PERSONAL VULNERABILITIES AND ASSORTATIVE MATE SELECTION AMONG NEWLYWED SPOUSES

Abstract

Assortative-mating theories propose that individuals select romantic relationship partners who are similar to them on positive and negative qualities. Furthermore, stress-generation and intergenerational transmission of divorce models argue that one’s depression history or family-of-origin relationship problems predict qualities of a marital partner that predispose them to relationship distress. We analyzed data from 172 newlywed couples to examine predictors and mediators of a marital partner’s risk index. First, an index of one’s own and one’s partner risk was created through factor analysis and was comprised of measures that indicate insecurity about oneself. This index was significantly correlated with baseline marital satisfaction and, among men, steps toward divorce at follow-up. Then, structural equation modeling tested direct and indirect pathways predicting partner’s risk index, analyzing prior depression history and family-of-origin relational impairment as predictors and one’s own risk index as the mediator. Results demonstrated that own risk index reliably predicted partner’s risk, while own risk index also mediated the relationship between own family-of-origin relational dysfunction/depression history and partner’s risk index. These results support assortative mating theories and suggest that the association between adverse family-of-origin relationships or depression history and the risk profile in one’s marital partner is explained by one’s own risk profile.

The majority of relationship research examines how features about the marital relationship—communication, stressful life events, or declines in marital satisfaction—affect subsequent relationship stability (Karney & Bradbury, 1995) and, in turn, aims to alter how couples communicate to strengthen relationships. These perspectives assign secondary significance to mate selection, under the assumption that interventions can teach couples how to better cope with stressful events, to communicate more effectively, or to overcome a partner’s psychopathology. The current project operates from the framework that variability in relationship satisfaction, stability, and the mental and physical health outcomes associated with marriage may be related to the qualities about the partner whom one decides to marry. The choice of a marital partner may be a crucial “bottleneck” that explains why some marriages remain steadfast and enhance well-being while others are characterized by significant distress/conflict and ultimately dissolve.

What characteristics in partners are most detrimental to later relationship satisfaction and stability? A review of the literature suggests that aspects related to emotional lability and insecurity about oneself emerge as key domains likely to affect intimate relationship functioning. Research indicates that traits of borderline and antisocial personality disorder, excessive anger, alcohol use, impulsivity, insecure attachment style, and low self-esteem are associated with interpersonal conflict and relationship disturbances for both the target individual and the partner. For example, women’s Cluster B personality traits, including borderline and antisocial symptoms, were associated with a greater number of conflict stressors with one’s partner and lower relationship satisfaction (Daley, Burge, & Hammen, 2000), while insecure attachment (Davila, Bradbury, & Fincham, 1998), neuroticism (Karney & Bradbury, 1997), and hostility and anger (Baron, Smith, Butner, Nealey-Moore, Hawkins, & Uchino, 2007) have also been associated with lower relationship satisfaction. A series of experiments (Murray, Derrick, Leder, & Holmes, 2008) demonstrated that in situations when low self-esteem individuals were subject to the possibility of interpersonal rejection from romantic partners, such individuals decreased connectedness with and support-seeking from their partners, suggesting that low self esteem negatively affects relationship quality. Furthermore, marriages where one partner had alcohol dependency were substantially more likely to end in separation or termination (Waldron, Heath, Lynskey, Bucholz, Madden, & Martin, 2011).

Drawing from these studies, a first step in the current project is to test a series of individual factors related to emotional lability and insecurity simultaneously to determine whether they cluster together into a higher order factor that we define as an index or profile of partner’s risk. We then correlate this risk index with baseline marital satisfaction and subsequent steps to divorce to validate our risk index as one that is associated with risk for lowered marital satisfaction and potential stability (i.e., relational distress). Next, we ask whether one’s own risk index (a) predicts partner’s risk index and (b) mediates the relationship between family-of-origin relational problems (parental conflict, parental divorce, and poor-quality maternal and paternal relationships with the target participant) or depression history and partner’s risk index. We review the relevant literature on how one’s own risk index, depression history, and dysfunction in family-of-origin relationships may each be associated with partner’s risk index and general intimate relationship dysfunction, in order to develop our framework for mediational analyses and validate our proposed risk factor.

