This is an open access article distributed under the terms of the Creative Commons
Attribution License (Several studies explored the connections between experiences within the family-of-origin and adult attachment patterns (Feldman, Gowen, & Fisher, 1998; Lapsley & Edgerton, 2002; Muraru & Turliuc, 2012), the way adults cope with their premarital or marital relationship (Holman, Larson, & Harmer, 1994; Lee & Ok, 2002; Martinson, 2005; Muraru & Turliuc, 2012; Sabatelli & Bartle-Haring, 2003; Whitton et al., 2008), as well as adjustment to couple or family life (Holmes & Anderson, 1994).
Attachment in infancy and childhood has attracted an increasingly interest from both researchers and practitioners (Blount-Matthews & Hertenstein, 2006; Crowell & Treboux, 1995; McCartney & Dearing, 2002; Roth-Hanania & Davidov, 2004). Basically, attachment was conceptualized as an emotional connection that develops between two individuals, one of them being capable of providing protection, comfort and support in times of need (McCartney & Dearing, 2002; Roth-Hanania & Davidov, 2004). Early attachment develops between an infant and a primary caregiver since the first year of life. Since 1980’s, there was a new trend in the research of attachment focusing on working models, as well as on emotional and behavioral styles associated with adults’ romantic attachment (Bartholomew & Shaver, 1998; Fraley & Shaver, 2000; Hazan & Shaver, 1987). In the last three decades, researchers have shown greater interest in conceptualizing the romantic relationship between two adults, taking into account attachment working models that adults acquire since childhood. It has been emphasized that attachment working models evolve over time, depending on the developmental tasks and on the relational experience of each individual (Bartholomew & Shaver, 1998; Crowell & Treboux, 1995; Fraley & Shaver, 2000; Hazan & Shaver, 1987). Although attachment patterns exist also in adulthood, relationship strategies adults use are not the same as those used by infants and children (Parker & Scannell, 1998). Thus, adults tend to have a broader repertoire of behaviors for their romantic relationships than infants do. Furthermore, in adults, romantic attachment patterns have a variety of functions and they are usually characterized by reciprocity, companionship, sexual bonds, and mutual goals (Crowell & Treboux, 1995).
It has been shown that the romantic attachment adults acquire has its roots in the attachment patterns developed during childhood (Crowell & Treboux, 1995; Hazan & Shaver, 1987) and plays an important role in the adults’ psychosocial adjustment (Crowell & Treboux, 1995). Based on John Bowlby and Mary Ainsworth’s models of attachment, Hazan and Shaver (1987) had described three styles of romantic attachment (secure, avoidant, and anxious). Researchers manifested a great deal of interest for the impact romantic attachment has upon the quality of adult couple relationships (Feeney, 1999; Fuller & Fincham, 1995; Marchand, 2004; Timm & Keiley, 2011; Volling, Notaro, & Larsen, 1998). Studies have shown that attachment patterns tend to be associated with variables such as: degree of affection expressed towards the partner, satisfaction regarding marital life, marital conflict-solving styles, control of emotions or frequency of positive emotions expressed by marital partners. Del Giudice (2011) suggested that romantic attachment plays a central role both in the long-term regulation of the affective connections within a marital couple and parenting style, yet this process takes place differently for men and women.
The role gender plays in the marital relationship build-up and evolution could be important, yet it has not been fully understood and empirical evidence is still contradictory (Larson & Holman, 1994). Thus, in a study on gender differences regarding the differentiation of self-experimented within the family-of-origin and the adjustment level in adulthood, Holmes and Anderson (1994) concluded that, for men, the level of differentiation of self in the family-of-origin is significantly associated with the level of subjective well-being, as well as with the level of personal functioning within the family system. However, for women, adjustment indices were not significantly related to the differentiation of self in the family-of-origin. In another study, Holman, Larson, and Harmer (1994) concluded that, only for married men, the better they perceived the quality of family-of-origin environment, the higher their marital relation quality (as measured after one year through the marital satisfaction and its stability). Sabatelli and Bartle-Haring (2003) argued that, when compared to the husbands’ experiences within the family-of-origin, the wives’ experiences in their families-of-origin had stronger relationships both with their own perceptions about marriage and with the perceptions of their partners. In a study aimed at testing a structural model of the relationships among differentiation of self within the marital relationship, romantic attachment styles, communication regarding sex life and sex and marital satisfaction, Timm and Keiley (2011) proved the invariance of the model depending on the participants’ gender.
