Maladaptive personality traits and older adult relationship satisfaction: A co-twin control approach to understanding associations
Abstract
Objective
Maladaptive personality traits have been implicated in romantic relationship dissatisfaction, but the etiology of those links and the degree to which they extend to other types of relationships are unclear. The purpose of this study was to examine associations between maladaptive personality traits and satisfaction in various relationships using a co-twin control design to identify potential environmental contributions.
Method
The sample consisted of 1340 older adult twin participants from the Minnesota Twin Registry (Mage = 70.3) that completed the Personality Inventory for DSM-5 Faceted Brief Form and Network of Relationships Inventory (Revised for Older Adults).
Results
Several maladaptive personality traits were phenotypically associated with relationship dissatisfaction, with detachment and negative affect having the largest effects. Further, within twin pair differences in detachment and negative affect were associated with greater relationship dissatisfaction, suggesting that observed associations were mediated partly by the unique environment, not solely the result of genetic and familial confounding. Both phenotypic and co-twin associations were strongest overall in the romantic partner relationship.
Conclusion
These findings support the notion that maladaptive personality traits are implicated in interpersonal dysfunction across multiple domains.
1 INTRODUCTION
Decades of research have explored factors associated with higher levels of satisfaction in close relationships and personality has consistently emerged as a relevant construct (Caughlin et al., 2000; Tavakol et al., 2017; White et al., 2004). The Big Five domains of personality have long been established as reliable predictors of romantic relationship satisfaction (Malouf et al., 2010; but see Holland & Roisman, 2008). More specifically, low neuroticism, high agreeableness, high conscientiousness, and high extraversion are all associated with greater relationship satisfaction and marital success (Schaffhuser et al., 2014). In addition to being consistently associated with normative personality domains, relationship dissatisfaction has been associated with personality disorders (PD; Bouchard et al., 2009; Chen et al., 2004; Stroud et al., 2010; Wilson et al., 2017). For example, disorders conceptualized as being in Cluster B in the DSM-IV's categorical approach to PDs (e.g., narcissistic and borderline personality disorders), in particular, have been associated with sustained interpersonal dysfunction and conflict (Chen et al., 2004).
The fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) includes two frameworks for diagnosing personality disorders (American Psychiatric Association, 2013). The first is a traditional, categorical framework essentially identical to the approach taken in the fourth edition (DSM-IV). The second entails maladaptive personality functioning (Criterion A) and trait dimensions (Criterion B) and is described as the Alternative Model of Personality Disorders (AMPD). In creating the AMPD's Criterion B, the DSM-5 Personality Disorders Workgroup focused on delineating maladaptive trait domains (Krueger et al., 2012). The Personality Inventory for DSM-5 (PID-5) assessment instrument was then created to provide clinicians and researchers with a tool for assessing those traits (Krueger et al., 2012; Markon et al., 2024). The maladaptive personality model has five higher-order domains and 25 lower-order facets. The domains are Negative Affect, Detachment, Antagonism, Disinhibition, and Psychoticism. These domains have been shown empirically to be well conceptualized as the pathologically low pole (usually maladaptive) of the Five-Factor Model (FFM; usually adaptive) domains (e.g., Suzuki et al., 2015). These classifications are emotional stability and neuroticism (negative affectivity), extraversion and detachment, agreeableness and antagonism, and conscientiousness and disinhibition. Although the AMPD psychoticism domain is not equivalent to FFM openness, psychoticism is empirically associated with the high creativity aspect of openness (Suzuki et al., 2015). Overall, the pathological pole of major personality domains is associated with lower quality of life compared to the adaptive pole emphasized by the FFM (Hobbs et al., 2023).
Higher levels of personality pathology have been associated with increased intimate partner aggression and dissatisfaction for both individuals and their partners (Smith et al., 2020). Smith et al. (2020) demonstrated that individuals who had higher levels of negative affectivity were less satisfied with their relationships and tended to commit more psychological intimate partner aggression. Similarly, participants who rated their partners higher in negative affectivity were more likely to be dissatisfied in their relationships. Higher levels of detachment and psychoticism were also linked to greater dissatisfaction in marital relationships. Decuyper et al. (2018) likewise found that detachment and negative affectivity had the most consistent negative associations with relationship satisfaction. Negative affectivity may contribute to relationship dissatisfaction through emotional dysregulation. Among the normative traits of personality, neuroticism is the most consistent predictor of relationship dissatisfaction (Decuyper et al., 2018).
Most personality pathology and relationship satisfaction research has been conducted using designs that offer little insight into the etiology of the association. For instance, an underlying genetic liability toward both personality pathology and relationship satisfaction may confound the observed cross-sectional associations. Twin studies can be used to explore the environmental basis of individual differences (McGue et al., 2010). A co-twin control approach can offer stronger causal inferences by controlling for genetic and shared environmental effects by comparing differences both between and within twin pairs. For example, if twins with higher personality pathology have lower relationship satisfaction compared to their twin who has lower levels of personality pathology, this would suggest that the relationship between personality traits and relationship satisfaction is not strictly due to a shared genetic liability but rather is partially mediated in the unique environment, consistent with personality having a causal effect. On the other hand, if relationship satisfaction is comparable between twins with different levels of personality pathology, the association might be better explained by a shared underlying liability towards dysfunction in interpersonal relationships (Wilson et al., 2018). Identical (MZ) twin pairs allow for the greatest control for confounding variables because they share both genetic and shared environmental effects. Nevertheless, other unmeasured confounders may be present, such that, although studying MZ pairs provides inferential leverage beyond that provided by studying unrelated persons, it is only one design that may contribute to maximizing the quality of our causal inferences in observational research (McGue et al., 2010).
In a co-twin study, Wilson et al. (2018) found that higher levels of maladaptive personality traits were associated with lower relationship satisfaction beyond the similarities due to environmental and genetic factors shared by twins. At the individual level, each of the five domains was negatively associated with romantic relationship satisfaction. After controlling for shared factors, higher levels of negative affectivity, detachment, disinhibition, and psychoticism were associated with lower levels of relationship satisfaction in couples. There was no significant association between within-twin pair differences in antagonism and relationship satisfaction, suggesting genetic or shared familial confounding may be driving the individual-level association. Although other confounders are possible, the associations with the other personality domains were not explained strictly by shared genetic factors, consistent with those traits having a causal impact on relationship dissatisfaction.