ASSORTATIVE MATING AND RELATIONSHIP SATISFACTION

The concept of assortative mating (Crow & Felsenstein, 1968) proposes that individuals choose partners who are more similar to them on various characteristics at a rate greater than expected by chance. Research supports this idea, as individuals marry partners who are like them in terms of demography (Blackwell & Lichter, 2000), agreeableness and openness (McCrae et al., 2008), and mental disorders (Butterworth & Rodgers, 2006). Furthermore, data with partners of a variety of relationship types (i.e., cohabitating, dating, married) has also shown similarity between partners on negative characteristics like antisocial behaviors (Kim & Capaldi, 2004), and unipolar and bipolar depression (Mathews & Reus, 2001). The length of an intimate relationship fails to predict similarity among married couples (Watson et al., 2004), suggesting that similarity— at least in terms of demographic variables—is due to initial assortment for particular qualities rather than convergence. Additional research has addressed the consequences of assortative mating on relationship stability, finding that a newlywed marital partner’s similarity on positive traits like agreeableness predicts higher relationship satisfaction (Luo & Klohnen, 2005), whereas similarity among marital partners on negative characteristics like unipolar depression predicts divorce (Butterworth & Rodgers, 2008). Taken together, assortative mating theory suggests a direct association between one’s own risk profile and that of his/her partner, with adverse relationship consequences when partners match on psychopathology.

DEPRESSION HISTORY AND ADVERSE PARTNER SELECTION

Depression also merits close attention as a predictor of partner’s risk profile: it is the most commonly-experienced mental disorder (Kessler et al. 2003) with profound social and economic consequences (Greenberg et al., 2003). Interpersonal models of depression have determined that depression is associated with subsequent intimate relationship problems, including declines in marital satisfaction (Fincham, Beach, Harold, & Osborne, 1997; Whisman, 2007), marital communication characterized by negative behaviors and affect (Rehman, Gollan, & Mortimer, 2008), increased interpersonal conflict stressors (Hammen, 1991), and negative feedback-seeking (Casbon, Burns, Bradbury, & Joiner, 2005) that is linked to interpersonal rejection (Coyne, 1976). These models suggest that cognitive and interpersonal deficits that accompany depression may be connected to the depressed person’s desire for/dependency on intimate relationships in ways that enhance maladaptive partner selection.

For example, one line of research has proposed that depressive individuals cope with interpersonal stressors through social withdrawal (Agoston & Rudolph, 2010) and to avoid social rejection (Slavich, Thornton, Torres, Monroe, & Gotlib, 2009). Social withdrawal is associated with interpersonal dependency (Darcy, Davila, & Beck, 2005), suggesting that depressed individuals who withdraw from relationships may be overly dependent on the limited relationships that they do form. This dependency may mean that such individuals are less able to scrutinize their partner for potentially maladaptive characteristics that increase the risk for relationship dissatisfaction. As interpersonal problem-solving deficits already characterize depressed individuals (Davila, Hammen, Burge, Paley, & Daley, 1995), these abilities may be even more compromised in the absence of substantial prior relationship experience.

In summary, cognitive and interpersonal deficits associated with depression may mean that a depressed person enters a relationship without the skills required to select a suitable partner or to navigate conflict once it occurs. Importantly, intimate relationship dysfunction is independent of current depressive symptoms, as even women with a history of depression but without current symptoms report lower marital satisfaction and a greater number of interpersonal stressors (Hammen & Brennan, 2002). These findings suggest the need, as our current study does, to examine depression history rather than current depressive symptoms as a predictor of partner’s risk index, particularly because individuals with a history of depression might enter into maladaptive romantic relationships irrespective of their level of current depressive symptomatology.

PARENTAL RELATIONSHIPS AS PREDICTORS OF RISKY PARTNER SELECTION

Sole emphasis on depression history as a determinant of partner’s risk profile and subsequent adverse relationship outcomes is likely to underestimate the complexity of mate selection. Indeed, alternative conceptual frameworks highlight factors other than depression as influences in partner selection. Prominent in this regard are the experiences that individuals have in their families as they grow up and the ways in which relationship dysfunction can be transmitted across generations. The emotional climate in the home, the nature of the relationship between one’s parents, children’s perceptions of their relationship with their parents, and parental divorce are known to predict depression (Oldehinkel, Ormel, Veenstra, de Winter, & Verhulst, 2008), aspects of mate selection (Wolfinger, 2003), and relationship functioning in adolescence and adulthood (Amato, 2001; Amato & Booth, 2001; Crockett & Randall, 2006). Based upon this literature, we will therefore test an array of family-of-origin relational problems (recollections of parental conflict, parental divorce, and the participant’s perceptions of their relationship quality with their mother and father) as predictors of a marital partner’s risk index.

THE CURRENT STUDY

Using a sample of 172 newlywed couples entering their first marriage, we will first derive an index of partner’s risk using factor analysis. In line with previous research that has argued for examining multiple risk factors rather than any specific risk factor as a predictor of adverse outcomes (Kraemer, Stice, Kazdin, Offord, & Kupfer, 2001; Rauer, Karney, Garvan, & Hou, 2008; Rutter, 1979), our dependent variable will represent the additive effects of several risk factors known to predict relationship dysfunction. We will also correlate this risk index with concurrent marital satisfaction and later steps toward divorce to validate this index as one denoting risk for concurrent and prospective relationship dissatisfaction and steps toward dissolution. This methodology also extends assortative mating research, which has frequently examined correspondence between individual traits among partners without considering their common variance or understanding that individuals often experience several risk factors simultaneously.

