Social disadvantage and borderline personality disorder: a study of social networks

Abstract

Examining differences in social integration, social support, and relationship characteristics in social networks may be critical for understanding the character and costs of the social difficulties experienced of borderline personality disorder (BPD). We conducted an ego-based (self-reported, individual) social network analysis of 142 participants recruited from clinical and community sources. Each participant listed the 30 most significant people (called alters) in their social network, then rated each alter in terms of amount of contact, social support, attachment strength and negative interactions. In addition, measures of social integration were determined using participant’s report of the connection between people in their networks. BPD was associated with poorer social support, more frequent negative interactions and less social integration. Examination of alter-by-BPD interactions indicated that whereas participants with low BPD symptoms had close relationships with people with high centrality within their networks, participants with high BPD symptoms had their closest relationships with people less central to their networks. The results suggest that individuals with BPD are at a social disadvantage: those with whom they are most closely linked (including romantic partners) are less socially connected (i.e., less central) within their social network.

Social disadvantage and BPD: a study of social networks

Chronic dysfunction in social relationships is a core feature of borderline personality disorder (BPD; Choi-Kain, Zanarini, Frankenburg, Fitzmaurice, & Reich, 2010). Researchers have recently worked to better characterize the nature of social difficulties, finding relationships among individuals with BPD are characterized by high levels of conflict (Stepp, Pilkonis, Yaggi, Morse, & Feske, 2009), marked by opposing fears of abandonment and dependency (Melges & Swartz, 1989), and are generally more unstable (Bouchard & Sabourin, 2009). Individuals with BPD are thought to particularly struggle in close relationships, which could otherwise confer strong mental health benefits (Beckes & Coan, 2011; Fonagy & Luyten, 2009). For example, attitudes towards individuals with BPD are often negative and stigmatizing (Bodner, Cohen-Fridel, & Iancu, 2011). Mental health care workers report less empathy towards individuals with BPD than other mental health patients and seek greater social distance from individuals with the disorder (Markham, 2009). This suggests that even in acute care settings, individuals with BPD are at a social disadvantage. The purpose of the current study was to examine, using social network analysis (SNA), whether BPD symptoms were associated with signs of social disadvantage on a broader level, within individual social networks.

Various methods have been used to assess interpersonal functioning in those with BPD, including global self-report measures, ecological momentary assessment (EMA), experimental paradigms, behavioral observation, and psychophysiological and neuroscience-based tasks (Lazarus, Cheavens, Festa, & Rosenthal, 2010). Each method has advantages and disadvantages and reveals different aspects of interpersonal phenomena in BPD. For instance, whereas global measures can provide an overall index of relationship functioning, EMA captures sequences of interpersonal behavior (Berenson, Downey, Rafaeli, Coifman, & Paquin, 2011). At the same time, it is difficult to parse specific information about relationship partners from EMA. In contrast, variation in the type of interpersonal dysfunction across specific types of relationships, is well captured by assessing individual’s social networks. Whole-network analyses involve querying everyone in a network (e.g., a community). However, this approach is expensive and time-intensive, particularly for mental health research, where assessing hundreds of participants may yield only a handful of people with the disorder of interest. Ego-based social network analysis (Borgatti, Everett, & Johnson, 2013) requires a single participant to generate their own social network. This approach provides a more tractable scale used to generate the people (alters) and connections (edges) among people from an individual’s perspective (i.e., the ego; see figure 1 for a visual depiction of the approach). Cross-sectional, ego-based SNA, provides a snapshot of a person’s social world. As opposed to other methods that may speak to levels of social support or conflict indirectly, this snapshot helps to obtain a more objective picture of social functioning in BPD by allowing the measurement of specific indices of network health.

Figure 1.

Glossary of social network terms.

Ego: the participant. Ego-centered networks are distinct from whole-networks in that in the former, only one person, the ego/participant, provides information on the network. In ego-based networks, the participant is necessarily connected to all other people in the network.

Alter: person in the participant’s network. These are people identified by the participant as belonging to their social network. In the present study, participants were asked to name 30 people “who have been the most significant in your life in the past year. At a minimum, these should be people who you can recognize by sight or by name, who recognize you by sight or by name, and who you could contact if you had to.”