Since the majority of relationship satisfaction research has focused on the FFM and romantic relationships in younger samples, this study aims to replicate the Wilson et al. (2018) results and extend these findings to an older sample. Older adulthood is a particularly relevant time to study personality and interpersonal functioning due to the unique challenges in late life, such as increased health concerns that can exaggerate affective or behavioral tendencies characteristic of personality disorders (Oltmanns & Balsis, 2011). Additionally, social support has also been linked to health outcomes by impacting perceptions of health issues, coping skills, and positive health behaviors. Personality pathology can play a role in late-life health outcomes by causing impairments in social functioning (Oltmanns & Balsis, 2011).
This study also expands on the Wilson et al. (2018) results by comparing levels of satisfaction in other close relationships beyond the romantic partner. This allows for a more comprehensive understanding of how maladaptive personality traits impact the quality of social connections. Given that maladaptive traits are associated with lower romantic relationship satisfaction and general interpersonal impairment, it would be reasonable to expect similar associations for other types of relationships. The Network of Relationships Inventory - Social Provisions Version (NRI-SPV; Furman & Buhrmester, 2009), allows for aspects of relationship quality to be assessed and compared across different relationships in individuals' social networks. Further, we aim to explore the etiology of those associations by using a co-twin control approach. It is expected that individuals with higher levels of personality pathology will report lower levels of satisfaction in each category of interpersonal relationships and that there will be significant within-pair difference effects indicating these associations are mediated in the unique environment.
2 METHOD
2.1 Participants
Participants were 1340 individuals from the Minnesota Twin Registry, which has recently been rebranded the Multi-Ethnic Twin Registry at the University of Minnesota (MTR). The original MTR was a birth-record-based registry that recruited twin pairs born in Minnesota from 1904 to 1964 (Krueger & Johnson, 2002; Lykken et al., 1990). The participants were recruited through locating birth records and comparing it to death records 6 months following the birth (to eliminate any pairs that did not survive infancy). The most recent wave of data collection occurred from 2017 to 2022 and consisted of twins born between 1936 and 1955. Participant ages ranged from 62 to 85 (M = 70.26, SD = 5.45) and consisted of both female and male twins (62.5% female). Consistent with the demographics of Minnesota births during this period, participants in the present study were predominantly white (99.5%). There are 416 intact MZ twin pairs used in this sample and 254 same-sex DZ twin pairs; participants whose twin did not provide any data on key constructs were removed from the original cohort of 1377 participants. Additionally, participants who had incomplete data for a key construct were removed from analyses in that given NRI relationship type (i.e., without imputation). In these cases, the participant's twin was also removed from analyses of that relationship type in order to keep individual and co-twin-level analyses consistent for sample sizes. However, those cases were rare, and missing data primarily related to the prevalence of different types of relationships in the sample (e.g., it was more common to report having a child than an opposite-sex friend). A total number of twin pairs used in each relationship type for MZ individual and co-twin-level analyses can be found in Table 4.
MZ twin results are considered the primary analyses because their shared genetic and environmental factors offer a built-in control to analyze the impact of unique environmental effects. DZ twins provide an imperfect control for genetic confounds, but comparisons of MZ and DZ discordance can provide additional insight into the role of additive genetics and common environment. Additionally, the NRI assesses support and negative features across different relationships in a participant's social network, so analyses could only be conducted when a twin endorsed having these relationships. The PID-5 and NRI measures were collected as part of a larger, comprehensive participant assessment supported by the National Institute on Aging with longitudinal follow-up underway.
2.2 Measures
2.2.1 Personality inventory for DSM-5 faceted brief form
The PID-5-FBF is a 100-item self-assessment scale that assesses the maladaptive domains and facets of AMPD (Maples et al., 2015). The five domains are negative affectivity, detachment, antagonism, disinhibition, and psychoticism. The PID-5 was scored by averaging answers from each lower-order facet that contributed to the higher-order domains (see Table 1). Example items for the domains include ‘I worry a lot about terrible things that might happen’ (negative affect), ‘I don't like spending time with others’ (detachment), ‘I use people to get what I want’ (antagonism), ‘I always do things on the spur of the moment’ (disinhibition), and ‘I have seen things that weren't really there’ (psychoticism). Each question asked the participant to rate the accuracy as Very or Often False (0), Sometimes or Somewhat False (1), Sometimes or Somewhat True (2), or Very or Often True (3). In the sample, the PID-5 had high internal consistency across domains (0.77 < α < 0.86).
| Variable | Mean (SD) | Range | α |
|---|---|---|---|
| Negative affect | 0.45 (0.45) | 0.00–2.16 | 0.86 |
| Detachment | 0.50 (0.49) | 0.00–2.33 | 0.86 |
| Antagonism | 0.23 (0.27) | 0.00–1.67 | 0.77 |
| Disinhibition | 0.36 (0.34) | 0.00–1.75 | 0.81 |
| Psychoticism | 0.17 (0.27) | 0.00–1.42 | 0.79 |
| Romantic partner support features | 4.09 (0.72) | 1.09–5.00 | 0.95 |
| Romantic partner negative features | 1.66 (0.64) | 1.00–5.00 | 0.90 |
| Twin support features | 3.31 (0.82) | 1.00–4.95 | 0.95 |
| Twin negative features | 1.48 (0.69) | 1.00–5.00 | 0.93 |
| Same-sex friend support features | 2.83 (0.66) | 1.00–4.71 | 0.93 |
| Same-sex friend negative features | 1.14 (0.29) | 1.00–3.17 | 0.81 |
| Child support features | 3.30 (0.58) | 1.00–4.76 | 0.91 |
| Child negative features | 1.33 (0.48) | 1.00–5.00 | 0.89 |
| Grand-child support features | 2.91 (0.61) | 1.00–4.71 | 0.91 |
| Grand-child negative features | 1.12 (0.28) | 1.00–3.67 | 0.81 |
| Opposite-sex friend support features | 2.60 (0.59) | 1.10–4.19 | 0.91 |
| Opposite-sex friend negative features | 1.14 (0.30) | 1.00–2.83 | 0.78 |
- Note: Scoring procedures for the instruments are given in the text.