After deriving the index of partner’s risk, we test several models that analyze own risk index as a mediating variable between own depression history/family-of-origin relationship dysfunction and partner’s risk index. The viability of one’s own risk index as a mediating variable has been established by previous research that demonstrates a relationship between family-of-origin relationship distress/depression history, adverse relationship outcomes and own psychopathology (i.e., components of our study’s composite indices for own risk index and partner risk index). For example, depression history is associated with psychopathology in relationship partners and dysfunctional relationship outcomes like intimate partner violence (Hammen & Brennan, 2002; Keenan-Miller, Hammen, & Brennan, 2007), and depression is frequently a chronic and recurrent disorder (Coyne, Pepper, & Flynn, 1999). An association between parental discord/divorce and subsequent psychopathology has also been determined (Amato & Sobolewski, 2001; Pilowsky, Wickramaratne, Nomura, & Weissman, 2006). Finally, adolescents’ self-esteem mediated the relationship between parent-adolescent relationship quality and subsequent intimate relationship outcomes in young adulthood (Johnson & Galambos, 2014). Taken together, this research suggests models by which the factors that comprise our construct of own risk mediate the relationship between distal variables like family-of-origin relationship dysfunction and depression history and subsequent risk profiles in one’s eventual marital partner.

METHOD

PARTICIPANTS

The total sample consists of 172 newlywed couples recruited from Los Angeles County marriage licenses between May 1993 and January 1994. A total of 3,606 letters were sent to identified couples, with 637 (17.8%) expressing interest in participating. Our sample, representative of Los Angeles in terms of ethnicity, consisted of a significant number of ethnic minorities (39% of wives and 33% of husbands). The mean age was 27.6 years for husbands and 26.0 years for wives, while median annual income was between $21,000 and $30,000 for husbands and between $11,000 and $20,000 for wives. Husbands averaged 15.3 years of education while wives averaged 15.5 years.

PROCEDURE

Within the first six months of marriage, couples participated in a laboratory visit where they completed self-report measures including the risk components and clinical interviews to assess current and past depressive symptomatology. This study uses cross-sectional data from the baseline/initial laboratory visit.

MEASURES

Depressive Diagnostic History

A clinical interview adapted from the Structured Clinical Interview for DSM-III (SCID; Spitzer, Williams, Gibbon, & First, 1992) assessed for current (“during the past month”) and previous (“have you ever experienced a time when”) diagnoses of major depression using a 0 to 3 point scale (0 = no symptoms; 1 = 1–2 symptoms; 2 = 3–4 symptoms; 3 = diagnosable depression). At least one of the depressive symptoms must have been either depressed mood or anhedonia; otherwise, a score of 0 was given. The majority of men (67.3%) endorsed no prior symptoms of depression; 12.5% endorsed one or two symptoms, 9.5% three to four symptoms, and 10.7% disclosed five or more prior depressive symptoms. Among wives, 51.5% endorsed no prior depressive symptoms, with 12.7% endorsing one or two symptoms, 18.2% three or four symptoms, and 17.6% five or more symptoms. The full 0–3 range of this measure was used in all analyses.

Parental Divorce

Parental divorce before the age of 16 was assessed through clinical interview. Thirty-seven wives (21.5%) and 32 husbands (18.6%) reported experiencing parental divorce.

Parental Conflict

The 15 item true/false Family of Origin subscale from the Marital Satisfaction Inventory (Snyder, 1979) assessed for general levels of perceived family-of-origin warmth and conflict (i.e., “I had a very happy home life” and “my parents had very few quarrels”). Relevant items were reverse-coded such that higher scores indicated a higher level of familial conflict. Coefficient alphas were .83 for husbands and .86 for wives.

Relationship Quality with Mothers and Fathers

Participants completed a laboratory-created questionnaire derived from Hazan and Shaver’s work (1987), where, on a five-point Likert scale ranging from 1 (never) to 5 (usually), they responded to 18 adjectives using the following prompt: “Take a moment to think about your relationship with each of your parents while you were growing up. What were their attitudes, feelings, and behaviors toward you?” Sample adjectives included loving, critical and disinterested. Participants responded to each of these 18 items separately for their mothers and fathers. Relevant items were reverse-coded such that higher scores indicated a more negative relationship with one’s mother or father. Coefficient alphas for relationship quality with both mothers and fathers were .88 for husbands and .91 for wives.

