Loneliness is the subjective experience of social isolation. In a comprehensive review of the evolutionary and heritability findings of loneliness, Cacioppo et al. (Reference Cacioppo, Cacioppo and Boomsma2014) argued that loneliness may have an adaptive function because loneliness is an ‘aversive state that, like hunger, thirst, and pain, promotes behavior change to increase the likelihood of the survival of one’s genes’ (p. 3). Qualter et al. (Reference Qualter, Vanhalst, Harris, Van Roekel, Lodder, Bangee and Verhagen2015) describe this need to reconnect with others when feeling lonely as the reaffiliation motive (reaffiliation model). However, the manner in which individuals cope with loneliness appears to vary. Although loneliness is a commonly experienced emotion, for many (with estimates of 15–30%), loneliness is a chronic condition that has a negative impact on mental and physical wellbeing (Hawkley & Cacioppo, Reference Hawkley and Cacioppo2010).
Loneliness appears to be a stable construct (Bartels et al., Reference Bartels, Cacioppo, Hudziak and Boomsma2008; Boomsma et al., Reference Boomsma, Cacioppo, Muthén, Asparouhov and Clark2007; Junttila & Vauras, Reference Junttila and Vauras2009) that has been found to be present in individuals at every stage of life and has been shown to be associated with a range of psychological constructs implicated in poor mental health or maladjustment. In childhood and early adulthood, loneliness is associated, for example, with insecure attachment to parents (Akdoğan, Reference Akdoğan2017; Mund & Neyer, Reference Mund and Neyer2016; Wiseman et al., Reference Wiseman, Mayseless and Sharabany2006). Vanhalst et al. (Reference Vanhalst, Soenens, Luyckx, Van Petegem, Weeks and Asher2015) reported that adolescents who were chronically lonely viewed inclusion into groups as due to the situation, while exclusion from others was seen as caused by internal and stable characteristics. In adults, loneliness increases rates of mortality (Holt-Lunstad et al., Reference Holt-Lunstad, Smith, Baker, Harris and Stephenson2015) and correlates positively with shyness (Cheek & Busch, Reference Cheek and Busch1981; Jackson et al., Reference Jackson, Soderlind and Weiss2000), depressive characteristics (Besser et al., Reference Besser, Flett and Davis2003), social anxiety (Jackson, Reference Jackson2007) and neuroticism (Saklofske & Yackulic, Reference Saklofske and Yackulic1989; Saklofske et al., Reference Saklofske, Yackulic and Kelly1986; Stokes, Reference Stokes1985). Loneliness has also been found to correlate negatively with extraversion, self-esteem (de Jong-Gierveld, Reference de Jong-Gierveld1987), self-confidence (Cheng & Furnham, Reference Cheng and Furnham2002), acceptance of others (Levin & Stokes, Reference Levin and Stokes1986) and mattering (Flett et al., Reference Flett, Goldstein, Pechenkov, Nepon and Wekerle2016). Lonely people tend to make fun of themselves (Schermer et al., Reference Schermer, Martin, Vernon, Martin, Conde, Statham and Lynskey2017) and engage in self-protective and social avoidant behavior (Knowles et al., Reference Knowles, Lucas, Baumeister and Gardner2015; Lucas et al., Reference Lucas, Knowles, Gardner, Molden and Jefferis2010); and lonely people tend to have higher anxiety because of their perceptions about their social skills (Solano & Koester, Reference Solano and Koester1989).