The Present Study
There is a growing body of literature suggesting that men and women are different concerning how they perceive their family-of-origin experiences (Holmes & Anderson, 1994; Sabatelli & Bartle-Haring, 2003; Volling, Notaro, & Larsen, 1998), romantic attachment pattern (Del Giudice, 2011; Volling, Notaro, & Larsen, 1998), as well as adjustment to romantic relationships (Larson & Holman, 1994; Volling, Notaro, & Larsen, 1998). Nevertheless, there is a lack of literature available on the differences in women and men concerning the role family-of-origin plays in their romantic relationships. The current study aimed at exploring the differences between women and men in relationships among family-of-origin, romantic attachment, and marital adjustment using multiple-group analysis through structural equation modeling (SEM). Along with the role gender plays, all these variables were explored separately, without combining them in one model. Figure 1 depicts our hypothetical working model of relationships among family-of-origin, romantic attachment pattern and marital adjustment. A previous study carried out on 164 married adults revealed that family-of-origin latent variable had a significant effect on romantic attachment which, in its turn, was significantly related to marital adjustment (Muraru & Turliuc, 2012).
Figure 1
The hypothetical model of relationships among family-of-origin, romantic attachment, and marital adjustment.
Method
Participants
Two hundred and forty nine adults completed the measures by which the indicators of latent variables introduced in the model were obtained. There were 79 men and 170 women. The mean age for participants was 35.5 years (SD = 7.8). All participants were Caucasian, most of them had a bachelor degree (80.3%) and were married for the first time (94%). Almost 59% of the participants had a marriage experience of more than five years, having on average 2.4 children (SD = .73).
Measures
Differentiation in the Family System Scale (DIFS, Anderson & Sabatelli, 1992)
Differentiation of self is reflected in the family system’s interactional patterns for maintaining interpersonal distance and, thus, in the tolerance for both individuality and intimacy (Anderson & Sabatelli, 1992). In other words, personal autonomy of the choices a family member makes is promoted in an interactional context based on respect, kindness, affection, and empathy, as well as on cooperative problem-solving (Sabatelli & Bartle-Haring, 2003). The assessment of patterns of family interactions is essential if researchers are to understand the influence of family environment on members and family system development. The 66-item Differentiation in the Family System Scale was used to measure the differentiation of self in the family-of-origin. Participants used a five-point Likert-type scale (1 - never to 5 - always) to rate items, such as “My father responds to my feelings as if they have no value” or “My mother showed respect for my father's viewpoint”. The scale is behavioral focused, so that interactional patterns can be emphasized. Starting from the relational dyads participant-father, father-participant, participant-mother, mother-participant, mother-father, and father-mother three composite scores were derived in order to be used as indicators of differentiation (see Anderson & Sabatelli, 1992). For entire scale, Anderson and Sabatelli (1992) reported alphas ranging from .84 to .94 across three studies. Construct validity of the DIFS was supported by significant correlations between differentiation scores and measures of family conflict and identity status. In the present study, the DIFS subscales proved a satisfactory internal consistency (Cronbach α between .71 and .90).
The Family Adaptability and Cohesion Evaluation Scales
The Family Adaptability and Cohesion Evaluation Scales - FACES III (included in Grey Smith, 1996) was used to measure adaptability and cohesion, as indicators of family-of-origin latent variable. Adaptability is defined as the ability of a family system or marital dyad to change its power structure, roles and rules in response to stressful situation (Olson & Gorall, 2003). Adaptability was measured using ten items (e.g., “Our family changes its way of handling tasks“). Also, FACES III allows the assessment of cohesion which may be viewed as the extent to which members of a family system are emotionally connected. Murray Bowen refers to this dimension by using the word “togetherness” (cf. Nichols & Schwartz, 2001). Cohesion is linked to other facets of family life, such as: coalitions, interests, recreational activities, decision-making or family boundaries (Olson & Gorall, 2003). The cohesion was measured using other ten items (e.g., “We like to do things with just our immediate family”). Answers were provided on a five-point Likert-type scale ranging from 1 (never) to 5 (almost always). For a non-clinical norming sample, D. H. Olson, J. Portner and Y. Lavee reported internal consistencies as high as .62 for adaptability and .77 for cohesion (cf. Grey Smith, 1996). Authors of the original scale reported a non-significant and negligible correlation (r = .03) between cohesion and adaptability, indicating independence of the two dimensions. The Cronbach α realiability for the current study was .79 for adaptability and .89 for cohesion.