Edge: connection between alters. In the present study, these connections are determined by asking the participant: “How close is the relationship between [Alter 1] and [Alter 2]”.

Centrality: refers to how well connected an alter is to other alters in the network. In the above network, A is the most central alter, having the most connections, whereas B is the least connected alter, having a connection only with the participant.

SNA has been used extensively, particularly in studies of physical health (Cohen, 2004), in which researchers have shown that characteristics of an individual’s social networks have robust effects on various diseases and health conditions (House, Landis, & Umberson, 1988). In conceptualizing the mechanisms of the effect of social networks on health, various theorists have developed models that can inform the study of the psychological impact of social networks in BPD. For instance, Cohen (2004) elaborated on a model in which social support, negative interactions and social integration impact health through psychological variables. In Cohen’s model, social support is defined as instrumental, informational and emotional support from others intended to help an individual cope with stress, and is thought to affect health through the mechanism of stress buffering. Negative interactions, including arguing, criticism and negative influence, increase stress, which increases risk for disease and psychological difficulties. Social integration is the degree to which a person is actively engaged in a range of relationships and feels a sense of belonging. Among other indicators, social integration can be measured by examining one’s connections to others who are well-connected. Cohen argued that social integration has a direct effect on health by encouraging the development of identity, purpose, self-worth and positive affect. Each of these psychological factors are characterized as problems amongst individuals with BPD. Examining the relationship between BPD and social factors could potentially reveal the social disadvantages faced by individuals with BPD symptoms.

Attachment theory represents another account of how the character of close relationships and an individual’s representation of those relationships can affect stress, coping and mental health. Theorists characterize secure attachment with another as seeking proximity to the person in times of stress, feeling anxiety regarding significant separation, being able to use the person as a secure base to return to when anxious and regarding the person as a safe haven with whom to explore problems (Fraley et al., 2006). Possessing mental models of close others as supportive, trustworthy, available and collaborative has been robustly associated with positive mental health outcomes (Shaver & Mikulincer, 2007). Though less central to attachment theory, researchers have even found attachment insecurity is associated with poorer physical health (McWilliams & Bailey, 2010), consistent with social support models of physical and mental health. Theorists and researchers have consistently characterized BPD as a severe attachment disturbance (Gunderson & Lyons-Ruth, 2008). Given difficulties in close relationships, BPD symptoms are likely associated with difficulty maintaining a strong attachment with others in their netowork.

The level of closeness, social support and conflict in a social network likely relates somewhat to how connected members of the network are with each other. Even more directly, graph measures – properties that are calculated from the patterns of connections within a network – are measures that provide information about whether the network members are well-connected overall (density), and how well each individual network member is connected to other members in the network (centrality). Density refers to the degree of interconnectedness amongst people in the network, and is thought to indicate group cohesion (Wasserman & Faust, 1994). Alternatively, degree centrality is among the most common ways for measuring the position of an individual (an individual alter) within a focal person’s social network. Alters who have higher centrality are more directly connected to others in the network, are more readily able to access network resources, and thus have greater influence and power (Brass & Burkhardt, 1993). Though the participant/ego has at least some connection to each alter listed, the strength of these connections varies, and thus, some people may have stronger connections to people in their networks who are more strongly connected to others in their network. Research has shown that people who are more depressed tend to be less central in social networks (Rosenquist, Fowler, & Christakis, 2010), while those who transition to greater centrality tend to later also become happier (Fowler & Christakis, 2008). Considering these findings, centrality is likely to be low amongst individuals with BPD, given stigma and interpersonal problems such as intense anger and unstable relationships (Bodner et al., 2011). Relatedly, researchers have also found that within organizations, women and people of color have fewer network ties, which limits their own status and opportunities within the organization (James, 2000; Seidel, Polzer, & Stewart, 2000). Similarly, people who are connected to more central figures are likely to experience more positive outcomes (Rosenfeld, Richman, & Bowen, 2000). Thus, assessing whether centrality of important relationship partners differs for those with BPD may be important for understanding potential contributors to social and emotional difficulties. Being closer to and spending more time with such a person is a social advantage, because that person is better socially connected, has more access to information and social resources. Having a well-connected romantic partner or strong-attachment figure who is central to the network is a significant advantage.