Within personality disorders, borderline personality disorder (BPD) is a commonly diagnosed disorder that is positively related to anxiety, depression, hostility and impulsiveness, and negatively related to self-discipline and deliberation (American Psychiatric Association [APA], 2013; Bach & Sellbom, Reference Bach and Sellbom2016; Lieb et al., Reference Lieb, Zanarini, Schmahl, Linehan and Bohus2004; Trull, Reference Trull2001). As explained by Bach and Sellbom (Reference Bach and Sellbom2016), BPD as defined by the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; APA, 2013) includes ‘fear of abandonment, unstable relations, identity disturbance, impulsivity, self-destructiveness, affective lability, chronic emptiness, uncontrolled anger, and paranoia or dissociation’ (p. 492, Table 2). This sensation of chronic emptiness and fear of abandonment may specifically contribute to loneliness for those with BPD (Lieb et al., Reference Lieb, Zanarini, Schmahl, Linehan and Bohus2004). Although loneliness and an intolerance of aloneness (Matthies et al., Reference Matthies, Schiele, Koentges, Pini, Schmahl and Domschke2018) are prevalent in those with BPD, in a maladaptive manner, it may actually lead to the avoidance of others (vs. affiliation with others). For example, Hauschild et al. (Reference Hauschild, Winter, Thome, Liebke, Schmahl, Bohus and Lis2018) found that the loneliness drive to affiliate with others is actually reversed in people with BPD. For example, mimicking behavior tends to increasing social interactions and affiliation but, as found by Hauschild et al., those who reported feeling lonely were less likely to mimic others, suggesting that the lonely people may have greater difficulty interacting with other people. Hauschild et al. also indicated that people with BPD tend to expect future relationships to be negative and therefore may avoid forming bonds with others. This process has been noted previously and labeled as rejection sensitivity (e.g., see Gao et al., Reference Gao, Assink, Cipriani and Lin2017) and may capture the general interpersonal ambivalence or anxiety often seen in those with BPD features (Lieb et al., Reference Lieb, Zanarini, Schmahl, Linehan and Bohus2004; Mącik, Reference Mącik2018).
While the use of categorical clinical diagnoses presents some advantages in the research and clinical contexts, the use of dimensional models has been recommended to gain a full understanding of personality disorders, their correlates and etiology (APA, 2013). Borderline personality features are present at different degrees in the general population, and as such, can be conceptualized and measured within a continuous or dimensional framework. These features cover the borderline spectrum and include self-harm behaviors, including suicide attempts, unstable affect or volatile emotions, identity problems or feelings of emptiness or lack of purpose, and negative relationships with others (e.g., intense sense of betrayal from others; Morey, Reference Morey1991). Borderline personality features are positively correlated with neuroticism, antagonism, irresponsibility and introversion (Distel, Trull et al., Reference Distel, Trull, Willemsen, Vink, Derom, Lynskey, Martin and Boomsma2009). Studies investigating the relationship between borderline personality features and loneliness find a positive correlation between the two dimensions, that is, higher levels of borderline personality features are associated with feeling lonelier. For example, Liebke et al. (Reference Liebke, Bungert, Thome, Hauschild, Gescher, Schmahl and Lis2017) investigated loneliness in women with BPD compared to control women and found that women with BPD were lonelier; and specifically, loneliness for those women with BPD was associated with less diverse social networks and greater difficulty with interpersonal communication.
Loneliness is heritable (see review by Cacioppo et al., Reference Cacioppo, Cacioppo and Boomsma2014), with genetic factors explaining ~35–48% of the variance in loneliness in adults (Boomsma et al., Reference Boomsma, Willemsen, Dolan, Hawkley and Cacioppo2005; Distel et al., Reference Distel, Rebollo-Mesa, Abdellaoui, Derom, Willemsen, Cacioppo and Boomsma2010; Schermer & Martin, Reference Schermer and Martin2019). The single-nucleotide polymorphism (SNP) heritability (the phenotypic variance accounted for by genotyped SNPs) of a categorical loneliness measure in the UK Biobank was 8.1% and of continuous loneliness measures, 4.2% (Abdellaoui et al., Reference Abdellaoui, Sanchez-Roige, Sealock, Treur, Dennis, Fontanillas and Boomsma2019). As outlined by Trull (Reference Trull2001), mood disturbance patterns and other borderline personality characteristics have familial components. Genetic factors explain between 40% and 45% of borderline personality features, according to twin studies from the Netherlands, Belgium and Australia (Distel et al., Reference Distel, Trull, Derom, Thiery, Grimmer, Martin and Boomsma2008; Distel, Rebollo-Mesa et al., Reference Distel, Rebollo-Mesa, Willemsen, Derom, Trull, Martin and Boomsma2009) and has been estimated to be 46% in a Norwegian twin study (Torgersen et al. Reference Torgersen, Myers, Reichborn-Kjennerud, Røysamb, Kubarych and Kendler2012). Typically, it has been found that the remaining variance is due to unique environmental effects, while the common environment shared by family members or cultural transmission from parents to offspring does not significantly account for variation in borderline personality features. These results, in conjunction with the reaffiliation model (Qualter et al., Reference Qualter, Vanhalst, Harris, Van Roekel, Lodder, Bangee and Verhagen2015), suggest that uncovering the genetic and environmental relationships between loneliness and BPD would add to our understanding of how the two constructs are related beyond the phenotypic correlations typically examined.