Trait Anger

Participants completed the 25-item Multidimensional Anger Inventory (MAI; Siegel, 1986) to assess levels of felt and expressed anger and hostility. Items ranged from 1 (totally false) to 5 (totally true), with higher total scores indicating higher levels of anger. Sample items include “It is easy to make me angry” and “When I get angry, I stay angry for hours.” Coefficient alphas for both husbands and wives in the current sample were .86.

Dysfunctional Impulsivity

Six items captured partner’s self-reported dysfunctional impulsivity, defined as acting without prior thought (Dickman, 1990). All items ranged from 1 (does not describe me) to 7 (describes me very well), with higher total scores indicating higher levels of impulsivity. Sample items included “I often make up my mind without taking the time to consider the situation from all angles” and “I often get into trouble because I don’t think before I act.” All 6 of the selected individual items loaded on Dickman’s larger conceptualization of a dysfunctional impulsivity factor with loadings of .53 to .85. Furthermore, Dickman’s original dysfunctional impulsivity factor was highly correlated with several other impulsivity measures, including the Narrow-I (Eysenck & Eysenck, 1977) and the Personality Research Form Impulsivity Scale (Jackson, 1967) demonstrating the measure’s validity. Coefficient alphas in the current sample were calculated at .82 for wives and .88 for husbands.

Alcohol Symptoms/Consequences

A laboratory-created measure was derived from The Michigan Alcoholism Screening Test (Selzer, 1971). This adapted 20-item measure assessed for consequences and problem behaviors resulting from alcohol use. Sample items included “How often have your friends complained or expressed concerns as a parent while drinking?” and “How often have you hit your spouse or gotten into a physical fight with your spouse while you were drinking?” Items were scored on a 6-point scale: (never, has happened but not in the past year, happened once in the past year, happened twice in the past year, happened three times in the past year, and happened four or more times in the past year) and summed, with higher total scores indicating higher levels of alcohol-related problems. Coefficient alphas for the current study’s sample were calculated at .81 for wives and .91 for husbands.

Neuroticism

Participants completed the 12-item neuroticism subscale of the NEO-Five Factor Inventory, Form S (NEO-N; Costa & McCrae, 1992). Items were scored on a 1 (strongly disagree) to 5 (strongly agree) scale, and relevant items were reverse-scored such that higher total scores represented higher levels of neuroticism. Sample questions included “I am not a worrier” (reverse-scored) and “When I’m under a great deal of stress, sometimes I feel like I’m going to pieces.” Reliability values measured through coefficient alpha in the present sample were .86 for husbands and .86 for wives.

Self-Esteem

The 10-item Rosenberg Self Esteem Scale (RSES; Rosenberg, 1965) measured participants’ self-esteem. The items ranged from 1 (strongly agree) to 4 (strongly disagree), and five items were reverse-scored to create a measure where higher scores indicated lower levels of self-esteem. Alphas were .87 for both husbands and wives.

Anxious Attachment

Participants completed the 18-item Revised Adult Attachment Scale (Collins & Read, 1990), from which the 6-item anxious attachment subscale was derived. On a five-point scale ranging from 1 (not at all characteristic) to 5 (very characteristic), participants answered sample items like “In relationships, I often worry that my partner does not really love me.” Alphas were .83 for both men and women.

Marital Satisfaction

In order to validate the composite indices of partner risk (see below), we correlated measures of marital satisfaction with the composite index and constituent scales. We used two self-report marital satisfaction measures administered at the initial laboratory session. First, the 15-item Marital Adjustment Test (MAT; Locke & Wallace, 1959) was used. Second, the Semantic Differential (SMD; Karney & Bradbury, 1997) was employed. On this measure, participants rated their marital relationship on 15 pairs of opposing adjectives (e.g., satisfied/dissatisfied) using a 7-point visual scale.

DATA ANALYSIS

We conducted our analyses using Mplus, Version 7.3 (Muthén & Muthén, 1998–2012). We tested four separate models and two simultaneous models to examine husbands’ and wives’ own risk as mediators of the relationship between one partner’s family-of-origin relational distress or depression history and their partner’s risk index. We determined 99% confidence intervals based on bootstrapping with 1,000 replications. Both partners’ risk variables were therefore theoretically meaningful components of our analyses, and our models can be considered gender-specific couple-level models.