There is limited research on the social networks among individuals with BPD. Clifton and colleagues (Clifton, Pilkonis, & McCarty, 2007) assessed the social networks of 22 participants (11 diagnosed with BPD). They found networks of those with BPD contained a greater number of former romantic partners and cutoff relationships compared to individuals not diagnosed with a PD. Networks of individuals with BPD were also characterized by less social support from more central relationships. Lazarus and colleagues have recently completed two additional reports, one with college females (n = 127) with varying levels of BPD features and the other with BPD patients (n = 27) and healthy controls (n = 26). In the first study {Lazarus:2016ik}, the authors found BPD features predicted lower ratings of social network satisfaction and more conflict and criticism a month later. In the second (Lazarus, Southward & Cheavens, 2016), compared to healthy control women, women with BPD had smaller social networks, rated relationships as less satisfying and supportive, and more conflictual and critical. In addition, BPD participants reported more relationship ruptures within the previous month. Overall, social network studies in BPD demonstrate more interpersonal disturbance in networks including less support and greater conflict.

Current Study

The current study extends these previous findings by evaluating structural aspects of social networks in a large, diverse sample with significant psychopathology. We had a number of hypotheses, which utilize Cohen’s framework of social influences on health (social support, negative interactions and social integration), and attachment theory as guides. First, we predicted that BPD symptoms would be associated with decreased social support, more negative interactions in the social network and perceptions of social network members, such as less strong attachment ratings. In addition, we anticipated that BPD symptoms would be associated with poorer social integration, due to the interpersonal difficulties and social cognitive problems associated with the disorder (Lazarus et al., 2010). The difficulties individuals with BPD experience, particularly in close relationships, were expected to limit opportunities with central figures. Specifically, we expected the most significant relationships of those with high BPD symptoms to be with people who were less central to their network. We also sought to examine this hypothesis by examining BPD as a moderator of centrality of romantic partners, predicting high BPD would be associated with low degree centrality among these significant figures.

Methods

Participants

Potential participants completed a screening interview, using the McLean Screening Instrument for Borderline Personality Disorder (MSI-BPD; Zanarini et al., 2003), either over the phone or face-to-face to determine eligibility. The MSI-BPD, using a minimum of seven endorsed items, has been shown to have good sensitivity (0.81) and specificity (.85) relative to a BPD diagnosis (Zanarini & Vujanovic, 2003). Our recruitment criteria were intended to capture the full spectrum of BPD features. The recruitment protocol used three strata (0–2, 3–4, or 5 or more diagnostic criteria endorsed on the MSI-BPD) and two groups (community sample and psychiatric sample). Mean MSI-BPD scores were 6.69 (SD = 2.51) for the clinical group and 3.20 (SD = 2.99) for the community group. The patient sample was recruited via promotional flyers in outpatient treatment facilities at the Western Psychiatric Institute and Clinic (WPIC) and through research networking systems. The community sample was recruited by telephone through the use of a random digit dialing method coordinated by the University Center for Social and Urban Research at the University of Pittsburgh. Random digit dialing was utilized to yield a probability sample representative of demographic characteristics reflected in the U.S. census for the Pittsburgh metropolitan area with oversampling of African Americans. Participants with psychotic disorders, organic mental disorders, mental retardation and major medical illnesses that influence the central nervous system (e.g., stroke, Parkinson’s disease) were excluded. All participants were between the ages of 21 and 60 at the time of the screening.

Using these methods, 150 participants enrolled in the study, including 75 psychiatric patients and 75 community members. More detail on the recruitment procedures and sample can be found elsewhere (Scott et al., 2013; Scott, Stepp, & Pilkonis, 2014). Recruitment criteria were designed to sample a broad range of BPD features within both clinical and community samples. The community sample was selected to represent a range of psychopathology in a non-treatment-seeking population, rather than a healthy control sample. Clinical consensus ratings were made for each of the nine DSM-IV borderline personality disorder criteria, rating each criterion on a scale of 0 (not present), 1 (present), or 2 (strongly present). These nine ratings were summed to form a total borderline score for each participant, with a possible range from 0 to 18. In the current sample, dimensional borderline scores ranged from 0 to 12, with a mean of 2.66 (3.34).