- Abbreviations: SD, standard deviation; α, Cronbach's alpha internal consistency coefficient.
2.2.2 Network of relationships inventory-social provisions version for older adults
The Network of Relationships Inventory-Social Provisions Version (NRI-SPV; Furman & Buhrmester, 2009) assesses relationship satisfaction through a 30-item self-report measure of two negative interaction features and seven support features. In the original version of the measure, up to six relationships were examined with respect to each of the 30 items: mother, father, romantic partner, same-sex, and other (opposite)-sex friend, and other. In collaboration with Wyndol Furman, we revised these relationship figures to be appropriate for older adults, and these figures included romantic partner, twin sibling, child, grandchild, same-sex friend, and other-sex friend. For each relationship, the two negative interaction scales are conflict and antagonism; the support feature scales are affection, reliable alliance, reassurance of worth, intimacy, instrumental aid, companionship, and nurturance. The two higher-level domains (support and negative features) were computed by averaging the items on each scale that contributed to the domain (see Table 1). Only the two higher-level domains are analyzed for this study.
As noted above, the same set of questions are asked for each person identified in the respondent's social network. This creates a matrix of scores that allows for comparisons between each type of relationship. Correlations between NRI variables are shown in Table 2. Some participants did not endorse having certain relationships and, therefore, did not complete ratings for those items. Sample sizes used for analyses varied across relationship types depending on the availability of complete sets of twin data where both participants identified having that relationship. Example items for the higher-level domains include “How much does this person show support for your activities?” (support feature) and “How often do you and this person argue with each other?” (negative feature). The answer choices are Little or None (1), Somewhat (2), Very Much (3), Extremely Much (4), and The Most (5). The NRI-SPV higher-level domains in this sample had high internal consistency (0.78 < α < 0.95).
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. Romantic partner support | – | −0.40** | 0.23** | 0.02 | 0.18** | 0.00 | 0.40** | −0.11* | 0.33** | −0.06 | 0.23** | −0.08 |
| 2. Romantic partner negative | −0.40** | – | −0.05 | 0.19** | −0.11* | 0.07 | −0.11 | 0.28** | −0.12* | 0.17** | −0.05 | 0.22** |
| 3. Twin support | 0.23** | −0.05 | – | −0.31** | 0.28** | −0.05 | 0.55* | −0.07 | 0.36** | 0.01 | 0.14* | 0.02 |
| 4. Twin negative | 0.02 | 0.19** | −0.31** | – | 0.08* | 0.24** | −0.08* | 0.33** | 0.02 | 0.08* | 0.04 | 0.14* |
| 5. Same-sex friend support | 0.18** | −0.11* | 0.28** | 0.08* | – | 0.05 | 0.42** | −0.01 | 0.39** | 0.06 | 0.68** | 0.08 |
| 6. Same-sex friend negative | 0.00 | 0.07 | −0.05 | 0.24** | 0.05 | – | −0.01 | 0.17** | −0.02 | 0.24** | 0.06 | 0.39** |
| 7. Child support | 0.40** | −0.11* | 0.55** | −0.08* | 0.42** | −0.01 | – | −0.11* | 0.71** | 0.04 | 0.42** | 0.04 |
| 8. Child negative | −0.11* | 0.28** | −0.07* | 0.33** | −0.01 | 0.17** | −0.11* | – | −0.09* | 0.31** | 0.00 | 0.17* |
| 9. Grand-child support | 0.33** | −0.12* | 0.36** | 0.02 | 0.39** | −0.02 | 0.71** | −0.09* | – | 0.00 | 0.35** | 0.10 |
| 10. Grand-child negative | −0.06 | 0.17** | 0.01 | 0.08* | 0.06 | 0.24** | 0.04 | 0.31** | 0.00 | – | 0.17* | 0.26** |
| 11. Opposite-sex friend support | 0.23** | −0.05 | 0.14* | 0.04 | 0.68** | 0.06 | 0.42** | 0.00 | 0.35** | 0.17* | – | 0.10 |
| 12. Opposite-sex friend negative | −0.08 | 0.22** | 0.02 | 0.14* | 0.08 | 0.39** | 0.04 | 0.17* | 0.10 | 0.26** | 0.10 | – |
- * p < 0.05;
- ** p < 0.001 for each correlation.
2.3 Data analyses
2.3.1 Preliminary analyses
Univariate ACE models
All data analyses were completed in R version 2022.12.0+353 (R Core Team, 2022). Preliminary analyses were conducted to provide univariate estimates for genetic and environmental sources of variance in each of the PID-5 and NRI domains. The preliminary results offer insight into the heritability of these traits and whether shared genetic and environmental confounds exist (as assumed in co-twin models). For each phenotype, we calculated MZ and DZ twin intraclass correlations (ICCs) to examine similarity within twin pairs. Higher ICC values among MZ twin pairs as compared to DZ twin pairs suggest genetic influences on a given phenotype. Next, we modeled univariate ACE (or ADE), AE, and CE variance decompositions using the OpenMx package in R (Neale et al., 2016). As recommended, age, sex, age2, and age × sex were regressed out for all variables to avoid overestimating twin intraclass correlations (McGue & Bouchard, 1984). The standardized residual values from those regressions were then used in the ACE models. When DZ ICCs were less than half the MZ ICC, an ADE model was fit rather than an ACE model. We used the Akaike information criteria (AIC) to compare models for each phenotype, and the model with the lowest AIC was reported as the best fit. In the case of inadmissible parameters (negative variance components) for the best-fit model (ADE or ACE), the model with the next lowest AIC (AE) was reported instead (these rare instances are noted within the results table).
