.70 (.47)	3.21 (.47)
M. Verbal aggression	1.19 (.63)	1.56 (.63)	1.90 (.63)	2.61 (.63)
M. Conflict: Leisure time	1.67 (.82)	1.92 (.82)	2.61 (.82)	2.94 (.82)
M. Conflict: Household labors	1.72 (.87)	2.19 (.87)	2.67 (.87)	3.14 (.87)
M. Conflict: Career	1.49 (.87)	1.84 (.83)	2.34 (.83)	2.92 (.83)
M. Conflict: Finances	1.57 (.83)	1.84 (.88)	2.41 (.88)	2.72 (.88)
M. Conflict: Relationship	1.43 (.75)	1.81 (.75)	2.39 (.75)	3.14 (.75)

Note. All of the means and SDs are significant at p < .01. SD = standard deviation.

Next, we examined the odd ratios of the control variables at Wave 2 predicting profile membership at Wave 3. These results are summarized here; full details are found in Table 4. Males’ depressive symptoms and conflict were associated with an increase in the odds of being in the incongruent male moderate conflict profile, while females’ depressive symptoms and conflict were associated with an increase in the odds of being in the incongruent females’ conflict group than the congruent low conflict profile. On the other hand, both males’ and females’ depressive symptoms and conflict were associated with larger increases of being in the congruent moderate conflict profile in comparison to the congruent low conflict profile. Of the controls, however, females’ relationship satisfaction was consistently associated with a decrease in the odds of being in the three profiles (incongruent female moderate, incongruent male moderate, and congruent moderate conflict) 1 year later in comparison to the congruent minimal conflict profile.

OPEN IN VIEWER

Table 4.

Partnered males’ and females’ unstandardized, ORs, and significance levels for each profile (N = 1,598).

	Incongruent female moderate conflict			Incongruent male moderate conflict			Congruent moderate conflict
Parameter estimate	b	SE	OR	b	SE	OR	b	SE	OR
Relationship duration	−0.05	0.04	0.95	−0.04	0.04	0.96	−0.08*	0.04	0.92
Married	0.34	0.45	10.40	0.51	0.36	1.43	0.39	0.38	10.48
Household incomea	−0.06	0.14	0.94	0.07	0.13	1.07	−0.41**	0.15	0.66
Females’ age	−0.05	0.05	0.95	−0.03	0.04	0.97	−0.08	0.04	0.92
Males’ age	−0.03	0.04	0.97	−0.05	0.04	0.95	−0.01	0.04	0.99
Females’ relationship satisfaction	−0.34*	0.11	0.71	−0.27**	0.10	0.76	−0.36**	0.11	0.70
Males’ relationship satisfaction	−0.05	0.10	0.95	−0.13	0.09	0.88	−0.16	0.09	0.85
Females’ depressive symptoms	1.62**	0.45	5.05	−0.012	1.01	0.89	1.47**	0.42	4.34
Males’ depressive symptoms	0.45	0.88	1.57	2.29**	0.44	9.88	1.65**	0.58	5.21
Females’ conflict	2.33**	0.46	10.28	0.89	0.50	2.43	2.60**	0.36	13.46
Males’ conflict	0.94*	0.48	2.56	1.78*	0.70	5.93	2.92**	0.35	18.54
Females’ educationb	0.30	0.29	1.35	0.12	0.28	1.13	−0.03	0.32	0.97
Males’ educationb	−0.54	0.41	0.58	−0.49	0.30	0.61	−0.14	0.33	0.87

Note. OR = odd ratio; SE = standard error. Reference is the congruent low conflict profile. All predictors were from Wave 2 and profiles were comprised of Wave 3 variables.

^a Euros.

^bUniversity-level education.

*p < .05. **p < .01.

Relationship instability

Finally, using the same four-profile model above, the four profiles then predicted males’ and females’ relationship instability at Wave 4, which revealed several differences between the profiles (see Table 5). Most noticeably, being in the congruent moderate conflict profile was associated with a 62% increase in the probability of females and 54% increase in the probability of males reporting relationship instability 1 year later. All of the other profiles had lower probabilities of males and females reporting relationship instability 1 year later. Additionally, logistic regression tests revealed that, in comparison to the congruent moderate conflict profile, the three other profiles were significantly associated with a decrease in the odds of males and females reporting relationship instability 1 year later.

OPEN IN VIEWER

Table 5.

Probabilities of the four profiles being in an instable relationship (N = 1,598).

	Congruent low conflict		Incongruent females’ conflict		Incongruent males’ conflict		Congruent moderate conflict
	P	SD	P	SD	P	SD	P	SD
Males’ relationship instability W4
0	.96**	.01	.82**	.04	.80**	.05	.45**	.05
1	.04**	.01	.18**	.04	.20**	.05	.54**	.05
Males’ relationship instability W4
0	.96**	.01	.66**	.05	.84**	.07	.38**	.06
1	.04**	.01	.34**	.05	.16**	.07	.62**	.06

Note. 0 = did not endorse any relationship instability items. 1 = endorsed at least one of the relationship instability items. SD = standard deviation.