Due to a computer error in which social network data did not write to the disk after task completion, data on six participants were lost. An additional participant provided complete data for only seven of the 30 social network members, whereas another included deceased relatives and pets. Both were removed from the analysis. The final analyses therefore include 72 psychiatric patients and 70 community members (total N=142). Ninety-three participants were rated as having two or fewer BPD criteria, 24 met three or four criteria (16 from the clinical sample), and 25 met five or more criteria (20 from the clinical sample). Demographics of the two subsamples were comparable (see Table 1). As in previous studies, the community and clinical groups were combined into a single sample of 142 participants, reflecting our dimensional sampling frame. Of the 4,260 alters collected for these participants, data on one or more questions regarding an alter was missing in 53 (1.2%) instances.

Table 1.

Demographics for participants sampled from clinical and community settings.

	Clinical Participants (n = 72)	Community Participants (n = 70)
Female, n (%)	50 (67%)	47 (63%)
Age, M (SD)	43.15 (9.79)	46.56 (10.82)
Education, n (%)
Education beyond HS	48 (67%)	54 (74%)
HS graduate	15 (21%)	12 (17%)
Did not complete HS	9 (13%)	4 (6%)
Ethnicity, n (%)
European American	41 (57%)	40 (47%)
African American	27 (38%)	27 (39%)
Biracial	4 (6%)	2 (3%)
Asian	0 (0%)	1 (1%)
BPD symptoms, M (SD)	3.93 (2.42)	1.31 (3.53)
BPD diagnosis, n (%)	20 (27%)	5 (7%)
Any PD diagnosis	62 (83%)	24 (32%)
Affective Disorders	48	13
Anxiety Disorder	29	17
Substance Use Disorder	20	11

Note. BPD symptoms refers to number of criteria met according to clinician rating. BPD Diagnosis refers to number of individuals rated as meeting threshold for full diagnosis.

Procedure

All participants were interviewed by trained clinicians who had a master’s degree or doctoral degree and at least five years of assessment/clinical experience. At the initial meeting, clinicians described the study in detail and obtained written, informed consent from the participant. During the intake process, participants were interviewed by a single interviewer for a minimum of four 2-hour sessions. Data collection methodology included structured interviews for assessing the nature and severity of Axis I and II diagnoses, using the Structured Clinical Inventory for DSM-IV disorders (Pfohl, Blum, & Zimmerman, 1997) and the Structured Interview for DSM-IV Personality Disorders (SIDP-IV; Pfohl et al., 1997), and semi-structured interviews outlining social and environmental history (Pilkonis, Heape, Pilkonis, Lambert, & Proietti, 1989). At the conclusion of the intake process, a consensus diagnostic case conference was conducted by a consensus team of at least three judges. At the case conference, the primary interviewer presented all historical and concurrent information collected during the intake process to the consensus team for review and discussion. A complete description of the consensus rating process used in this research group has been provided in previous reports (Pilkonis, 1995).

Participants completed an ego-centered social network assessment, using the computer program EgoNet (McCarty, 2004). They were first asked to list 30 individuals “who have been the most significant in your life in the past year. At a minimum, these should be people who you can recognize by sight or by name, who recognize you by sight or by name, and who you could contact if you had to.” After listing each name, participants completed a brief survey about each of the 30 alters, including demographic information and the nature of the relationship, length of the relationship, whether they had ever been romantically or sexually involved, and whether the participant had cut off contact with the alter in the past year. Participants then estimated the frequency (on a 6-point scale from “Not at all in the past year” to “about once per day or more”) of various interactions with each alter, including frequency of face-to-face interaction, and arguments or fights. For each alter, participants were then asked a series of relationship questions including ratings (on a 4-point scale from “Not at all” to “Very”). These prompts queried how much the participant trusts, relies on for practical advice, relies on for emotional support, and feels criticized by each alter, as well as the overall perceived closeness of his or her relationship with each alter. Finally, for each alter, the participant also completed four questions adapted from Tancredy and Fraley’s (2006) attachment measure. These questions included assessment of proximity seeking (“It is important to me to see or talk with X regularly”), difficulty with separation (“X is a person whom I do not like to be away from”), safe haven (“X is the first person that I would turn to if I had a problem”), secure base (“X is the person that I would actually count on to always be there for me and care for me no matter what”). For each of these, participants rated (on a 7-point scale from “Strongly disagree” to “Strongly agree”) the extent to they agree with each statement.