In summary, borderline personality and loneliness have been found to be connected at a phenotypic level (Liebke et al., Reference Liebke, Bungert, Thome, Hauschild, Gescher, Schmahl and Lis2017), but the etiology of the relationship between the two constructs is unknown. Both borderline personality features (Distel et al., Reference Distel, Trull, Derom, Thiery, Grimmer, Martin and Boomsma2008; Distel, Rebollo-Mesa et al., Reference Distel, Rebollo-Mesa, Willemsen, Derom, Trull, Martin and Boomsma2009) and loneliness (Cacioppo et al., Reference Cacioppo, Cacioppo and Boomsma2014) are partially heritable, as reviewed above. The present study examines the genetic and environmental (both unique and shared) contributions to the correlation between borderline personality features and loneliness.
Method
Participants
Twins took part in studies on health and wellbeing at the QIMR Berghofer Medical Research Institute in Australia (QIMRB; see Lynskey et al., Reference Lynskey, Agrawal, Henders, Nelson, Madden and Martin2012) and the Netherlands Twin Register (NTR; Boomsma et al., Reference Boomsma, Vink, Van Beijsterveldt, de Geus, Beem, Mulder and van Baal2002; Ligthart et al., Reference Ligthart, van Beijsterveldt, Kevenaar, de Zeeuw, van Bergen, Bruins and Boomsma2019). The Australian sample included 4820 twins from 3047 families distributed in 1773 complete pairs (265 monozygotic male pairs (MZM), 546 monozygotic female pairs (MZF), 176 dizygotic male pairs (DZM), 412 dizygotic female pairs (DZF), and356 opposite sex pairs (OS) and 1274 single twins with a mean age of 28.93 years (SD = 5.8, range = 15–40). The Dutch sample included 6509 twins from 4142 families distributed in 2367 complete pairs (361 MZM, 159 DZM, 1044 MZF, 431 DZF, and372 OS) and 1775 single twins, with a mean age of 33.71 years (SD = 12.44, range = 14 to 86). All twins, from complete and incomplete pairs, were included in the study. The data for this study are available by contacting the directors of the two twin registries who are listed as the last two co-authors on the manuscript.
Measures and Procedure
As part of studies on health and lifestyle, individuals from both countries completed the Borderline Features scale from the Personality Assessment Inventory (PAI-BOR; Morey, Reference Morey1991; see also Distel et al., Reference Distel, Trull, Derom, Thiery, Grimmer, Martin and Boomsma2008) and the three questions assessing loneliness, as suggested by Hughes et al. (Reference Hughes, Waite, Hawkley and Cacioppo2004), from the revised UCLA Loneliness Scale (Russell et al., Reference Russell, Peplau and Cutrona1980). The questionnaires were completed online or via mail. The protocols were approved by the QIMR Ethics Research committee and by the Central Ethics Committee on Research Involving Human Subjects of the VU University Medical Centre, Amsterdam, an Institutional Review Board certified by the US Office of Human Research Protections (IRB number IRB00002991 under Federal-wide Assurance-FWA00017598; IRB/institute codes, NTR 03-181). In the Australian sample, the information reported at the youngest age was selected when several reports were available. The data collection in the Netherlands was part of NTR Survey 2004 (seventh survey within a longitudinal survey study in adolescent and adult participants). All participants provided informed consent and the data were screened for missing patterns and outliers (defined as observations outside a distance 1.5 times the interquartile range) by cohort and those cases were excluded in the analyses.