Experiences in Close Relationships Scale (ECR, Brennan, Clark, & Shaver, 1998)
In the present study, romantic attachment was measured using a short version of The Experiences in Close Relationships Scale (Wei, Russell, Mallinckrodt, Vogel, 2007). According to data reported by Wei et al. (2007), ECR displayed satisfactory psychometric qualities. The scale includes 12 items to which participants answered using a seven-point Likert-type scale ranging from 1 (strong disapproval) to 7 (strong approval). Six of the items focus on anxiety as a romantic attachment pattern (e.g., “I need a lot of reassurance that I am loved by my partner”), while the other six focus on avoidance (e.g., “I am nervous when partners get too close to me”). Across six studies carried out on college students samples, Wei et al. (2007) reported internal consistencies equal to .78-.88 for avoidance and .77-.86 for anxiety. Also, confirmatory factor analyses indicated two factors and a good fit to the data for measurement models across studies. In the present study, Cronbach α reliability was .82 for avoidance and .66 for anxiety.
Revised Dyadic Adjustment Scale (RDAS, Busby et al., 1995)
Marital adjustment has been described as a dynamic process, the outcome of which is determined by the degree of: a) troublesome differences in marital relationship; b) interpersonal tensions and personal anxiety; c) satisfaction of marital partners; d) cohesion; e) consensus on important couple’s matters (Spanier, 1976). A 32-item scale (Dyadic Adjustment Scale/DAS) has been developed to operationalize four dimensions of marital adjustment: consensus, satisfaction, cohesion and affectional expression. In the present study, the 14-items Revised Dyadic Adjustment Scale was used to measure marital adjustment. This instrument is a psychometrically improved shortened version of DAS (Spanier, 1976). Participants answered on a six-point Likert-type scale, except for one item formatted on a five-point scale. Consensus on matters of importance to marital relationship (decision making, values, and affection) was measured using six items. Four items (e.g., “How often do you discuss terminating your relationship?”) measured marital satisfaction. This dimension refers to stability of a couple and conflicts between spouses. Other four items (e.g., “How often do you have a stimulating exchange of ideas?”) operationalized marital cohesion which included discussions and activities engaging both marital partners. The RDAS was selected because it had proved good psychometric qualities (Busby, Christensen, Russell Crane, & Larson, 1995). In the original sample, values of internal consistency were .81 for consensus, .85 for satisfaction, .80 for cohesion, and .90 for total RDAS. Busby et al. (1995) reported a correlation equal to .97 (p < .001) between RDAS and DAS. Also, authors provided evidence for criterion-related validity of RDAS, suggesting that RDAS and DAS are equal in their ability to correctly classify cases as either distressed or non-distressed married adults. In the present study, the internal consistency values were: .83 for marital consensus, .92 for marital satisfaction, and .77 for marital cohesion.
Procedure
Participants were recruited by a convenience non-probabilistic sampling plan from various occupational settings (e.g., education, public administration, services etc.). The participation was voluntary and based on informed consent. Participants were asked to complete anonymously the four self-reporting scales in their homes. Questionnaires were administered in the same order: DIFS - FACES III - ECR - RDAS. Eighty percent of protocols with responses were returned to investigators. The rest of the participants did not respond to our request to return the protocols, without giving any reason for doing so. No other additional information was obtained from these participants. A supplementary number of 12 protocols were excluded from subsequent analyses because of missing responses. Our initial aim was to collect data from a large sample of marital couples but, due to the sampling constraints, only a small portion of participants was represented by spouses. Thus, the unit of analysis was the individual.
Statistical Analyses
The means and standard deviations of observed variables for females and males were computed and the Student’s t-Test for independent samples was used in order to perform comparisons by gender. For each comparison, the effect size was estimated using the Cohen’s d coefficient (Cohen, 1992). Cohen’s d values around .20 indicate a small effect size, values around .50 indicate a moderate effect size, while values as high as .80 indicate a large effect size. For both females and males, zero-order correlations among the indicators of all latent variables were computed using the product moment correlation coefficient (r).