To determine the structural characteristics of the social network, participants were asked to rate how well each pair of alters knew one another. For each pair of alters, participants were asked “How close is the relationship between [Alter 1] and [Alter 2]”. Response options ranged from 1 (“They have never met”) to 5 (“Very close”). Because ratings of 1 (“They have never met”) or 2 (“They barely know each other”) indicate lack of meaningful, functional connection among alters, both of these were coded as “not connected”, whereas connections rated as 3–5 were retained as different weights. This information was then used to generate a network diagram. This diagram was shown to the participant, who was given the opportunity to identify any errors in his or her structural tie ratings. Each participant reported on exactly 30 alters in his or her social network. From the network data, we calculated graph-level density and alter-level centrality. Density is simply calculated by dividing the number of possible connections between alters in the social network by the number of actual connections present. Degree centrality is the most common measure of centrality used in SNA (Wasserman & Faust, 1994). Degree centrality is the product of the number of connections between an alter and each other alter in the network and the strength of those connections (connections present × how close), divided by the product of connections and weights possible in the network (connections possible × max closeness).

The social network included a number of a priori constructs. Social support was defined by several indicators: perception of less closeness, trust, support and advice among alters in the network overall. Negative interactions were defined by two indicators: more arguing and criticism between the participant and alters. Attachment strength was defined by lower proximity seeking, separation anxiety, and perceptions of the alter as a safe haven and secure base. We examined individual variables, rather than combined scores, in assessing the relationship between these variables and BPD in order to assess more specific patterns of social perception associated with BPD (using combined scores for each construct would not qualitatively alter the results). Social integration was operationalized in two ways, consistent with definitions in the literature (Berkman & Glass, 2000; Minzenberg, Vinogradov, & Poole, 2006; Valente & Foreman, 1998), by examining the relationship between BPD and network density, and the relationship between BPD and the degree centrality of close others.

Results

Data analytic strategy

We first conducted mixed effects random intercept models (see Table 1) assessing the relationship between BPD dimensional scores and ratings of closeness, trust, instrumental advice, and emotional support (closeness and support), frequency of arguments with alter, amount of criticism from alter (negative interactions), and proximity seeking, separation, safe haven and secure base (attachment strength). Using the igraph package in R (Csardi & Nepusz, 2006), we calculated density on each participant’s network data. Additionally, using the same package, we calculated degree centrality for each of the alters in each participant’s network. For the latter, we used these measures as dependent variables in several mixed effects models evaluating BPD as a moderator of the relationship of indicators of closeness with alters and centrality. We selected a number of variables that are indicators of closeness with a particular alter. These include face-to-face time, attachment ratings, closeness and support and frequency of arguments. For brevity, we chose to use the combined scales of attachment ratings and closeness and support. However, because previous research (Cohen & Wills, 1985), and our current data suggests people argue more with those with whom they are close, we selected this individual variable, rather than combining it with a perception of criticism from an alter. Results are organized for congruence with hypotheses.

Social Support

We tested the hypothesis that BPD dimensional scores would be associated with reports of lower social support. A mixed effects random-intercept model using the combined closeness and support variable revealed higher BPD was associated with perception of less closeness, trust and support of others (see Table 1). Higher BPD was also associated with reports of being less likely to go to others for emotional support. Unexpectedly, BPD scores were not associated with reports of being less likely to seek instrumental support from others.