2.3.2 Primary analyses
Individual-level analyses
To estimate the bivariate association between relationship satisfaction and personality pathology while accounting for the within-family data structure, individual-level linear mixed models were created for MZ and DZ twins separately in R using the lme4 package (Bates et al., 2015). We interpreted the main effect (beta coefficient) of personality pathology predicting relationship satisfaction. Random intercepts corresponding to family ID were included to account for similarities within twin pairs. Only results with p ≤ 0.01 are considered statistically significant to account for multiple testing for these and all subsequent analyses. DZ individual analyses are considered supplemental and are discussed in the supplemental results section.
Co-twin control analyses
Co-twin control analyses were also performed separately for MZ and DZ twins using the lme4 package (Bates et al., 2015) in R with random intercepts corresponding to family ID included. This analysis controls for shared genetic and environmental experiences within a twin pair while isolating any twin difference effects. In this linear mixed model, the twin pair's mean score for each variable and each individual twin's difference from the pair's mean are calculated to estimate within-twin pair effects (associations within pairs) and between-twin pair difference effects. Within-pair effects in this model reflect unique nonshared environmental effects free of shared experiences and genetic variation that typically confound correlational research. We interpreted the beta coefficients to test the within-twin pair effects of personality pathology on relationship satisfaction. All data were standardized before running regressions, and, thus, regression coefficients are standardized in the reported results. DZ co-twin analyses were also conducted to further examine the role of genetics (rG) and shared environment (rC). DZ co-twin analyses are considered supplemental and are discussed in the supplemental results section.
2.3.3 Supplemental analyses
Co-twin control analysis with sex interaction effects
To examine differences across sex in the effect of personality pathology on relationship satisfaction (while controlling for shared genetic and environmental contributions), co-twin control analyses were performed for MZ twins and included sex interaction terms. Sex interaction terms were created by multiplying a sex dummy variable (0 = male, 1 = female) by the within-pair difference term for a given PID-5 variable. These analyses were run across each PID-5 variable with each NRI outcome. All data were standardized before running regressions and we interpreted the beta coefficients of sex interaction effects.
Co-twin control analysis with zygosity interaction effects
The co-twin control analyses were conducted for MZ and DZ twins together to include zygosity by (within-pair difference in) personality interaction term. This allows us to make further inferences on the relative contributions of genetic and environmental influences on the covariance. The zygosity term was coded 1 = MZ and 0 = DZ so a significant, negative interaction would suggest genetic confounding in the association (i.e., within-pair effects are more pronounced in DZ twins). All data were standardized before running regressions and we interpreted the beta coefficients of zygosity interaction effects.
3 RESULTS
3.1 Preliminary analyses
3.1.1 Univariate ACE models
For most PID-5 and NRI domains (15 of 17), the MZ ICC (−0.06 ≤ r ≤ 0.72) was greater than the DZ ICC (−0.14 ≤ r ≤ 0.59), suggesting genetic influences on those phenotypes (see Table 3). However, for the same-sex friend and child relationship negative features, the DZ correlation was greater. Additionally, the opposite-sex friend relationship negative features for both MZ and DZ twins had negative ICCs, perhaps due to a smaller sample available as compared to other relationships.
| Variable | MZ pairs | DZ pairs | MZ ICC | DZ ICC | A [95% CI] | C [95% CI] | E [95% CI] | AIC |
|---|---|---|---|---|---|---|---|---|
| Negative affect | 415 | 254 | 0.33 | 0.19 | 0.33 [0.25–0.41] | – | 0.67 [0.59–0.75] | 3747.65 |
| Detachment | 413 | 249 | 0.37 | 0.16 | 0.36 [0.28–0.44] | – | 0.64 [0.56–0.72] | 3711.62 |
| Antagonisma | 412 | 253 | 0.30 | 0.02 | 0.27 [0.18–0.36] | – | 0.73 [0.64–0.82] | 3673.87 |
| Disinhibition | 415 | 254 | 0.37 | 0.24 | 0.39 [0.30–0.46] | – | 0.61 [0.54–0.70] | 3723.98 |
| Psychoticism | 410 | 254 | 0.28 | 0.10 | 0.28 [0.19–0.36] | – | 0.72 [0.64–0.81] | 3726.73 |
| Romantic partner support | 268 | 143 | 0.21 | 0.03 | 0.20 [0.09–0.31] | – | 0.80 [0.69–0.91] | 2913.78 |
| Romantic partner negative | 269 | 143 | 0.23 | 0.07 | 0.23 [0.11–0.34] | – | 0.77 [0.66–0.89] | 2912.33 |
| Twin support | 410 | 249 | 0.72 | 0.57 | 0.25 [0.10–0.42] | 0.48 [0.32–0.61] | 0.27 [0.23–0.32] | 3230.71 |
| Twin negative | 415 | 252 | 0.66 | 0.59 | – | 0.64 [0.59–0.68] | 0.36 [0.32–0.41] | 3444.02 |
| Same-sex friend supporta | 308 | 189 | 0.30 | 0.05 | 0.28 [0.18–0.38] | – | 0.72 [0.62–0.82] | 3161.98 |
| Same-sex friend negative | 313 | 190 | 0.06 | 0.14 | – | 0.09 [0.00–0.17] | 0.91 [0.83–1.00] | 3245.12 |
| Child support | 314 | 170 | 0.40 | 0.15 | 0.40 [0.31–0.48] | – | 0.60 [0.52–0.69] | 3090.39 |
| Child negative | 319 | 174 | 0.08 | 0.20 | – | 0.11 [0.03–0.20] | 0.89 [0.80–0.97] | 3197.38 |
| Grand-child support | 254 | 139 | 0.30 | 0.03 | 0.28 [0.17–0.39] | – | 0.72 [0.61–0.83] | 2721.21 |
| Grand-child negative | 259 | 140 | 0.10 | 0.01 | 0.08 [−0.03–0.19] | – | 0.92 [0.81–1.03] | 2797.81 |
| Opposite-sex friend support | 81 | 54 | 0.41 | 0.19 | 0.45 [0.25–0.60] | – | 0.55 [0.40–0.75] | 1538.95 |
| Opposite-sex friend negativeb | 82 | 54 | −0.06 | −0.14 | – | −0.09 [−0.27–0.10] | 1.09 [0.90–1.27] | 1552.45 |
- Note: The PID-5 variables are the higher-order domains assessed in the inventory. The NRI variables are the higher-level domains of support and negative features composited from subscales of items assessed in the inventory. The number of MZ and DZ pairs is dependent on complete twin pairs that have data for a given phenotype. Differences in total twin pairs number between phenotypes are due to missing or incomplete data.