In order to estimate the parameters and degree of statistical fitting for our SEM model across gender, multiple-group analysis through maximum likelihood (ML) technique was used (Byrne, 2010). The SEM with ML procedure assumes that observed variables have normal distributions (Bowen & Guo, 2012; Byrne, 2010). In the present study, the univariate normality of distribution for each observed variable was assessed by examining the skewness and kurtosis values. The SPSS package uses 0 as a reference value for skewness and kurtosis, to decide if a distribution is normal or not. There is no clear cutoff to indicate an acceptable level of skewness and kurtosis (Byrne, 2010; West, Finch, & Curran, 1995). In a conservative approach, the researcher might conclude that a skewness value greater than 1 or less than – 1 is problematic (Bowen & Guo, 2012). The value of skewness tends to impact tests of means. More problematic than skewness is kurtosis, which severely impacts tests of variances and covariances. As in the case of skewness, if kurtosis is greater than 1 or less than -1 (in software packages using 0 as reference value for a normal distribution), one might conclude that the distribution could be problematic. However, several simulations (Muthén & Kaplan, 1985; West, Finch, & Curran, 1995) have found significant problems in the ML procedures arising with univariate skewness at least equal to 2 and kurtosis to 7 (in software packages using 3 as reference value for a normal distribution) or 4 (in programs using 0 as cutoff). One approach to handle the presence of non-normal observed variables in SEM suggests the use of a bootstrapping procedure which yields more accurate ML estimates of parameters from a model (Byrne, 2010).
The fitting between the configural model and observed data was assessed by examining the following indexes (Browne & Cudeck, 1993; Byrne, 2010; Hu & Bentler, 1998; MacCallum & Austin, 2000): χ2 (critical level of significance was set at .05, two-tailed), Bentler-Bonnett normed fit index (NFI) and non-normed (Tucker-Lewis) fit index (NNFI/TLI), comparative fit index (CFI) and the classical root mean square error of approximation (RMSEA). RMSEA is one of the most important indicators showing the degree to which estimated parameters of a SEM model are representative for the whole population from which the sample was drawn. Since RMSEA is sensitive to misspecifications of relationship among variables and it is accompanied by a confidence interval which provides an indication of precision of estimation, its use in applied research is strongly encouraged (MacCallum & Austin, 2000). A non-significant χ2 value, as well as values greater than .95 for NFI, NNFI and CFI, and RMSEA value lower than .05 indicate a good fit of the configural model to the observed data (Byrne, 2010). Following suggestions from the literature, it was considered that a value of RMSEA as high as .08 indicates an acceptable fit of the SEM model (Browne & Cudeck, 1993). Also, TLI and CFI values ranging from .90 to .95 indicate an acceptable model fit (Hu & Bentler, 1998).
The configural model provides the baseline to which all subsequent tests for invariance are compared. The classical approach in arguing for evidence of invariance is based on the χ2 difference (∆χ2) test (Byrne, 2010). The value ∆χ2 represents the difference between the χ2 value for the configural model and the χ2 values for the other models in which equality constraints have been imposed on particular parameters. According to Byrne (2010), evidence of invariance is claimed if ∆χ2 value is not statistically significant. In the criticism of the ∆χ2 procedure, some researchers have argued that, from a practical perspective, the χ2 difference test represents an excessively stringent test of invariance, because SEM models, at best, are only approximations of reality (Byrne, 2010). It was consequently argued that it may be more reasonable to base decision on a difference in CFI rather than on the ∆χ2 value (Cheung & Rensvold, 2002; cited in Byrne, 2010). Evidence of invariance should be based on a ∆CFI value not exceeding .01. Although this practical approach has been criticized, its use is increasingly reported in the family and couple psychology literature (Lucas et al., 2008; South, Krueger, & Iacono, 2009).
Results
Preliminary Analyses
Table 1 shows descriptive values for distributions of all observed variables. Participants tended to obtain a higher score in the differentiation between them and their mothers, while the score in differentiation between their parents was lower and more heterogeneous. The average anxious attachment score was higher than the one for avoidance, although the distributions for both dimensions displayed the same amount of variation.