Negative Interactions

We expected that BPD symptoms would be associated with more negative interactions. As expected, higher BPD scores were associated with more frequent arguments with and criticism from alters (Table 2).

Table 2.

Mixed Effects Models: borderline personality disorder dimensional scores predicting participant self-report on social network members.

	Estimate	SE	F(1, 141)
Closeness and Support
Closeness	−0.04	0.01	12.15***
Trust	−0.04	0.01	15.14***
Advice	−0.03	0.01	7.83**
Support	−0.02	0.01	ns
Attachment Strength
Proximity Seeking	0.00	.02	ns
Separation	0.00	.00	ns
Safe Haven	0.04	0.02	ns
Secure Base	−0.01	0.03	ns
Negative Interactions
Argue	0.03	0.01	7.75**
Criticism	0.04	0.02	5.58*

^***p < .001,

^**p < .01,

^*p < .05,

ns = not significant

Note: Closeness= “How close do you currently feel to X?”, Trust= “How much do you trust X?”, Advice= “How likely would you be to go to X for advice about a professional or practical problem?”, Support= “How likely would you be to go to X for emotional support if you were feeling sad or upset?”, Proximity Seeking= “It is important to me to see or talk with X regularly”, Separation= “X is a person whom I do not like to be away from”, Safe Haven= “X is the first person that I would turn to if I had a problem”, Secure Base= “X is the person that I would actually count on to always be there for me and care for me no matter what”, Argue= “On average, how often do you argue or fight with X?”, Criticism= “How critical is X of you?

Attachment strength

We had hypothesized that higher BPD scores would be associated with rating other network members as having lower attachment strength. Contrary to hypotheses, BPD was not associated with perception of lower attachment strength on any variable.

Social Integration

We anticipated that individuals with BPD would have more sparsely connected social networks, measured using network density. On a network level, density represents the proportion of connections present in the social network out of the number of possible connections. Network density was not significantly correlated with BPD, r(141) = −.04, p > .05.

We hypothesized that BPD dimensional scores would moderate the relationship between indicators of closeness (i.e., face-to-face time) in a relationship and the degree centrality of the alter. Specifically, we hypothesized that low-BPD would be associated with a pattern in which there was a strong effect between how close the alter was rated and how central their position was in the participant’s network. Further, we hypothesized high-BPD scores would be associated with a pattern of less differentiation in centrality based on closeness ratings of the alter. Using mixed-effects random intercept models, we found support for these predictions. In all analyses, we controlled for the amount of time the participant knew each alter, with the assumption that alters who had been in a network longer, were more likely to be more central. We found significant interactions between BPD and each relationship variable tested in predicting degree centrality: face-to-face time (F(1, 4099) = 20.48, p < .001), attachment ratings (F(1, 4108) = 4.10, p = .043), closeness and support (F(1, 4093) = 6.09, p = .014) and frequency of arguments (F(1, 4108) = 6.20, p = .012). Graphs of the interactions are found in Figure 2, where low and high BPD are defined as 1 standard deviation above and below the mean. In each case, whereas presence of few BPD symptoms was associated with a strong relationship between centrality and face-to-face time, attachment ratings, closeness and support and frequency of arguments, high BPD scores were associated with less differentiation between central and peripheral figures on these variables.

Figure 2.

Interaction between relationship variables and BPD.

Dependent variable for each interaction is degree centrality. All interactions are level-1 alter characteristics by level-2 participant BPD dimensional scores. Face-to-face time is participant rating of how much time spent with alter. Attachment Rating is the sum of ratings of attachment strength (see table 1). Closeness/support is the sum of the four closeness/ support items. Frequency of arguments is the participant’s rating of how often they argue with each alter. High and low BPD is defined as 1 standard deviation above the mean, and one standard deviation below the mean.

Significant relationships

We also hypothesized that romantic partners of participants with high-BPD symptoms would be less centrally located within networks than partners of participants with low-BPD symptoms. Again, using multilevel, random-intercept models we found BPD-by-romantic partner interactions. As can be seen in Figure 3, whereas current romantic partners were central to the networks of those with low BPD symptoms, current romantic partners were peripheral to the networks of those with high BPD symptoms (F(1, 4054) = 28.43, p < .001). Because there is more turnover in the social networks of individuals with BPD, we investigated whether any of the interactions may be due to individuals with high BPD scores having known people in their networks for less time. Though years known was a strong predictor of most relationship dependent variables (i.e., closeness and support), it did not account for significant interactions in any instance.