RESULTS

CREATION OF COMPOSITE PARTNER RISK OUTCOME AND OWN RISK PREDICTOR

Individual scores on each of the six aforementioned risk scales (anxious attachment, self-esteem, dysfunctional impulsivity, anger, alcohol use symptoms, and neuroticism) were initially standardized by gender. Exploratory factor analysis using a quartimax rotation (results were similar using other rotation methods; i.e., varimax, promax) was conducted on these six standardized scales. Results revealed one factor that explained 41.7% of the total variance. The included variables were partner reports of neuroticism (loading = .86), low self-esteem (.79), anxious attachment (.73), and anger (.61), demonstrating a coherent factor comprised of measures of insecurity about oneself, especially given research demonstrating an association between low self-esteem and aggression (Donnellan, Trzesniewski, Robins, Moffitt, & Caspi, 2005). The coefficient alpha for the composite measure that consisted of these four standardized scales was α = .75. The partner risk outcome variable was therefore defined as the sum of the partner’s standardized responses on these variables, while the own risk predictor was computed as the sum of the target’s responses on these variables. Further details of this exploratory factor analysis can be found in Table 1.¹

Table 1.

summary of exploratory Factor Analysis Results for Partner Risk Index Using Principal Components Analysis with Quartimax Rotation

Item	Factor Loadings
	Factor 1	Factor 2
Partner anxious attachment	.73	.07
Partner low self-esteem	.79	−.12
Partner anger	.61	.37
Partner neuroticism	.86	.08
Partner dysfunctional impulsivity	.31	.51
Partner alcohol use consequences	.05	.87
Eigenvalues	2.50	1.02
Percentage of variance explained	41.68	17.07

Note. Factor loadings above .40 are noted in bold.

Values for women’s risk index ranged from −6.06 to 7.85, with a mean of 0 and a standard deviation of 3.13, while the values for men’s risk index ranged from −6.16 to 9.77 with a mean of 0 and a standard deviation of 2.92. Scores on the individual four scales correlated with the composite risk index. For women’s risk index, these correlations ranged from r = .75 to r = .85 with a median correlation of r = .77, while correlations for men’s risk index ranged from r = .65 to r = .84 with a median correlation of r = .71. Taken together, these results indicate a wide degree of variability in partner risk across this factor and demonstrate that the constituent scales that comprised the larger index are highly correlated with the overall measure.

Validating the Risk Measure

We correlated raw scores from each of the individual scales that comprised our outcome variable, and the composite outcome, with the MAT and SMD marital satisfaction measures. From Table 2, scores from the overall composite risk factor and many of its consistent subscales were significantly correlated with these two measures of baseline marital satisfaction. We also examined correlations between the composite risk outcome and scores from 9-items taken from the Marital Status Inventory (Weiss & Cerreto, 1980) administered during the longitudinal portion of the larger study, at the fourth year of marriage. Relevant items were reverse-coded such that scores ranged from 0, indicating no steps to divorce, to 9, indicating that all of the steps had been taken, including divorce. Sample items included “I have occasionally thought of divorce or wished that we were separated, usually after an argument or other incident,” and “I have discussed the question of my divorce or separation with someone other than my spouse (trusted friend, psychologist, minister, etc.).” Among men, the measure of steps taken to divorce was significantly correlated with wives’ overall risk outcome (r = .22, p < .05), while the relationship between wives’ steps to divorce and husbands’ overall risk was non-significant (r = .03, p = .74). These analyses help validate the overall risk measure as being negatively associated with concurrent marital satisfaction and with men’s report of steps to divorce after four years of marriage, thus clarifying our risk index as one related to risk for concurrent reduced marital satisfaction and prospective steps to divorce in men.

Table 2.

Cross-Partner Correlations Between Partner Risk subscales and Composite Risk Index with self-Reported Baseline Marital satisfaction

Measure	MAT		SMD
	Men	Women	Men	Women
Overall risk outcome	–.39***	–.31***	–.30***	–.18*
Anxious attachment	–.36***	–.17*	–.22**	–.05
Low self-esteem	–.27***	–.22**	–.18*	.13
Anger	–.35***	–.20*	–.33***	–.11
Neuroticism	–.25**	–.32***	–.22**	–.23**

Note. MAT = Marital Adjustment Test; SMD = Semantic Differential.

^*p < .05.;

^**p < .01;

^***p < .001

PRELIMINARY ANALYSES

Inter-correlations between the partner risk outcome and all potential predictors for the full sample are shown in Table 3. These results indicated a significant correlation between own risk index and partner’s risk index (r = .26) but also demonstrated that neither depression history nor any of the family-of-origin variables were significantly correlated with partner’s risk index. Small-to-medium correlations were observed between depression history and the family-of-origin variables, alongside generally medium-to-large correlations between the various family-of-origin variables.

Table 3.

Inter-Correlations Between Partner’s Risk outcome and All Predictors for Full sample

Variables	1	2	3	4	5	6	7
1. Partner’s risk	–
2. Own risk	.26***	–
3. Depression history	.02	.19***	–
4. Parental divorce	–.01	–.02	.08	–
5. Parental conflict	.02	.21***	.18**	.45***	–
6. Relationship quality with mothers	.06	.29***	.19**	.12*	.51***	–
7. Relationship quality with fathers	–.03	.20***	.18**	.21***	.57***	.42***	–

Note.