Table 1
Descriptive Statistics for the Observed Variables
| Variables | M | SD | Min | Max | Skewness | Kurtosis | |
|---|---|---|---|---|---|---|---|
| 1. | Family-of-origin cohesion | 37.57 | 7.66 | 14 | 50 | -.83 | .13 |
| 2. | Family-of-origin adaptability | 24.88 | 6.38 | 10 | 46 | -.04 | -.29 |
| 3. | Participant-father differentiation | 1772.44 | 598.56 | 240 | 3025 | -.16 | -.72 |
| 4. | Participant-mother differentiation | 1811.16 | 527.82 | 504 | 3025 | -.08 | -.54 |
| 5. | Mother-father differentiation | 1750.71 | 678.11 | 180 | 3025 | -.21 | -.87 |
| 6. | Avoidant attachment | 12.31 | 6.18 | 6 | 34 | 1.35 | 1.55 |
| 7. | Anxious attachment | 17.93 | 6.41 | 6 | 40 | .59 | .19 |
| 8. | Marital consensus | 23.69 | 4.48 | 5 | 30 | -1.11 | 1.64 |
| 9. | Marital satisfaction | 11.76 | 6.18 | 0 | 20 | -.49 | -1.39 |
| 10. | Marital cohesion | 12.07 | 3.75 | 1 | 19 | -.40 | -.15 |
Family-of-origin adaptability, participant-father, participant-mother, as well as mother-father differentiation had quasi-symmetric distributions. Marital satisfaction and cohesion, as well as anxious attachment showed modest departures from the reference value for skewness, while the distribution for family-of-origin cohesion had a more prominent skew. Only avoidant attachment and marital consensus had a skewness above the absolute value of 1. On the other hand, family-of-origin cohesion, anxious attachment and marital cohesion had quasi-kurtic distributions, while participant-father and mother-father differentiation were platykurtic in their distributions. Avoidant attachment and marital consensus had more prominent leptokurtic distributions (kurtosis > 1), while marital satisfaction was more platykurtic. Thus, none of the skewness values for observed variables were greater than 2, while none of the kurtosis values were greater than 7. Therefore, we considered that the distributions of the observed variables did not significantly depart from normality. Furthermore, we used the ML procedure without performing the bootstrapping.
There were no significant differences between females and males in all observed variables (see Table 2). Cohen’s d values were close to zero, indicating negligible effect sizes.
Table 2
Gender Comparisons in Observed Variables
| Variables | Females
|
Males
|
t | p | D | |||
|---|---|---|---|---|---|---|---|---|
| M | SD | M | SD | |||||
| 1. | Family-of-origin cohesion | 37.45 | 8.07 | 37.82 | 6.73 | -.34 | .728 | .05 |
| 2. | Family-of-origin adaptability | 24.97 | 6.82 | 24.69 | 5.36 | .35 | .726 | .04 |
| 3. | Participant-father differentiation | 1779.47 | 625.57 | 1757.32 | 539.32 | .27 | .786 | .04 |
| 4. | Participant-mother differentiation | 1814.85 | 552.84 | 1803.24 | 472.73 | .16 | .872 | .02 |
| 5. | Mother-father differentiation | 1722.57 | 721.24 | 1811.26 | 574.02 | -1.04 | .298 | .13 |
| 6. | Avoidant attachment | 12.24 | 6.18 | 12.46 | 6.22 | -.26 | .788 | .04 |
| 7. | Anxious attachment | 18.02 | 6.14 | 17.72 | 6.99 | .35 | .725 | .05 |
| 8. | Marital consensus | 23.68 | 4.42 | 23.69 | 4.65 | -.01 | .990 | .00 |
| 9. | Marital satisfaction | 11.88 | 6.10 | 11.49 | 6.38 | .46 | .640 | .06 |
| 10. | Marital cohesion | 11.99 | 3.87 | 12.24 | 3.48 | -.48 | .631 | .07 |
Table 3 displays the zero-order correlations among all observed variables. The correlations were computed separately for each gender. There were significant negative correlations among family-of-origin indicators, avoidant attachment and anxious attachment. For males, only three out of 11 correlations were significant (p < .05, two-tailed). For both female and male participants, the strongest correlations were among family-of-origin indicators and consensus, as an indicator of marital adjustment.
Table 3
Zero-Order Correlations Among Observed Variables
| Variables | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. | Family-of-origin cohesion | - | |||||||||
| 2. | Family-of-origin adaptability | .56*** (.25*) | - | ||||||||
| 3. | Participant-father differentiation | .67*** (.57***) | .45*** (.36**) | - | |||||||
| 4. | Participant-mother differentiation | .56*** (.58***) | .46*** (.28*) | .59*** (.79***) | - | ||||||
| 5. | Mother-father differentiation | .75*** (.66***) | .48*** (.23*) | .81*** (.83***) | .66*** (.75***) | - | |||||
| 6. | Avoidant attachment | -.15* (-.26*) | -.20** (-.11) | -.29*** (-.16) | -.27*** (-.11) | -.32*** (-.15) | - | ||||
| 7. | Anxious attachment | -.05 (-.07) | -.14 (-.26*) | -.15** (-.21) | -.26*** (-.12) | -.14 (-.15) | .42*** (.48***) | - | |||
| 8. | Marital consensus | .28*** (.37**) | .25** (.20) | .36*** (.24*) | .29*** (.25*) | .39*** (-.26*) | -.68*** (-.67***) | -.49*** (-.42***) | - | ||
| 9. | Marital satisfaction | .07 (-.10) | -.01 (.24*) | .12 (-.03) | .22** (.16) | .15* (-.01) | -.11 (.04) | -.10 (-.005) | .06 (.01) | - | |
| 10. | Marital cohesion | .12 (.31**) | .27*** (.09) | .22** (.15) | .18* (.15) | .19** (.19) | -.50*** (-.62***) | -.35*** (-.35**) | .54*** (.71***) | .04 (.03) | - |
Note. Correlations placed out of brackets are for females, while the correlations placed inside the brackets are for males.