The PAI-BOR consists of 24 items that require responses on a four-point scale from 0 = False, not true at all to 3 = Very true. There are four six-item subscales comprising the PAI-BOR including: affect instability (high negative mood baseline, deficits in emotional regulations, high scores indicate highly emotional individuals who are quick to anger; Australian alpha = .76 and Dutch alpha = .71); identity problems (unstable sense of self that can manifest as chronic feelings of emptiness or boredom; high scores indicate the tendency of individuals to feel that they do not have a purpose in life; Australian alpha = .66 and Dutch alpha = .56); self-harm (high scores indicate individuals who tend to engage in risky behavior and possibly self-mutilation and suicide attempts; Australian alpha = .68 and Dutch alpha = .65); and negative relationships (fail to have stable and fulfilling relationships with other people; Australian alpha = .59 and Dutch alpha = .50). To avoid inflating the correlations with loneliness, the item ‘I rarely feel very lonely’ was removed from both the negative relationships scale total as well as the total borderline score. The internal consistency (coefficient alpha) value for the PAI-BOR scale without the loneliness item was .85 in the Australian sample and .78 in the Dutch sample (vs. .85 and .79, respectively, including the item) and alpha without the loneliness item did not change for the negative relationships scale in either sample.
The three items from the revised UCLA Loneliness Scale (Russell et al., Reference Russell, Peplau and Cutrona1980) by Hughes et al. (Reference Hughes, Waite, Hawkley and Cacioppo2004) ask how often an individual feels that they lack companionship, how often the individual feels that they were left out and how often the individual feels that they are isolated from others. Responses to these items were on a 1 = Hardly ever to 3 = Often scale. The internal consistency (coefficient alpha) for the loneliness scale was .82 for the Australian sample and .74 for the Dutch sample.
Statistical Analyses
The classic twin design was employed to decompose the variance of borderline features and loneliness into additive genetic (A), nonadditive or dominant genetic (D), shared environmental (C) and unique environmental (E) variance components (note the E component includes measurement error). This is achieved with a model that compares monozygotic (MZ) and dizygotic (DZ) twins within twin pair correlations. MZ twins share all their genetic variants, while DZ twins share, on average, half of their additive genetic variants. Based on biometrical genetic theory, this means that all additive and dominant genetic variances are shared by MZ twins. DZ twins share half of the additive genetic and a quarter of the dominant genetic variance. Both MZ and DZ twins experience environments that are shared, and therefore in common by both twins in a pair, which may contribute to their resemblance. In addition, each twin within a pair has experiences that are unique and also contribute to trait variation. From these relationships, a structural equation model is specified. However, with only MZ and DZ correlations, all four sources of variation cannot be identified and either C or D is excluded from the model after inspection of the twin correlations. If the DZ twin correlation is less than half the MZ twin correlation, then D is estimated along with A and E, otherwise C is estimated. From a full model with three parameters, such as an ACE model, nested models are used to test the significance of variance components; for example, evaluating the significance of C by running an AE submodel. With the data in this study, parameters were estimated by raw data maximum likelihood, and a likelihood ratio test was applied to test if the nested (or reduced) model fit the data less well than the full model (Posthuma et al., Reference Posthuma, Beem, deGeus, vanBaal, von Hjelmborg, Iachine and Boomsma2003). Age and sex were included as fixed effects in the variance component analyses. All analyses were conducted in R and OpenMx (Boker et al., Reference Boker, Neale, Maes, Wilde, Spiegel, Brick and Fox2011).