Figure 3.

Relationship characteristics with romantic partners for participants with high and low BPD symptoms.

Supplemental to understanding the social position of romantic partners of people with BPD, we were also interested in the nature of these relationships. Our primary question in this instance was whether participants rated having less closeness and support, spending less time, and feeling less strongly attached to romantic partners. We investigated other differences in romantic relationships as seen in Figure 2. Significant interactions were also present for romantic partners when considering attachment ratings (F(1, 4050) = 7.46, p = .006) and face-to-face time (F(1, 4073) = 5.02, p = .025, but not for closeness. High-BPD was associated with weaker attachment and less face-to-face time with romantic partners.

Discussion

In this study of the character of social networks in a sample selected for a high degree of problems with personality functioning, we found evidence that individuals with BPD show a number of differences in the qualities of relationships, particularly with central figures in their social networks. Our overarching proposition was that the social networks of individuals with high BPD symptoms would have several indicators of social disadvantage. We hypothesized that participants with BPD would report less social support, more negative interactions, less strong attachment to alters and less social integration. On the whole, except for our expectation regarding attachment, these hypotheses were supported.

We found higher BPD symptoms were related to perception of poorer social support from alters in their networks. Social support was operationalized as having close, trusting relationships in which they felt they could go to alters for support. In addition, they perceived more negative interactions with alters in their networks, operationalized as more frequent arguing and criticism. In the emerging research available on the social networks and BPD, participants with BPD have consistently reported feeling less supported by, and experiencing more negative interactions with the alters in their networks (Clifton et al., 2007). The present study adds to this literature by confirming these findings in a large sample with high rates of psychopathology. It is unclear whether reports of reduced social support in this or previous studies is a reflection of an actual deficit in the presence of supportive others or simply perception of reduced support despite adequate resources for the provision of social support. BPD has frequently been characterized as a preoccupied style of disorganized attachment (Gunderson & Lyons-Ruth, 2008; Melnick, Patrick, Lyons-Ruth, & Peter, 2007). Preoccupied attachment is associated with feeling less benefit from equally supportive behavior of a partner (Collins & Feeney, 2000), as well as seeking more support, seeing oneself as in need of help, and being chronically dissatisfied with help provided (Campbell, Simpson, Boldry, & Kashy, 2005). Though this distinction between perception and objective behavior is important for developing effective treatments, it may not be a critical distinction in predicting mental health. Perception of insufficient social support, regardless of objective observations is a key predictor of poor mental health (Cohen & Wills, 1985). Although the current research also cannot parse perception of greater conflict from objectively greater conflict in the social networks of individuals with BPD, previous research supports that BPD is associated with indices of conflict, such as more aggression, using objective laboratory measures (Beeney, Levy, Gatzke-Kopp, & Hallquist, 2014; Dougherty, Bjork, Huckabee, Moeller, & Swann, 1999).

Inconsistent with our hypotheses, BPD was not associated with attachment strength. One likely possibility for lack of support for this hypothesis is that we did not fully consider the nature of attachment difficulties among individuals with BPD in making this hypothesis. Rather than global problems in attaching to others, a better consideration of the literature suggests that individuals with BPD will experience greater difficulty in attachment-based relationships (e.g., romantic partners, therapists; (Lyons-Ruth, Melnick, Patrick, & Hobson, 2007). In addition, our attachment questions (e.g., X is someone I would not want to be away from) appear to measure strength of attachment, rather than attachment disturbance. It is possible to have a strong attachment to someone, at the same time that the attachment is insecure, as would be expected among individuals high in preoccupied attachment.