- Abbreviations: A, additive genetic; AIC, Akaike information criteria; C, common environment; CI, confidence intervals; DZ ICC, dizygotic intraclass correlation; DZ pairs, number of dizygotic twin pairs; E, unique environment; MZ ICC, monozygotic intraclass correlation; MZ pairs, number of monozygotic twin pairs; NRI, network of relationships inventory; PID-5, personality inventory for DSM-5.
- a AE models are interpreted despite higher AIC because the ADE model provided nonsensical parameters (negative variance components), perhaps due to lower than expected DZ ICCs. Models are estimated using direct symmetric approach, which does not place a lower bound of 0 on variance component estimates to avoid upward bias in estimates, at the potential risk of decreasing interpretability (Verhulst et al., 2019).
- b The best-fit model was CE despite having nonsensical parameters; other models yielded similar results such that there may not be an appropriate model for this phenotype.
The majority of univariate decompositions had a lower AIC for the AE model than the ACE or ADE models. Each of these PID-5 and NRI domains had genetic and unique environmental variance estimates that were greater than zero. For the twin relationship support features, an ACE model was indicated to be the best fitting and includes variance estimates for the common environment. For twin, same-sex friend, child, and opposite-sex friend relationship negative features, a CE model was indicated to be best fitting. These models suggest variance is not accounted for by additive genetics, but results may be impacted by the skew of variables (most NRI-negative features are left skewed) or lower sample sizes. Univariate parameter estimates and fit statistics can be found in Table 3.
3.2 Primary analyses
3.2.1 Individual-level analyses
Results of the MZ individual-level analyses indicated that negative affect, detachment, and disinhibition were consistently associated with relationship dissatisfaction, as measured by greater negative features and fewer support features for most relationships assessed in the NRI. Notably, psychoticism and antagonism had larger nominal associations with greater negative features as opposed to fewer support features (see Table 4).
| PID-5 variable | NRI variable | Romantic partner | Twin | Same-sex friend | Child | Grand-child | Opposite-sex friend | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| β (SE) | p | N | β (SE) | p | N | β (SE) | p | N | β (SE) | p | N | β (SE) | p | N | β (SE) | p | N | ||
| Negative affect | Support | −0.21 (0.04) | <0.001 | 536 | −0.07 (0.03) | 0.01 | 820 | −0.12 (0.04) | <0.001 | 616 | −0.10 (0.04) | 0.01 | 628 | −0.07 (0.04) | 0.12 | 508 | −0.17 (0.07) | 0.01 | 162 |
| Negative | 0.28 (0.04) | <0.001 | 538 | 0.10 (0.03) | <0.001 | 830 | 0.10 (0.04) | 0.01 | 626 | 0.15 (0.04) | <0.001 | 638 | 0.20 (0.04) | <0.001 | 518 | 0.12 (0.07) | 0.07 | 164 | |
| Detachment | Support | −0.34 (0.05) | <0.001 | 534 | −0.11 (0.03) | <0.001 | 818 | −0.13 (0.04) | <0.001 | 616 | −0.19 (0.04) | <0.001 | 626 | −0.22 (0.05) | <0.001 | 506 | −0.27 (0.07) | <0.001 | 162 |
| Negative | 0.32 (0.05) | <0.001 | 536 | 0.10 (0.03) | <0.01 | 826 | 0.15 (0.04) | <0.001 | 624 | 0.18 (0.04) | <0.001 | 634 | 0.15 (0.05) | <0.01 | 514 | 0.12 (0.07) | 0.08 | 164 | |
| Antagonism | Support | −0.13 (0.04) | <0.01 | 534 | −0.05 (0.03) | 0.08 | 816 | −0.01 (0.04) | 0.90 | 614 | −0.08 (0.04) | 0.05 | 624 | 0.01 (0.04) | 0.81 | 504 | −0.06 (0.08) | 0.47 | 160 |
| Negative | 0.14 (0.04) | <0.01 | 536 | 0.06 (0.03) | 0.03 | 824 | 0.20 (0.04) | <0.001 | 622 | 0.11 (0.04) | <0.01 | 632 | 0.17 (0.05) | <0.001 | 512 | 0.14 (0.08) | 0.07 | 162 | |
| Disinhibition | Support | −0.14 (0.04) | <0.01 | 536 | −0.09 (0.03) | <0.01 | 820 | −0.11 (0.04) | 0.01 | 616 | −0.14 (0.04) | <0.01 | 628 | −0.07 (0.04) | 0.10 | 508 | −0.16 (0.07) | 0.02 | 162 |
| Negative | 0.22 (0.04) | <0.001 | 538 | 0.11 (0.03) | <0.001 | 830 | 0.13 (0.04) | <0.001 | 626 | 0.15 (0.04) | <0.001 | 638 | 0.22 (0.04) | <0.001 | 518 | 0.09 (0.07) | 0.18 | 164 | |
| Psychoticism | Support | −0.17 (0.04) | <0.001 | 530 | −0.08 (0.03) | <0.01 | 814 | −0.04 (0.04) | 0.29 | 612 | −0.16 (0.04) | <0.001 | 626 | −0.07 (0.05) | 0.11 | 506 | −0.02 (0.07) | 0.77 | 162 |
| Negative | 0.22 (0.04) | <0.001 | 532 | 0.12 (0.03) | <0.001 | 820 | 0.11 (0.04) | 0.01 | 618 | 0.15 (0.04) | <0.001 | 632 | 0.31 (0.05) | <0.001 | 512 | 0.16 (0.06) | 0.01 | 164 | |
- Note: β = Main effects of the PID-5 higher-order domains on the higher-level domains of the NRI (support and negative features). Observation numbers are dependent on twin pairs that have data available for both variables included in the regression. Differences in observation number between regressions are due to missing data in which case both twins in a pair were removed. Only results with p ≤ 0.01 are considered statistically significant to account for multiple testing.