First, we assessed within each country whether means, variances and covariation could be equated across birth order, sex and zygosity. Next, MZ and DZ twin correlations were estimated, and ACE or ADE univariate models for loneliness and each of the borderline measures were fit to the data within each cohort. We then tested, in a multiple group model, whether variance component parameters could be equated across the Australian and Netherlands samples and the data combined for subsequent bivariate analyses.
Bivariate models were fitted to the data to quantify the extent that genes and the environment contribute to the observed phenotypic correlations between loneliness and borderline features. Five bivariate Cholesky decomposition analyses were conducted between loneliness and each of the borderline subscales and the borderline total score. A Cholesky decomposition partitions the variation of the initial variable (the borderline score) into genetic and environmental sources and estimates the extent those sources also contribute to the second variable (loneliness). Any remaining variation in loneliness that is independent of the borderline is also decomposed into genetic and environmental sources. With this model, we tested whether the phenotypic covariation between borderline and loneliness was due to genes or to the environment. Based on the results, a genetic and an environmental correlation were estimated between the two traits.
Results
Table 1 reports the descriptive statistics for each scale for men and women separately for the Australian and Dutch samples. Means and variances tended to be higher in the Australian cohort, both in the total sample and stratified by sex. For this reason, the variables were standardized within cohort to a mean of 0 and a standard deviation of 1. Because the data on all scales showed a somewhat skewed distribution, a square root and a logarithmic transformation were performed for the borderline personality variables and loneliness, respectively.
Table 1. Descriptive statistics by country and sex
Note: 1The item asking if people rarely feel lonely.
Assumptions of the twin design were checked, including the homogeneity of the means and variances of first- and second-born twins and across zygosity groups, and it was found that the assumptions were met. The effect of sex and age on the mean of the variables was also evaluated within cohort. With the exception of identity problems in the Dutch cohort, age had significant negative correlations with the variable scores. With respect to sex differences, females had higher means for the scale’s identity problems, negative relationships and self-harm within the Australian sample, and had higher means for all of the variables, except identity problems, in the Dutch sample.
Twin correlations could be equated for MZ pairs regardless of sex, and the same constraint was allowed for all DZ pairs. When testing, a sex limitation model and assuming the same sources of variances of the same magnitude for males and females, these constraints were allowed, thus discarding the need to stratify the variance components by sex. The first panel of Table 2 provides the twin correlations for MZ and DZ twins for the borderline personality scales and loneliness. For each scale, the MZ correlation was higher than the DZ correlation, suggesting genetic effects.
Table 2. Monozygotic (MZ) and dizygotic (DZ) twin correlations with 95% confidence intervals (CIs) by cohort for each of the phenotypes (note the twin correlations within zygosity could be equated regardless of sex) and proportions of variance explained by additive genetic (A) and residual variation (E) with 95% CIs from the univariate best fitting models (note the Australian and Dutch parameters could be equated)
Note: 1The item asking if people rarely feel lonely.
According to the observed twin correlations and comparing the magnitude of the within twin pair correlations for MZ versus DZ twins, ACE univariate models were fit for the negative relationships scale in both the Australian and Dutch samples. An ACE univariate model was also fit for the loneliness scale scores for the Australian sample. ADE univariate models were fit for the rest of the scales. The variance components of the full models could be equated across cohorts, meaning that the proportion of each of the sources of variance was equivalent and there was no need to stratify the analyses by cohort. Nested models were then tested against the full ACE or ADE models. The best fitting model in all cases was an AE model, which explains the variance observed by additive genetic (A) and unique environmental factors (which include the measurement error; E). The right-hand panel of Table 2 summarizes these results, showing that additive genetic effects account for between 31% (for the self-harm scale) and 41% (for the total Borderline scale score) of the variance and that most of the variance was accounted by unique environmental factors and measurement error.