^*p < .05

^**p < .01;

^***p < .001

In light of the medium-to-large correlations between family-of-origin variables, we explored whether the four independent variables could be collapsed together through exploratory factor analysis. We followed the general procedures we employed in computing the partner risk outcome measure (i.e., standardizing variables by gender, running an exploratory factor analysis using principal components analysis). Results suggested that all four individual constructs could be comprised into a single factor (eigenvalue: 2.15) that explained 53.7% of the variance. Factor loadings were .88 for parental conflict, .78 for relationship quality with fathers, .70 for relationship quality with mothers, and .53 for parental divorce. A second factor was not determined, and the coefficient alpha of the resulting single factor was .70. We therefore created one variable to comprise family-of-origin relationship problems, composed of the sum of z-standardized values for our variables of parental divorce, parental conflict, and participants’ relationship quality with mothers and fathers.

Values for men’s family-of-origin relationship problems ranged from −4.70 to 8.51, with a mean of 0 and a standard deviation of 2.87, while these values for women ranged from −4.22 to 6.34 with a mean of 0 and a standard deviation of 2.92. Scores on the individual four scales correlated with the composite factor. Among men, these correlations ranged from r = .60 to r = .81 with a median correlation of r = .75, while correlations for women ranged from r = .62 to r = .90 with a median correlation of r = .72.

MAIN ANALYSES

In the first step of our analyses, four separate models were run that examined own risk as a mediator between the relationship between partner’s family-of-origin relationship problems or depression history and partner’s risk index. Initially, these models were inspected to first examine the relationship between husbands’ and wives’ levels of risk. Across all models, this value ranged from b = 0.24 to b = 0.30, with a p-value < .01. Therefore, own risk was a reliable predictor of partner’s risk index.

We examined indirect effects of family-of-origin relational problems or depression history via own risk to partner’s risk index (based on a bias-corrected bootstrap procedure implemented in MPlus 7.3). These separate analyses of family-of-origin relational problems and depression history provided initial support for the hypothesized indirect effects via own risk to partner risk index in three out of four models.

We therefore examined two models that simultaneously tested family-of-origin relational problems and depression history as predictors of partner risk index via own risk separately for men and women. Table 4 reports the results of these two simultaneous models. The results suggested that husbands’ family-of-origin relational problems (β = .19) and husbands’ depression history (β = .23) significantly predicted husbands’ risk index. In turn, the husbands’ risk predicted wives’ risk index (β = .28), whereas direct effects of family-of-origin relational problems (β = -.04) and depression history (β = −.10) on wives’ risk index were non-significant. The indirect effect from family-of-origin relational problems via husbands’ risk to wives’ risk index was significant at the 5% alpha level (b = .06, β = .05, t = 1.76, 95% CI: .01, .15) but not at the 1% alpha level (99% C.I.: −.01, .16), whereas the effect of depression history via husbands’ risk to wives’ risk index was significant at the 1% alpha level (b = .19, β = .06, t = 2.58, 99% CI: .03, .42).

Table 4.

Summary of Structural Equation Modeling for Simultaneous Mediational Models

Parameter Estimate	Model 1				Model 2
	b	β	t	99% CI	b	β	t	99% CI
FOIH->RISKH	0.20	0.19	2.19*	[−0.04, 0.44]
DEPH->RISKH	0.64	0.23	3.30**	[0.11, 1.13]
RISKH->RISKW	0.30	0.28	3.76***	[0.08, 0.46]
FOIH->RISKW	−0.04	−0.04	−0.51	[−0.27, 0.17]
DEPH->RISKW	−0.30	−0.10	−1.35	[−0.86, 0.30]
FOIW->RISKW					0.23	0.22	2.89**	[0.03, 0.45]
DEPW->RISKW					0.21	0.08	1.03	[−0.33, 0.70]
RISKW->RISKH					0.25	0.27	2.95**	[0.02, 0.47]
FOIW->RISKH					−0.07	−0.07	−0.89	[−0.27, 0.13]
DEPW->RISKH					0.11	0.05	0.62	[−0.35, 0.58]

Note: FOIH = Husbands’ Family-of-Origin Relationship Problems; RISKH = Husbands’ Risk Index; DEPH = Husbands’ Depression History; RISKW = Wives’ Risk Index; FOIW = Wives’ Family-Of-Origin Relationship Problems; DEPW = Wives’ Depression History.