*p < .05. **p < .01.

Furthermore, we ran additional analyses with the four-profile model predicting males’ and females’ individual 3 items of relationship instability in place of the previous dichotomous males’ and females’ relationship instability variables. This four-profile model had a higher BIC (47,049.74) than our previous four-profile model, suggesting that our above analyses used a more parsimonious model, which makes sense because it used two outcome variables rather than six. However, this additional analysis revealed the following. Similarly, the congruent moderate conflict profile had not only all significant probabilities with all six outcome variables but much higher probabilities. For example, being in the congruent moderate profile was associated with an increase in the probability of reporting the relationship was in trouble 1 year later (77% for females, 64% in males), seriously considering separation/divorce (53% for females, 36% in males), seriously suggested separation/divorce (38% for females, 27% in males). The other profiles were associated with an increase in the probability of not endorsing the relationship instability items. However, there were only two exceptions that had very small probabilities. Such as those in the congruent low conflict profile were associated with a small increase in the probability of reporting their relationship to be in trouble 1 year later (2% for females, 4% for males). Together, these analyses reveal that the congruent moderate conflict profile to be more strongly associated with males’ and females’ relationship instability 1 year later than the other profiles.

Depressive symptoms and conflict behaviors vary by profile

First, our analyses revealed four distinct profiles: congruent low conflict, incongruent female moderate conflict, incongruent male moderate conflict, and congruent moderate conflict. The congruent low conflict and congruent moderate conflict profile demonstrated a high degree of interspousal agreement, where males and females reported similar levels of both depression and conflict. Additionally, males’ and females’ prior depressive symptoms, as well as conflict, were linked with the congruent moderate conflict profile, when compared to the congruent low profile. This suggests that couples with depressive symptoms and conflict were more likely to be in the profile where couples had increased depressive symptoms and conflict 1 year later. This finding corroborates stress generation theory and much of the literature (e.g., Hammen, 2006). This is not surprising in that when comparing both low and moderate congruent profiles, those with greater depressive symptoms and conflict are more likely to generate more depressive symptoms and conflict later on. What is interesting, however, is that when comparing the associations between depressive symptoms and conflict, both males’ and females’ conflict had a much higher increase in the odds of being in the congruent moderate conflict profile than males’ and females’ depressive symptoms. Continuing with the stress generation theory framework, this could suggest that prior conflict—as a stressor—may be more of a predictor than prior depressive symptoms of future depressive symptoms and conflict in couples. One possible explanation for this is due to the fact that, overall, our sample reported higher levels of conflict, relative to depressive symptoms. Even so, this is consistent with growing support for, when viewed longitudinally, relational conflict is linked with more depressive symptoms over time (Beach et al., 2003; Whisman, 2013).

Second, the incongruent female moderate conflict and incongruent male moderate conflict profiles were defined by a lack of interspousal agreement. This suggests that within couples, partners may differ in how much they perceive conflict in their relationship and experience depressive symptoms. It is known, particularly from clinical studies, that partners can differ on their levels of depressive symptoms (e.g., Denton et al., 2010), but partners differing on the level and type of conflict is intriguing. This may suggest that these couples are interpreting the same conflict differently, with one partner perceiving the conflict as more severe, or worse, than the other partner does (Sillars et al., 2004). Furthermore, our findings also support a stress generation theory framework in that males’ and females’ prior depressive symptoms and conflict were associated with them being in the profile where they, respectively, experienced greater depressive symptoms and conflict. This suggests that even in a romantic relationship, the partner with more depressive symptoms and more perceived conflict is at risk of generating more depressive symptoms and conflict later on. Similar to the previous congruent profiles, conflict had a stronger increase in the odds of being in the profile than depressive symptoms. This could suggest that when examining these profiles, conflict seems to be a stronger predictor of the profiles than depressive symptoms.

Some profiles are more relationally unstable

Third, in terms of relationship instability, the congruent moderate conflict profile had the highest probability of males and females reporting relationship instability 1 year later compared to the other three profiles. Specifically, this group of couples was more likely to report that their relationship was in trouble, seriously considered separation/divorce, or seriously suggested separation/divorce than the other three profiles. This supports stress generation theory and supports previous literature on relationship instability (e.g., Sharabi et al., 2016) as well as relationship uncertainty (e.g., Knobloch & Delaney, 2012) in that higher depressive symptoms and conflict are associated with relationship instability for both partners. This is consistent with previous research in that even among samples of less depressive symptoms, depressive symptoms do not need to be at clinical levels in order to have effect on the couples’ relationship (Kouros & Cummings, 2010). Previous literature suggests that both partners can be in agreement, or a consensus, about how they handle minor and major conflicts (Du Rocher Schudlich et al., 2011). Particularly, these couples with greater depressive symptoms struggle to resolve their conflicts and tend to use more negative conflict tactics (Du Rocher Schudlich et al., 2011). Hence, we found a profile of couples that not only agreed on their level of conflict but perceived problematic levels of conflict and depressive symptoms. Furthermore, when both partners are in agreement on moderate levels of depressive symptoms and conflict, it can increase the risk of both partners expressing relationship instability 1 year later. From a stress generation theory perspective, it is possible that relational conflict, in the presence of a system with two depressed partners, could be harder to work through than if at least one partner was not experiencing depressive symptoms.