The results of the bivariate analyses between loneliness and the borderline personality dimensions are reported in Table 3. All aspects of borderline personality were found to have significant positive phenotypic correlations (r ph) with loneliness, with the lowest estimate for the subscale self-harm (r ph = .17) and the highest estimates for the total score of the borderline personality questionnaire and identity disturbance showing r ph = .51 and r ph = .54, respectively. Consistent with the univariate analyses, the bivariate genetic models showed that the AE model had the best fit to the data. As detailed in Table 3, the results indicate substantial genetic influences on loneliness that were shared with borderline features (A12) as well as genetic influences on loneliness that were independent of borderline features (A22). Unique environment effects also influenced loneliness as well as borderline features, yet most of these influences on loneliness were independent of any influence on borderline features (E22). The genetic correlations between loneliness and borderline features ranged from 0.30 for the self-harm scale to 0.74 for the identity disturbance scale. Common unique environmental influences between loneliness and borderline features were lower than the common genetic sources, with the highest unique environmental correlations for identity disturbance (r e = .41) and the lowest for self-harm (r e = .1). The proportion of the phenotypic correlation between the total score of the PAI-BOR and loneliness due to genetic factors was 51%. For the subscales, the proportion of the phenotypic correlation with loneliness due to genetic factors was 54%, 47%, 57% and 60% for affective instability, identity disturbance, negative relationships and self-harm, respectively. Thus, approximately half of the correlation value observed between loneliness and the borderline personality dimensions is due to genetic factors.
Table 3. Model-fitting results for bivariate models (Cholesky decomposition) for the borderline traits and loneliness and proportions of variance and covariance explained by additive genetic (A) and residual variatio (E) with 95% confidence intervals (Australian and Dutch samples combined)
Note: 11 The proportion of variance explained in the borderline trait; 12 The proportion of covariance explained in the borderline trait and loneliness; 22 The proportion of variance explained in loneliness. r ph, phenotypic correlation, r a, genetic correlation, and r e, environmental correlation.
Discussion
This study examined the relationship between borderline personality features and loneliness at the phenotypic, genetic and environmental levels and confirmed that individual differences in the PAI-BOR subscales and total score and in loneliness were explained by a combination of additive genetic and environmental factors. The heritability estimates, which ranged from 31% for the borderline subscale self-harm to 41% for the borderline total score, are either the same or somewhat lower than estimates reported previously (Boomsma et al., Reference Boomsma, Willemsen, Dolan, Hawkley and Cacioppo2005; Distel, Rebollo-Mesa et al., Reference Distel, Rebollo-Mesa, Willemsen, Derom, Trull, Martin and Boomsma2009; Schermer et al., Reference Schermer, Martin, Martin, Lynskey, Trull and Vernon2015; Reference Schermer, Martin, Vernon, Martin, Conde, Statham and Lynskey2017; please note that the samples partially overlap, but that the Australian sample in the present study is substantially larger than what has been used in previous research).
Consistent with previous findings of the relationship between BPD and loneliness (Liebke et al., Reference Liebke, Bungert, Thome, Hauschild, Gescher, Schmahl and Lis2017), the phenotypic correlation between loneliness and the total scale of BPD was .51. At the subscale level, loneliness was found to have positive and significant correlations, ranging from .17 for self-harm to .54 for identity disturbance. The factors underlying the phenotypic variance–covariance structure were partially genetic and partly due to unique environmental factors. Specifically, loneliness was found to have a genetic correlation with the total BPD scale of .64 and a unique environmental correlation of .40. For the BPD subscales, the genetic correlations with loneliness ranged from .30 with self-harm to .74 with identity disturbance.