^*p < .05;

^**p < .01;

^***p < .001

Furthermore, wives’ family-of-origin relational problems (β = .22) significantly predicted wives’ risk, but not wives’ depression history (β = .08). Wives’ risk, in turn, predicted husbands’ risk index (β = .27). Again, direct effects of wives’ family-of-origin relational problems (β = −.07) and depression history (β = .05) on husbands’ risk index were non-significant. The indirect effect from family-of-origin relational problems via wives’ risk to husbands’ risk (b = .06, β = .06, t = 2.05, 99% CI: .01, .16) was significant at the 1% alpha level, while the indirect effect from depression history via wives’ risk to husbands’ risk index was not significant (b = .05, β = .02, t = .93, 99% CI: −.08, .24). Figures 1 and 2 represent path diagrams that display our mediational analyses.

FIGURE 1.

Simultaneous models testing husbands’ risk index as a mediator between husbands’ depression history and husbands’ family-of-origin relational problems with wives’ risk index. H = husband, W = wife. Bold lines indicate significant paths, broken lines indicate non-significant paths. The indirect effect from family-of-origin relational problems via husbands’ risk to wives’ risk was significant at the 5% alpha level (b = .06, β = .05, t = 1.76, 95% CI: .01, .15) but not at the 1% alpha level (99% CI: −.01, .16), whereas the effect of depression history via husbands’ risk to wives’ risk was significant at the 1% alpha level (b = .19, β = .06, t = 2.58, 99% CI: .03, .42). Error terms have been omitted from the figure for greater clarity.

FIGURE 2.

Simultaneous models testing wives’ risk index as a mediator between wives’ depression history and wives’ family-of-origin relational problems with husbands’ risk index. H = husband, W = wife. Bold lines indicate significant paths, broken lines indicate non-significant paths. The indirect effect from family-of-origin relational problems via wives’ risk to husbands’ risk (b = .06, β = .06, t = 2.05, 99% CI: .01, .16) was significant at the 1% alpha level, while the indirect effect from depression history via wives’ risk to husbands’ risk was not significant (b = .05, β = .02, t = .93, 99% CI: −.08, .24). Error terms have been omitted from the figure for greater clarity.

DISCUSSION

The current study expands upon assortative mating literature by not only establishing a robust relationship between one’s own risk profile and that of one’s eventual marital partner, but also testing this own risk construct as a mediator of the relationship between a history of depressive symptoms or family-of-origin relationship problems and subsequent partner’s risk. We first defined risk through a factor analysis that indicated one outcome factor consisting of low self-esteem, anxious attachment, anger, and neuroticism. This factor was characterized by constructs related to insecurity about oneself and was significantly inversely correlated with concurrent measures of marital satisfaction and directly correlated with later steps to divorce in men. We followed similar factor analytic procedures to also construct an index of family-of-origin relationship problems. We then examined whether, as assortative mating theory suggests, own risk predicts partner risk’s index. Results indicated that own risk was a reliable predictor of partner’s risk index, supporting mate selection theories based upon similarity. In order to consider additional theoretical perspectives on partner selection, we also analyzed whether own risk mediated the relationship between more specific qualities such as depression history or relationship dysfunction in one’s family-of-origin and partner’s risk index. We formally tested individual and simultaneous mediational models that employed both depression history and family-of-origin relationship problems as predictors. Our simultaneous models indicated that own risk significantly mediated the relationship between own family-of-origin relationship problems and partner’s risk index in men and women, while men’s risk significantly mediated the relationship between men’s depression history and women’s risk index.

Before discussing these results, some limitations must be addressed. First, our sample was a community sample of couples newly entering their first marriage. As such, our sample was generally a highly-functioning group in terms of marital satisfaction and depression, especially at the initial time point. It is possible that we would have seen larger effect sizes or a greater proportion of significant results had we had a population with more extreme distributions on our predictors or the variables that comprised the risk index. A further consequence of sampling heterosexual newlywed couples means that we cannot account for other kinds of partnerships like same-sex couples, re-married couples, and cohabitating couples. These relationships may differ markedly in terms of a number of factors (i.e., barriers to relationship entry/exit; status of legal recognition) that may impact relationship quality/stability or the characteristics of individuals in these relationships. Our exclusion of certain variables from our overall risk outcome may also mean that specific constructs with clear relationship consequences were not captured, potentially skewing some results if our predictors were associated with these excluded constructs. Finally, the concurrent/cross-sectional nature of our predictions prevents making longitudinal predictions about how risk well before marriage predicts initial partner selection. We therefore could not ask questions about, for example, how aspects of earlier intimate relationships (both with previous relationship partners and the current spouse) affect ultimate partner selection.

On the other hand, it is likely that our risk indices and predictors represent relatively stable qualities over time, and we conducted a validity check through correlating individual scales and the composite factor with measures of relationship quality and steps toward dissolution to demonstrate relationship consequences to initial partner risk. Notable strengths of our study include its integration of assortative mating theory alongside models about the interpersonal nature and consequences of depression and the impact of parental relationship dysfunction on offspring’s relationship outcomes. We have addressed theoretical and methodological gaps in previous literature by considering partner risk beyond just correspondence in one domain (i.e., the relationship between one’s own depressive symptoms and those of one’s partner) and instead examined a multi-faceted risk index as suggested by several researchers (Rauer et al., 2008; Rutter, 1979).