*p < .05. **p < .01. ***p < .001.
Except for cohesion, all family-of-origin indicators correlated significantly and positively with marital cohesion in females. For males, there was a significant correlation only between family-of-origin cohesion and marital cohesion.
Regardless of gender, the correlations between family-of-origin indicators and marital satisfaction were inconsistent. For both sexes, the pattern of associations among avoidant/anxious romantic attachment and indicators of marital adjustment was similar. There were significant correlations among romantic attachment styles and marital consensus and cohesion.
Exploring the Invariance Across Gender
The current findings suggest that emphasizing stereotyped differences between women and men is not fruitful in clinical practice. Family and couple therapists could rather work on deconstructing the social discourses about gender differences, enabling a sense of personal agency for their clients. In the present study, romantic attachment was a negative predictor of marital adjustment for both women and men. These results are consistent with those reported in previous studies that have examined the relationship between adult romantic attachment and indicators of marital quality (Feeney, 1999; Marchand, 2004; Muraru & Turliuc, 2012; Volling, Notaro, & Larsen, 1998). In agreement with the data reported in a prior study (Muraru & Turliuc, 2012), but contrary to the study carried out by Sabatelli and Bartle-Haring (2003), the family-of-origin had no significant contribution to the marital adjustment prediction for either women or men who participated in the present study.
When the configural (unconstrained) model was taken as reference, marital satisfaction was not a significant indicator of marital adjustment, neither for women nor for men (see Table 4). The other nine observed variables were significant indicators of latent variables. For both women and men, romantic attachment was a significant and negative predictor of marital adjustment. Otherwise, family-of-origin had no significant contribution to the prediction of marital adjustment. Family-of-origin was also a significant predictor of the romantic attachment only in females. The configural model proved a satisfactory fit to the observed data across gender groups: χ2 = 109.87, df = 64, p < .001, CFI = .960, NFI = .991, NNFI = .943, RMSEA = .054 (90% CI = .036; .071). When measurement and structural weights, structural covariances and residuals were constrained to be equal across gender groups and the Δχ2 criterion was used, multiple-group analyses yielded four models. All of them indicated a satisfactory fit to the observed data (see Table 5). The differences between the χ2 value for the configural model and the χ2 value for constrained models were not statistically significant (p > .05), thus supporting the invariance of hypothesized model across gender groups. Moreover, the values of ΔCFI were lower than .01, strengthening the evidence of invariance.
Table 4
Measurement and Structural Weights for the Configural Model of Relationships Among Family-of-Origin, Romantic Attachment, and Marital Adjustment
| Hypothetical measurement and structural relationships |
Females
|
Males
|
||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Unst. est. | S.E. | St. est. | C.R. | p | Unst. est. | S.E. | St. est. | C.R. | p | |
| Family-of-origin → cohesion | .01 | .001 | .80 | 14.36 | < .001 | .009 | .001 | .68 | 7.32 | < .001 |
| Family-of-origin → adaptability | .006 | .001 | .55 | 8.03 | < .001 | .003 | .001 | .33 | 3.00 | .003 |
| Family-of-origin → participant-father differentiation | .79 | .04 | .86 | 16.61 | < .001 | .94 | .07 | .91 | 12.55 | < .001 |
| Family-of-origin → participant-mother differentiation | .58 | .05 | .71 | 11.59 | < .001 | .76 | .07 | .84 | 10.67 | < .001 |
| Family-of-origin → mother-father differentiation | 1.00 | - | .94 | - | - | 1.00 | - | .91 | - | - |
| Romantic attachment → avoidant style | 1.00 | - | .76 | - | - | 1.00 | - | .88 | - | - |
| Romantic attachment → anxious style | .71 | .10 | .54 | 6.89 | < .001 | .68 | .16 | .54 | 4.26 | < .001 |
| Marital adjustment → cohesion | .62 | .07 | .62 | 8.07 | < .001 | .67 | .08 | .80 | 7.69 | < .001 |
| Marital adjustment → consensus | 1.00 | - | .87 | - | - | 1.00 | - | .89 | - | - |
| Marital adjustment → satisfaction | .18 | .13 | .11 | 1.44 | .149 | .01 | .18 | .007 | 0.06 | .951 |
| Family-of-origin → romantic attachment | -.003 | .001 | -.39 | -4.16 | < .001 | -.002 | .001 | -.22 | -1.71 | .086 |
| Family-of-origin → marital adjustment | .0001 | .001 | .05 | .58 | .560 | .001 | .001 | .13 | 1.33 | .183 |
| Romantic attachment → marital adjustment | -.83 | .12 | -1.01 | -6.52 | < .001 | -.62 | .12 | -.82 | -4.80 | < .001 |
Note. Unst. est. – unstandardized estimate; S.E. – standard error; St. est. – standardized estimate; C.R. – critical ratio.