Self-harm at the extreme level involves the risk of suicide and injury, but at more moderate levels includes impulsive spending and other risk-taking behaviors (Morey, Reference Morey1991). Suicidal ideation is a common characteristic of individuals at the extreme level of self-harm. In a large sample of Danish students, Lasgaard et al. (Reference Lasgaard, Goossens, Bramsen, Trillingsgaard and Elklit2011) reported that suicidal ideation was associated with both peer-related and family-related loneliness but that deliberate self-harm was only associated with family-related loneliness. Possibly, the weaker correlation found in the present study may be due to not assessing the type of social loneliness experienced by the participants in relation to the self-harm scale. Specifically, as stated above, participants in this study were asked the degree to which they felt that they lacked companionship, felt left out and felt isolated from others. Possibly, a stronger relationship might have been found if the items specified certain groups, such as feeling left out of family activities versus social groups. Of interest, future research may assess the possible differences in the magnitude of the phenotypic, genetic and environmental correlations between borderline personality scales and types of social loneliness. For example, Liebke et al. (Reference Liebke, Bungert, Thome, Hauschild, Gescher, Schmahl and Lis2017) found that individuals with BPD were less lonely when they had a greater social network. Asking participants to provide more detail about their social interactions and the possible causes of their loneliness would aid in the understanding of how borderline personality features are linked with types of loneliness.
The moderately strong correlation between loneliness and the BPD subscale of negative relationships at the phenotypic and genetic levels, with low to moderate unique environmental correlations, is not surprising in light of the finding that people with BPD tend to report failed past relationships and predict that future social interactions will be negative (Hauschild et al., Reference Hauschild, Winter, Thome, Liebke, Schmahl, Bohus and Lis2018). Loneliness alone has been linked with a tendency to produce negative evaluations of relationships (Jones et al., Reference Jones, Sansone and Helm1983). In addition, Mellor et al. (Reference Mellor, Stokes, Firth, Hayashi and Cummins2008) reported that lonely people had a greater difference between the need to belong and relationship satisfaction than less lonely people. These results do suggest that future research may be needed to examine if cognitively altering perceived negative relationships for those with borderline personality features will, in turn, result in reports of lower scores on loneliness measures.
Affect instability, characteristic of rapid and extreme moods, had a moderately strong correlation with loneliness at both the phenotypic and genetic levels and weaker correlations due to unique environmental factors. Recently, Anupama et al. (Reference Anupama, Bhola, Thirthalli and Mehta2018) concentrated on the affect regulation and difficulties with interpersonal relationships for patients with BPD and found that individuals with BPD had difficulty recognizing positive and neutral emotions depicted in photographs. Possibly having difficulty comprehending the emotions of others leads to more extreme reactions for people with BPD, which in turn may result in an increase in perceived isolation and loneliness. Providing training or guidance for those with borderline personality features on how to interpret facial expressions correctly may reduce feelings of loneliness by improving interpersonal interactions.
The BPD subscale of identity disturbance characterizes individuals who have problems with their sense of self-worth. Morey (Reference Morey1991) describes people who score high on the identity problems scale as ‘feeling empty, bored, or unfulfilled’ (p. 17). Strong phenotypic and genetic correlations were found between loneliness and identity disturbance with a moderately strong unique environmental correlation. This pattern of correlations may reflect the finding reported by loneliness researchers who describe lonely people as socially avoidant and self-protective (Knowles et al., Reference Knowles, Lucas, Baumeister and Gardner2015; Lucas et al., Reference Lucas, Knowles, Gardner, Molden and Jefferis2010). Similarly, Flett et al.’s (Reference Flett, Goldstein, Pechenkov, Nepon and Wekerle2016) concept of mattering also explains the strong correlations found as those who feel that they do not matter also report being lonely. Although the genetic basis of mattering appears to be unknown, it would be interesting for future research to examine the genetic and environmental links between mattering, loneliness and BPD. Because a low sense of self-worth is strongly linked with loneliness, then treatment programs aimed at improving a sense of self-worth and/or feelings of mattering to others would facilitate an individual’s expressions of loneliness. In addition, although the genetic correlations were very robust, the unique environmental correlation between loneliness and identity disturbance was also significant suggesting that changes within the individual’s environmental experiences may influence scores on both variables. Future studies may make use of longitudinal data, molecular genetic data and analysis techniques such Mendialian randomization to further explore for models of causality.