In all models examining partner risk, we first confirmed that our partner risk index derived from factor analysis was significantly correlated with one’s own current relationship satisfaction and with men’s subsequent steps to divorce. These results confirm prior research that has examined specific components of our risk index as being associated with relationship distress (Baron et al., 2007; Davila et al., 1998; Karney & Bradbury, 1997; Murray et al., 2008) and validate the detrimental effects of anger, insecure attachment, neuroticism, and poor self-esteem on global relationship functioning and stability. These results also extend research—some using more established married couples, others using undergraduates in exclusive dating relationships—into a group of newlywed couples within the first six months of their first marriage and further associate the risk index with steps to divorce as opposed to solely relationship dissatisfaction.

Next, we tested direct links between one’s own risk index and that of one’s partner to test assortative mating predictions. We found that own risk index predicted that of one’s partner. These results support and extend assortative mating research (e.g., Crow & Felsenstein, 1968), which has demonstrated similarity between a target and one’s partner on individual characteristics like mood disorders (Mathews & Reus, 2001), antisocial personality traits (Kim & Capaldi, 2004), and general mental health (Butterworth & Rodgers, 2006). Our work demonstrates that individuals who score highly on an amalgamation of vulnerability factors characteristic of insecurity about oneself are likely to marry a partner who also displays a high level of these traits. We also extend prior research (i.e., Butterworth & Rodgers, 2008) by examining multi-faceted risk indices rather than focusing on a specific adverse quality like poor self-esteem, anxious attachment, or depressive symptomatology.

Not only did own risk predict partner risk, but own risk index significantly mediated the relationship between family-of-origin relationship problems and partner’s risk for both men and women, and the relationship between depression history and partner’s risk index in men. These results indicate that earlier experiences of relational distress in one’s family of origin, as well as depression history, are associated with a tendency to develop insecurity about oneself that predicts marriage to a partner with insecurity about him/herself. Furthermore, our findings extend prior research findings demonstrating that family-of-origin relationship problems (Amato, 2001; Amato & Booth, 2001; Wolfinger, 2003) or depression history (Hammen & Brennan, 2002) are associated with adverse partner selection and negative relationship outcomes by proposing a mediational pathway through one’s own risk profile. These results support prior literature, given that our own and partner risk index was comprised of an array of factors that indicate insecurity about oneself, and that negative family-of-origin relationship experiences and depression history are associated with patterns of insecure attachment and negative self-worth that are likely closely aligned with our risk indices (Diehl, Elnick, Bourbeau, & Labouvie-Vief, 1998; Dinero, Conger, Shaver, Widaman, & Larsen-Rife, 2008; Orth & Robins, 2013; Sowislo & Orth, 2013). Although the concurrent nature of this study prevents a temporal association between these variables, our analyses represent novel findings about the association between adverse experiences before marriage and subsequent maladaptive characteristics about a marital partner that are associated with concurrent marital dissatisfaction and might enhance the potential for later dissolution.

These results suggest opportunities for future inquiry while also informing theory and clinical practice. First, future research should take advantage of longitudinal data linking one’s own risk and the risk profiles of dating partners, cohabitating partners, and eventual marital partners. In collecting data about psychopathology, relationship risk factors, and the variety of intimate relationship experiences well before marriage (perhaps in adolescence), research might better understand how characteristics of past relationships or relationship partners shape eventual partner selection. Assessing experiences before marriage might also help to elucidate mechanisms of partner selection or examine a more intentional quality to partner selection. Although own risk was a reliable predictor of partner’s risk index, the correlation of r = .26 suggests that there were many individuals whose own risk profile did not match that of their spouse. The moderate level of correspondence between own risk and partner risk indices suggests that mate selection may in fact be an idiosyncratic process, in that two people might assess the same partner and find different things to be attracted to and repelled by.

These results imply further examining what characteristics predict a mismatch between own and partner’s risk index. How do people without an initially risky profile end up with risky mates? Furthermore, what about people who come from adverse experiences but nonetheless choose wisely in their partner selection? Examining these questions might inform theory about how individuals enter into a relationship in order to specifically buffer against their own risk profile or in spite of a relatively low risk profile. In terms of clinical implications, these results encourage interventions aimed at treating the personal and dyadic impact of features like neuroticism, poor self-esteem, and anxious attachment that are related to insecurity about oneself. Our results also suggest that treatment aimed at improving one’s self-esteem and reducing the impact of negative attachment or early-family-of-origin experiences might promote better partner selection or prevent marital problems before they occur or are exacerbated by a partner’s characteristics.