- Abbreviations: NRI, network of relationships inventory; PID-5, personality inventory for DSM-5.
Romantic partner
For MZ twins, all PID-5 domains were significantly associated with greater negative features (0.14 ≤ β ≤ 0.32) and fewer support features (−0.34 ≤ β ≤ −0.13) for the romantic partner. The domains with the largest associations with both features assessed in the NRI were negative affect (0.21 ≤ |β| ≤ 0.28) and detachment (0.32 ≤ |β| ≤ 0.34). Disinhibition, psychoticism, and antagonism were also associated with greater negative features and fewer support features, although to a lesser degree (|β| ≤ 0.22).
Twin
For the twin relationship, almost all PID-5 domains were significantly associated with greater negative features (0.10 ≤ β ≤ 0.12) and fewer support features (−0.07 ≤ β ≤ −0.11). However, antagonism did not have a significant association with twin support features. Negative affect (0.07 ≤ |β| ≤ 0.10), detachment (0.10 ≤ |β| ≤ 0.11), disinhibition (0.09 ≤ |β| ≤ 0.11), and psychoticism (0.08 ≤ |β| ≤ 0.12) were all predictive of relationship dissatisfaction.
Same-sex friend
In the same-sex friend relationship, detachment was the strongest predictor of greater negative features and fewer support features (0.13 ≤ |β| ≤ 0.15). Antagonism also did not have a significant association with relationship support features. Negative affect and disinhibition were implicated in same-sex friend relationship dissatisfaction, although to a lesser degree (0.10 ≤ |β| ≤ 0.13).
Child
For the child relationship, negative affect, detachment, disinhibition, and psychoticism were all significantly associated with relationship dissatisfaction (0.10 ≤ |β| ≤ 0.19). Antagonism was also associated with greater negative features and fewer support features, although to a lesser degree than other personality domains (0.08 ≤ |β| ≤ 0.11).
Grandchild
Similar results are indicated for the grandchild relationship compared to the child relationship, with negative affect (0.07 ≤ |β| ≤ 0.20) and detachment (0.15 ≤ |β| ≤ 0.22) as the strongest predictors of relationship dissatisfaction. Antagonism, disinhibition, and psychoticism specifically had stronger associations with negative relationship features (0.17 ≤ |β| ≤ 0.31) than with support features (|β| ≤ 0.07). Additionally, those associations with support features were not statistically significant.
Opposite-sex friend
For opposite-sex friend, negative affect, detachment, and disinhibition were significantly associated with greater support features (0.16 ≤ |β| ≤ 0.27), and psychoticism was predictive of greater negative features (|β| = 0.16). All other regressions were nonsignificant for this relationship type in MZ twins. Of all the relationships in the NRI, the fewest twin pairs identified having an opposite-sex friend (82 MZ pairs).
3.2.2 Co-twin analyses
Co-twin analyses were conducted to analyze twin differences in level of maladaptive personality pathology. As with individual-level analyses, within-twin pair differences in negative affect and detachment were consistently associated with greater negative features and fewer support features for most relationship types (see Table 5).
| PID-5 variable | NRI variable | Romantic partner | Twin | Same-sex friend | Child | Grand-child | Opposite-sex friend | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| β (SE) | p | β (SE) | p | β (SE) | p | β (SE) | p | β (SE) | p | β (SE) | p | ||
| Negative affect | Support | −0.17 (0.06) | 0.01 | −0.07 (0.03) | 0.02 | −0.14 (0.05) | 0.01 | −0.08 (0.05) | 0.14 | −0.16 (0.06) | 0.01 | −0.10 (0.09) | 0.25 |
| Negative | 0.20 (0.06) | 0.00 | 0.04 (0.03) | 0.29 | 0.04 (0.07) | 0.57 | 0.03 (0.07) | 0.63 | 0.17 (0.07) | 0.02 | 0.15 (0.11) | 0.17 | |
| Detachment | Support | −0.27 (0.08) | <0.001 | −0.09 (0.03) | <0.01 | −0.09 (0.06) | 0.13 | −0.11 (0.06) | 0.05 | −0.21 (0.07) | <0.01 | −0.26 (0.10) | 0.01 |
| Negative | 0.31 (0.07) | <0.001 | 0.07 (0.04) | 0.04 | 0.12 (0.07) | 0.11 | 0.19 (0.07) | 0.01 | 0.06 (0.08) | 0.43 | 0.05 (0.13) | 0.68 | |
| Antagonism | Support | −0.04 (0.06) | 0.48 | 0.00 (0.03) | 0.90 | 0.02 (0.06) | 0.80 | −0.06 (0.05) | 0.23 | −0.01 (0.06) | 0.83 | −0.15 (0.10) | 0.15 |
| Negative | 0.08 (0.06) | 0.21 | 0.04 (0.04) | 0.28 | 0.16 (0.07) | 0.02 | −0.01 (0.07) | 0.82 | 0.09 (0.07) | 0.22 | 0.30 (0.13) | 0.02 | |
| Disinhibition | Support | −0.10 (0.07) | 0.12 | −0.07 (0.03) | 0.03 | −0.15 (0.06) | 0.01 | −0.14 (0.05) | 0.01 | −0.15 (0.06) | 0.01 | −0.11 (0.10) | 0.26 |
| Negative | 0.14 (0.07) | 0.04 | 0.05 (0.04) | 0.17 | 0.00 (0.07) | 0.94 | 0.01 (0.07) | 0.93 | 0.22 (0.07) | <0.01 | 0.13 (0.12) | 0.28 | |
| Psychoticism | Support | −0.20 (0.06) | <0.01 | −0.03 (0.03) | 0.23 | −0.08 (0.05) | 0.16 | −0.16 (0.05) | <0.01 | −0.12 (0.06) | 0.05 | −0.01 (0.09) | 0.95 |
| Negative | 0.17 (0.06) | 0.01 | 0.05 (0.03) | 0.15 | 0.05 (0.06) | 0.46 | 0.03 (0.06) | 0.66 | 0.20 (0.07) | 0.01 | 0.09 (0.11) | 0.43 | |
- Abbreviations: NRI, network of relationships inventory; PID-5, personality inventory for DSM-5.