Table 5
Tests for Invariance of Hypothetical Model
| Model | χ2 | df | Δχ2 | Δdf | CFI | ΔCFI | NFI | NNFI/TLI | RMSEA | 90% CI of RMSEA |
|---|---|---|---|---|---|---|---|---|---|---|
| Unconstrained model (baseline) | 109.87 *** | 64 | - | - | .960 | - | .911 | .943 | .054 | .036; .071 |
| Measurement constrained | 122.08 *** | 71 | 12.21 ns | 7 | .955 | .005 | .901 | .943 | .054 | .037; .070 |
| Structural weights constrained | 123.51 *** | 74 | 13.64 ns | 10 | .957 | .003 | .899 | .947 | .052 | .035; .068 |
| Structural covariance constrained | 126.91 *** | 75 | 17.04 ns | 11 | .954 | .006 | .897 | .945 | .053 | .037; .069 |
| Structural residuals constrained | 129.52 *** | 77 | 19.65 ns | 13 | .954 | .006 | .895 | .946 | .053 | .036; .068 |
***p < .001. ns – non-significant differences from unconstrained model.
Discussion
Our study was designed to be exploratory. The purpose was not to test any hypothesis regarding invariance across gender when it comes to relationships among family-of-origin, romantic attachment, and marital adjustment, but to explore if there is a difference regarding the impact early family-of-origin experiences have on later adjustment to romantic relationships in the case of women and men. By incorporating suggestions from well-known couple and family therapists (Nichols & Schwartz, 2001; Sabatelli & Bartle-Haring, 2003), our hypothetical model assumed that experiences individuals undergo in their families-of-origin represent a legacy which tends to influence adjustment to developmental tasks throughout their lifespan, including to romantic relationships. From an intergenerational perspective, Murray Bowen has developed a comprehensive theoretical framework to explain how constructions individuals have about their families-of-origin are reflected in the choices they make in everyday life, including their romantic relationships (cf. Nichols & Schwartz, 2001). Bowen’s theory has influenced the practice of many couple and family therapists who emphasized the role family resources have in the process of rewriting stories related to the personal lives of clients. In line with suggestions from the literature, in the current study it was assumed that constructions married adults hold about their families-of-origin tend to be associated both with patterns of romantic attachment and their adjustment to marital relationships.
Data from a multiple-group analysis using structural equation modeling suggested the invariance of our hypothetical model across gender. The data revealed that romantic attachment had a significant negative effect upon marital adjustment for both women and men. This result is consistent with previous studies (Feeney, 1999; Marchand, 2004; Muraru & Turliuc, 2012; Volling, Notaro, & Larsen, 1998). There is a growing body of literature suggesting that romantic attachment patterns impact the quality of a couple’s relationship (Feeney, 1999; Marchand, 2004; Volling, Notaro, & Larsen, 1998). In a marital couple, a secure attachment expressed by the two spouses implies an active, affectionate, and reciprocal relationship in which both marital partners mutually provide closeness and emotional comfort. Adults expressing a secure attachment toward their partners tend to have relationships characterized by high levels of reciprocal trust, support, and intimacy, as well as shared feeling and ideas (Hazan & Shaver 1987; Simpson et al. 1992). In contrast, adults with prevailing insecure attachment tend to fear abandonment, to experience emotional ups and downs, to be obsessively jealous, and overly dependent on their partner (Parker & Scannell, 1998). Adults with an avoidant attachment pattern tend to deny attachment needs, are reluctant to trust others, avoid closeness and, often, are overinvolved with activities, such as professional work. In a relationship, two adults tend to bring with them their own working models about attachment and related behaviors. Thus, the relationship tends to be shaped by the partners’ style of relating. Moreover, the interaction of their personal styles tends to influence the way the two partners are experiencing their relationship.