Limitations
Although this study adds to the understanding of BPD and loneliness, the study is cross-sectional and therefore cannot make claims as to whether BPD characteristics have a causal influence on loneliness or if loneliness increases the likelihood of an individual to score higher on BPD items. The results do attest that for some of the variance, there is a common basis for both loneliness and BPD features to co-occur, but as the significant unique environmental correlations demonstrate, not all of the associations are due to common genetic factors. How experiences increase feelings of loneliness or contribute to borderline personality features, such as identity disturbance requires further research.
In addition, a possible limitation of the present study is that the results for the borderline personality features scales reflect those within a normal range and cannot be generalized to a clinical population. A further limitation in the present study was the use of the three-item loneliness scale. Although the three items from the UCLA Loneliness Scale (Russell et al., Reference Russell, Peplau and Cutrona1980) used in the present study from Hughes et al. (Reference Hughes, Waite, Hawkley and Cacioppo2004) is a frequently used measure in the loneliness literature, findings such as the ones reported by Lasgaard et al. (Reference Lasgaard, Goossens, Bramsen, Trillingsgaard and Elklit2011) do suggest that different aspects of loneliness (peer-related versus family-related) may have differing effects on other psychological expressions. Future studies may want to further investigate the borderline personality correlates with multiple areas of loneliness to further understand how the social networks and loneliness relate to borderline personality dimensions.
Conclusions
In conclusion, the present study found positive correlations between BPD dimensions and loneliness. As significant heritability for both constructs was confirmed, bivariate genetic analyses were conducted and found that for all of the scales, approximately half of the covariance at the phenotypic level was due to common genetic factors, with the remainder of the phenotypic covariance due to experiences unique to each twin within the pair. Importantly, the present study provides information about the sources of the phenotypic correlations between borderline personality and loneliness demonstrating substantial and significant genetic and unique environmental correlations between the borderline personality features and loneliness, meaning that these traits partially share causal pathways. This is a promising finding for interventions aiming to decrease loneliness in people with high scores in borderline personality traits, since it implies that there is a substantial proportion of the correlation that is due to factors that can potentially be modified. Given the impact of loneliness in health outcomes and the relationship with mental health, the findings of the present study add important information for understanding the factors influencing these relationships.
Acknowledgments
QIMRB warmly thank the twins and their family members for their continued support, generosity of time and interest in our research. Within the QIMRB studies, the ‘Cannabis’ and ‘Personality disorders’ studies were funded by National Institute on Drug Abuse (NIDA) grant: DA18267 and facilitated through access to the Australian Twin Registry, a national resource supported by an Enabling Grant (ID 628911) from the National Health & Medical Research Council. We thank Dixie Statham, Richard Parker, Soad Hancock, Judith Moir, Sally Rodda, Pieta-Maree Shertock, Heather Park, Jill Wood, Pam Barton, Fran Husband and Adele Somerville, who worked on this project. We acknowledge the assistance of research assistants Natalie Garden and Reshika Chand and the support staff of the Genetic Epidemiology Laboratory at QIMR Berghofer Medical Research Institute. The ‘16Up’ study was supported by the Young and Well Cooperative Research Centre, which was established and funded under the Australian Government’s Cooperative Research Centres Program. The ‘25Up’ study was funded by NHMRC Project Grant APP1069141 to IBH and NGM. We also thank David Smyth and Scott Gordon for IT support.
Funding
NTR warmly thanks all twin families and would like to acknowledge the National Institutes of Health (NIH, R37 AG033590-08 to J Cacioppo), NWO-Groot 480-15-001/674: Netherlands Twin Registry Repository; the Avera Institute for Human Genetics, Sioux Falls, South Dakota (USA) for zygosity assessment and the Amsterdam Public Health (APH) research institute.
DIB acknowledges her Royal Netherlands Academy of Science Professor Award (KNAW PAH/6635). LC-C is supported by a QIMR Berghofer fellowship. IBH is supported by a NHMRC Australia Fellowship (No. 511921).
Open access