- Note: The PID-5 variables are the higher-order domains assessed in the inventory. The NRI variables are the higher-level domains of support and negative features composited from subscales of items assessed in the inventory. β = Main effect of within-twin pair differences in the given PID-5 variable on support or negative features in the relevant relationship. Only results with p ≤ 0.01 are considered statistically significant to account for multiple testing.
Romantic partner
For the MZ romantic partner relationship, within-twin pair differences in negative affect (0.17 ≤ |β| ≤ 0.20), detachment (0.27 ≤ |β| ≤ 0.31), and psychoticism (0.17 ≤ |β| ≤ 0.20) were significantly associated with relationship dissatisfaction. Antagonism did not have significant associations with either support or negative relationship features, while disinhibition only had a significant association with greater negative features (β = 0.14).
Twin
Within-twin pair differences in negative affect were associated with fewer support features (β = −0.07) and with-twin pair differences in detachment were associated with greater negative features and fewer support features for the twin relationship (0.07 ≤ |β| ≤ 0.09), although the effects are weaker in comparison to other relationship types. Within-pair differences in antagonism and psychoticism did not have significant associations with MZ twin relationship satisfaction.
Same-sex friend
Within-twin pair differences in negative affect (β = −0.14) and disinhibition (β = −0.15) for the same-sex friend were associated with fewer support features. Within-twin pair difference effects in antagonism were associated with greater negative features (β = 0.16), while all other regressions were not statistically significant.
Child
For the child relationship, within-twin pair differences in detachment were predictors of both greater negative features and fewer support features (0.11 ≤ |β| ≤ 0.19). Twin differences in disinhibition (β = −0.14) and psychoticism (β = −0.16) were significantly associated with fewer support features for child relationships. All other regressions were not statistically significant.
Grandchild
Twin difference effects in negative affect (0.16 ≤ |β| ≤ 0.17) and psychoticism (0.12 ≤ |β| ≤ 0.20) were associated with both NRI features for the grandchild relationship. Although within-twin pair differences in disinhibition were associated with fewer support features (β = −0.15), twin difference effects in antagonism and detachment did not have significant associations with relationship satisfaction.
Opposite-sex friend
In the opposite-sex friend relationship, within-twin pair differences in detachment were predictive of fewer support features (β = −0.26). Twin difference effects in antagonism were significantly associated with greater negative features (β = 0.30). All other associations between twin differences in personality pathology and relationship satisfaction were not statistically significant for MZ twins.
3.3 Supplemental analyses
3.3.1 DZ individual-level analyses
DZ individual-level analyses reflect a similar trend to MZ individual analyses, although with fewer results being statistically significant, presumably due to the smaller sample size. Negative affect and detachment were associated with relationship dissatisfaction for most relationships assessed in the NRI. Psychoticism also followed a similar trend to MZ twins, with larger nominal associations with greater negative features (see Table S1). Of all the relationships identified in the NRI, the fewest twin pairs identified having an opposite-sex friend (54 DZ pairs).
3.3.2 DZ co-twin analyses
For DZ twins, within-twin pair differences in detachment, disinhibition, and psychoticism were associated with greater negative features across half of the NRI relationships (Table S2). DZ twin PID-5 within-pair associations with relationship dissatisfaction tended to be nominally larger, but yielded fewer significant results as compared to MZ twins given the smaller sample size.
3.3.3 Co-twin analyses with sex interaction effects
Sex was not a significant moderator of the vast majority of personality-relationship associations. Sex interaction terms were not significant for all associations within the romantic partner, same-sex friend, or grandchild relationships (Table S3). For the twin relationship, the sex by (within-twin pair differences in) psychoticism term was a statistically significant predictor of greater negative features, such that there was a stronger association in men versus women. For the child relationship, sex interaction terms were significant for disinhibition and relationship dissatisfaction, consistent with a stronger association in men as compared to women. Lastly, sex interaction effects were significant for psychoticism and opposite-sex friend relationship dissatisfaction, suggesting a stronger association in men versus women.
3.3.4 Co-twin analyses with zygosity interaction effects
Zygosity was not a significant moderator of the vast majority of personality–relationship associations. In co-twin analyses (MZ and DZ twins combined) with zygosity by within-pair personality interaction effects (Table S4), there were essentially no significant results (only one significant result out of 60 tests). Zygosity interaction effects were significant for antagonism and relationship dissatisfaction for the romantic partner (β = −0.26) in the expected direction (i.e., within-pair differences in personality were stronger predictors of relationship dissatisfaction in DZ twins). Overall, effect sizes were nominally small to moderate (0.00 ≤ |β| ≤ 0.24).
4 DISCUSSION
In the present study, higher levels of maladaptive personality traits in older adulthood were associated with lower relationship satisfaction as measured by greater negative features and fewer support features. In our preliminary analyses, we established that genetic and environmental influences contributed to variance in both the PID-5 and NRI phenotypes. We then expanded on these results by using an MZ co-twin analysis to examine the unique environmental contributions of personality pathology predicting relationship satisfaction. MZ individual and co-twin level analyses are primarily interpreted below.
Overall, personality pathology main effects on relationship satisfaction were most salient for the romantic partner, child, and grandchild relationships. Associations between personality pathology and relationship satisfaction were nominally larger for the opposite-sex friend compared to the same-sex friend. Personality pathology had the weakest associations with twin relationship satisfaction compared to all other NRI relationship types.
Results of both the MZ individual and co-twin level analyses found that detachment and negative affect had the largest associations with relationship dissatisfaction. The results for the individual-level analyses aligned with previous research supporting consistent associations between negative affect and detachment with relationship satisfaction at younger adulthood to midlife (Decuyper et al., 2018). Other PID-5 domains had small to moderate correlations.