Family-of-origin had no significant effect upon marital adjustment for either women or men. At least one explanation might be advanced in relation to this result. In Romania, the family-of-origin exerts a great influence on the way adults give meanings to their lives and relationships. The postmodern tendencies regarding changes in the pattern of family values are rather weak, being characteristic for a minority (e.g., young people from urban areas) within the Romanian population (Popescu, 2008). However, in the present study, over 87% of the participants were under forty, representing the cohort born since 1970. When democracy was reinstated in Romania, these participants were around 20 years old. Therefore, we assumed that the majority of participants had opportunities to be exposed to the new ideological representations about family and its role in society. This could have an impact on how participants in our study represent their families-of-origin and the role early experiences play in later personal development. In addition, in the present study, FACES and DIFS were adapted to allow the exploration of constructions participants had about their own family-of-origin experiences. This approach involved an effort to remember the early past, which could be impacted by the subjectivity of participants, regardless of their gender. At the same time, the RDAS asked participants to assess their current marital relationship. It is possible that the two factors above have a significant contribution to the way adults surveyed in this study perceived their own past (including the meaning of experiences specific for their families-of-origin) and related it to their perception of the current marital relationship. However, further explanations for this finding remain to be found.
Moreover, no significant effect of the family-of-origin upon romantic attachment was revealed in the case of male participants. This finding is consistent with data reported by Feldman, Gowen, and Fisher (1998). Why was the family-of-origin significantly related to romantic attachment only for women? Searching for an explanation, we presumed that gender could moderate the relationship between family-of-origin experiences and romantic attachment styles. Thus, we observed that correlations among differentiation of self within family-of-origin, adaptability, cohesion, and avoidant attachment were negative and significant only for women. Additionally, the correlations for men were lower than those for women. As far as anxious attachment was concerned, the pattern of correlations was quite similar, except for differentiation between participants and their fathers (for this variable the correlation was slightly greater for men, even if non-significant). The moderating role of gender could be related both to working models of attachment (Crowell & Treboux, 1995) and transitional changes over time, which, in conjunction with family-of-origin experiences, could have a different impact on the functioning of women and men. Although, regardless of gender, working models and romantic attachment behaviors (especially those related to the secure attachment) tend to be relatively stable over a short-term period (Fuller & Fincham, 1995; Lopez & Gormley, 2002; Scharfe & Bartholomew, 1994), relationship experiences can initiate some changes over time. Thus, getting married usually involves a major transition for the individual and many changes are likely to occur over the early transitional years of marital life. Women and men may differently integrate the legacy of experiences they undergo in their family-of-origin into the transitional stage of the first years of marriage, including in terms of restructuring models of romantic attachment. It is possible that, in structuring working models of romantic attachment, women tend to be more rooted in experiences acquired within their family-of-origin, while men tend to avoid recognizing the connection between the “relational present” and their early experiences. It is just one of the hypotheses that we intend to explore in future studies.
A limitation of the current study was the disproportion between number of male and female participants that could affect the accuracy of estimated parameters. Secondly, the findings were based on data from a relatively small and homogeneous (e.g., residence or level of education) convenience sample of participants, thus limiting the generalization of results. Adults married for the first time, living in urban areas and having a bachelor degree, may not be representative of all married adults. A replication of the present study on a more extended and heterogeneous sample of adults is needed. Another limitation refers to the unit of analysis, which was the individual. A study conducted on marital couples, using self-reports and structured interviews with both partners, would allow nuancing the explanations for possible differences or similarities when it comes to relationships among family-of-origin, adult attachment, and marital life. The marital couple seems to be a more appropriate unit of analysis, because it is more probable for the two spouses to be tied by an emotional and transactional bond, including by past experiences. As far as the structured interview is concerned, such a technique provides the clinician the opportunity to acquire a deeper understanding of personal shared meanings partners give to their relationship and, also, to past experiences. Thus clinicians, together with the couples, may co-construct a coherent exploration regarding the legacy of the family-of-origin. Finally, the cross-sectional nature of our design limits the possibility to infer causal relationships among variables included in the hypothetical model.