Surprisingly, both incongruent profiles did not have a high association with relationship instability 1 year later. Although it would be initially thought that differing perspectives between partners would be problematic, conflict is not always problematic. For example, couples with minimal conflict and relationship instability can suddenly divorce (Amato & Hohmann-Marriott, 2007), while couples can remain together even though there is a lot of conflict and they are not satisfied with their relationship (Hawkins & Booth, 2005). Additionally, when couples experience depressive symptoms, conflict can act as a buffer for the relationship (Kouros et al., 2008) because couples may end up resolving the conflict and not put their relationship in jeopardy. Subsequently, these within-couple differences in how partners perceive mood and conflict appear to affect couple functioning; perhaps partners’ ability to resolve issues or persist through difficult outcomes may buffer them from the potentially detrimental effects of depression and conflict.

Implications

Although our sample had minimally to moderately depressive symptoms, these findings illustrate that couples with nonclinical depressive symptoms as well as with moderate conflict may have some effect on their relationship 1 year later. Furthermore, these findings are among a general population, which can suggest implications among a general population of couples. Particularly, the profiles found in this study have the potential to inform the ways practitioners assess, treat, and educate couples with varying levels of conflict and depression. For example, if an assessment reveals the presence of depressive symptoms in one or both partners and/or couple conflict, the clinician can conduct a more thorough depressive symptoms and conflict assessments to gauge the couple’s position, or profile. These profiles suggest that partners can be congruent or incongruent in their experiences with depressive symptoms and conflict. This, in turn, can inform the case conceptualization and treatment planning processes. Specifically, for couples in the congruent moderate conflict profile, it is possible that they have been experiencing depressive symptoms and conflict for some time, and are at risk for relationship instability in the future. For these couples, couple therapy may be appropriate, as it has been found to be effective in treating not only the couple relationship but also in treating individual partners’ depressive symptoms. Previous research has found that when partners attend couple therapy to treat one (or both) partners’ depressive symptoms, they experience an improvement in relationship satisfaction—an improvement that is superior when compared to partners seeking individual treatment for depression (Denton et al., 2003). It has also been found that for female partners who attend couple therapy for depression, couple therapy was superior to individual cognitive-based treatment (Beach & O’Leary, 1993). These studies must be considered in light of more recent findings, however, that found no significant increase in relationship satisfaction for a depressed sample of couples (Atkins et al., 2009). Thus, based on these studies, a combined treatment approach (i.e., couple therapy with concurrent individual therapy) may be helpful for couples in the congruent moderate conflict profile, as well as either incongruent profiles.

For practitioners who work with couples in either incongruent profiles, it may be challenging to help the other partner to understand the partner who is under more distress from perceived conflict and depressive symptoms. There are a number of clinical interventions that could be useful in helping partners understand one another better, such as enactments from emotionally focused couples therapy (EFT; Johnson, 2004) and integrative behavioral couples therapy (IBCT; Christensen & Jacobson, 1996). Although different in the approach, the goal is the same in that partners turn to each other and share their experiences. From an EFT lens, the aim would be to identify underlying emotions in each partner and then guide them to share those emotions with their partner. This potentially allows for vulnerability in the relationship and understanding of their perspectives. From an IBCT lens, an enactment is a tool to help foster not only understanding but acceptance of themselves and their partner. Enactments can highlight their roles in the relationship and thus sharing their perspectives help move toward acceptance of their relationship roles. It is also pertinent to note that not all unstable couples experience depressive symptoms. For some couples considering separation and divorce, the idea of relationship dissolution may be alleviating rather than debilitating, and may not experience depressive symptoms (Cohen et al., 2007).

Limitations

Despite these findings, there are several limitations. First, the results are correlational and causation cannot be inferred. Second, despite the large sample (N = 1,598), this is a German sample, and the generalizability of the sample to other countries is limited. Third, the depression measure used assessed depressive mood and did not cover all criteria for a depression diagnosis. Additionally, this was a general population sample and not a clinical sample, as such our sample had generally, on average, mild to moderate levels of depressive symptoms. Similarly, these couples, on average, did not endorse severe levels of conflict. We recommend future research to further examine these profiles among clinical samples of depressed and highly conflictual couples. Fourth, couples who report separation are more likely to drop out of the Pairfam study, however, this was not the case for couples who reported higher relationship instability. This means that to a small degree, attrition of the Pairfam sample can be attributed to relationship instability (Müller & Castiglioni, 2015). Additionally, consistent with previous literature on this measure, we dichotomized our relationship instability variables. Although dichotomizing the variables addressed our lack of highly instable couples in our sample, we acknowledge that doing so limited the variation of males’ and females’ relationship instability. Future research can explore how depressive symptoms and conflict are linked with couples who are highly instable. Understanding these limitations, this study contributes to what we know about couples and depressive symptoms by identifying different profiles of couples with depressive symptoms and conflict, and their association with relationship instability in the future.