The pattern of results linking maladaptive personality to romantic relationship dissatisfaction (Wilson et al., 2018) generalized to other relationship types. At both the individual and co-twin levels, personality pathology had notable associations with child and grandchild relationship satisfaction. Weaker associations can be seen for twin, same-sex friend, and opposite-sex friend relationships. These weaker associations could be attributed to lower levels of contact compared to the romantic partner. While most participants likely have daily interactions with their romantic partner, relationships with the friend or twin could be impacted by physical distance or limited to online or phone contact. It is possible that the instrumental aid and companionship facets of the NRI largely contributed to the lower mean scores in MZ support features for the twin (M = 3.31), same-sex friend (M = 2.83), or opposite-sex friend (M = 2.60) since those items require a more physical presence. Additionally, personality pathology was more predictive of relationship dissatisfaction with opposite-sex friends compared to same-sex friends. However, there are fewer MZ and DZ twin pairs that identified an opposite-sex friend, so those participants may feel more strongly about those relationships. The decrease in power as a result of a smaller sample size for opposite-sex friend regressions could have impacted main effects. Overall, these findings support the notion that the maladaptive personality traits assessed in the Alternative Model of Personality Disorders (AMPD) are implicated in interpersonal dysfunction across multiple important domains.
The MZ co-twin analysis suggested that the observed associations are mediated in the unique environment and are not strictly the result of genetic confounding. The magnitudes of the MZ within-pair effects of negative affect on a romantic partner, twin, and same-sex friend support features were similar to the individual-level effects. Additionally, within-pair effects of detachment on relationship dissatisfaction were similar at the individual level for romantic partner, twin, grandchild, and opposite-sex friend relationships. Other observed similarities between within-pair effects and individual-level effects include psychoticism for the romantic partner and child relationship, as well as disinhibition in child and grandchild relationships. Overall, this pattern of results is consistent with personality pathology, most notably detachment and negative affect, having a causal effect on relationship satisfaction, and are most salient in romantic relationships. However, given the cross-sectional nature of the data, reverse causation (i.e., relationship dissatisfaction leading to personality pathology) cannot be ruled out.
Supplemental analyses were conducted to examine MZ and DZ discordance for insight into the role of additive genetics and shared environment as well as interaction effects of sex and zygosity. Main effects of twin differences in personality pathology and relationship satisfaction for the DZ romantic partner relationship were nominally larger than for the MZ relationship, which may indicate a significant genetic component to these associations. An overall lack of significant sex interaction effects may suggest that the genetic and environmental influences between male and female twins were comparable or that the study was underpowered to detect those differences. Similarly, a lack of significant zygosity interaction effects may be due to limitations in statistical power.
Limitations of this study include inability to account for gene–environment interactions, sample-related issues, and using only self-reported data. In the context of this study, gene–environment correlations refer to a genetic propensity to be less satisfied in relationships, which may then invoke unshared environmental factors to influence further relationship dissatisfaction. While these gene–environment interactions (e.g., evocative rGE) are an important source of variation to consider for behavioral traits, it is outside the scope of current analyses. In terms of sample-related issues, the prevalence of missing data likely impacted the results and lower levels of racial diversity limits generalizability to more diverse samples. Certain relationships assessed in the NRI (i.e., opposite-sex friend) were identified by fewer twin pairs compared to the romantic partner. This may be attributed to less common relationships held by the older adult population. For relationship types that used a smaller sample size, given the need to remove twin pairs with incomplete data, we may be underpowered to detect significant findings. The sample used in this study was also predominantly white and recruited exclusively from Minnesota. Expanding to a sample with greater racial diversity could increase the generalizability of results. Lastly, the data used in this study were self-reported; collecting informant-reported data or using other modes of assessment could offer another perspective on these factors.
Future directions of related research could address the bidirectionality of personality pathology effects on relationship satisfaction by conducting a longitudinal twin study using older adults. Directionality cannot be determined in cross-sectional data used in the present study. This would provide more insight into the potential causality of these variables. Further research could also address the alternate hypothesis by Wilson et al. (2018) that individuals with higher levels of maladaptive traits establish relationships with others who have high levels of maladaptive traits, leading to greater overall dysfunction.
Overall, results of this study suggest that maladaptive personality traits are implicated in a wide range of interpersonal dysfunction, with closer relationships (the romantic partner) being the most impacted. The co-twin analysis also provides more insight into unique environmental factors that indicate a potentially causal relationship between PID-5 traits and relationship satisfaction. The mechanisms involving unique environmental factors through which personality impacts relationship satisfaction may be educational and work experiences, life events, or peer influences, which can shape personality traits and perceptions of their relationships. The use of an older adult sample highlights the impact of personality pathology on relationships in later life and the relationships where interpersonal impairment is particularly salient.
AUTHOR CONTRIBUTIONS
Helen H. Yu: Data Analysis, Interpretation of Results, Writing—Original Draft. Colin D. Freilich: Analysis, Interpretation, Writing—Reviewing and Editing. Sylia Wilson: Writing—Reviewing and Editing. Matt McGue: Conceptualization, Obtained Funding, Writing—Reviewing and Editing, Study Design. Glenn I. Roisman: Conceptualization, Obtained Funding, Writing—Reviewing and Editing, Study Design. Robert F. Krueger: Conceptualization, Obtained Funding, Writing—Reviewing and Editing, Study Design, Resources, Supervision.
ACKNOWLEDGEMENTS
This study was supported by a grant from the National Institute on Aging (R01AG053217) to RFK and GIR. CDF was likewise supported as a trainee by the National Institute on Drug Abuse (T32DA050560). That said, the content of this paper is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
CONFLICT OF INTEREST STATEMENT
We have no known conflict of interest to disclose.
ETHICS STATEMENT
Data collection for The contributions of personality and social relationships to late-life health: A twin study approach was reviewed and approved by IRB protocol number 1604S86262 at the University of Minnesota.
Open Research
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available at https://osf.io/a86hg/. This study was not preregistered.