We also anticipated that BPD would be associated with signs of poorer social integration. We found support for this hypothesis. Central figures in social networks are individuals with the greatest potential for bestowing support, information and influence (Tsai & Ghoshal, 1998). We expected those with BPD to have their closest relationships with people less central to their social networks. For people low in BPD symptoms variables related to closeness (time spent together, closeness/support, arguing) were strongly associated with greater degree centrality. For those with high BPD symptoms these variables were weaker predictors of centrality of the alter. This suggests individuals with BPD are closest with people who are more peripheral to their own social network. Further examining this hypothesis among one type of primary relationship figure, we found individuals high in BPD symptoms evidenced centrality differences among romantic partners. The current romantic partners of those high in BPD were peripheral to their social networks, whereas romantic partners of those with low BPD symptoms were among the most central figures. We did not find that the networks of individuals with BPD were less connected overall (i.e., lower density). However, high density (our measure of social network interconnection) does not necessarily mean that the network is functioning well or is globally supportive. This interpretation is consistent with our results, in which individuals with BPD reported less support among an equally connected network. In addition, the people with whom people with high BPD symptoms were closest, were less well connected to others in the network. Overall, individuals with high-BPD symptoms had their closest relationships with people who are not the “hubs” of their social network.

In total, poorer social support, greater conflict, and poorer social integration, suggests those with BPD lack the benefits that a social environment is capable of providing: belongingness, support, and, more generally, the multitude of benefits of connections to others who are well-connected. Paris (1992) previously hypothesized that social risks such as social disintegration and breakdown of social norms is linked to increased incidence of behaviors associated with BPD (attempted suicide, substance use). The current results provide some support for this hypothesis. Past research has shown that a person’s social networks impact health and mental health (House et al., 1988). In addition, recent theories have underlined the social aspect of emotion regulation (Hughes, Crowell, Uyeji, & Coan, 2011). The current research suggests that the social context of individuals with BPD may not be supportive of health, mental health and emotion regulation. Consistent with the connection between social networks and physical health, in a nationally representative sample, BPD was associated with heightened incidence of numerous medical conditions, after controlling for sociodemographic variables and other disorders (El-Gabalawy, Katz, & Sareen, 2010). One mechanism of that link may be the social networks of individuals with BPD. Indeed, the perceived deficit in close, supportive and influential alters in the social network may exacerbate aspects of the disorder, from employment and work functioning to substance use.

Though a number of clear social differences exist between individuals with high BPD symptoms and those with few symptoms, the source of these differences is not altogether clear. An interpretive question seems evident: what causes people with BPD to be most closely connected to people who are more peripheral to their network? We consider three possibilities to both illuminate interpretations of the current results and present a framework for future research.

Social cognitive deficit

Our results could reflect a ‘deficit’ in social cognition among individuals with BPD: those with high BPD symptoms are simply less alert to social factors, such as influence and social power, compared to others. In support of this interpretation, past research has found that individuals with BPD show neural and behavioral insensitivity to social norms (King-Casas et al., 2008; Roepke, 2013; Unoka, Seres, Aspán, Bódi, & Kéri, 2009), and some research has demonstrated poor mentalization (Fonagy, Luyten, & Strathearn, 2011; Preißler, Dziobek, Ritter, Heekeren, & Roepke, 2010; Preißler et al., 2013) and emotion recognition (Domes, Schulze, & Herpertz, 2009) among those with the disorder. This may mean that while others gravitate more towards figures who are well-connected in their networks, those with BPD are less alert to this dimension in selecting partners.

Interpersonal difficulties

Another possibility is that interpersonal difficulties experienced by individuals with BPD may lead well-connected others to focus on more harmonious relationships. BPD is associated with significant interpersonal problems characterized by oscillations between neediness and angry withdrawal (Melges & Swartz, 1989), which can generate difficult feelings, even for many trained clinicians. Interpersonal difficulties related to BPD are likely to have evocative effects on social partners. Researchers have provided evidence that people high in rejection sensitivity frequently evoke rejection from others (Downey et al., 1998). Having many social options, some well-connected people may consciously or unconsciously distance themselves from a person with interpersonal difficulties. The possibility is also consistent with the high stigmatization experienced by individuals with BPD (Rüsch et al., 2006), which theorists and researchers suggest prompts emotional distancing among social partners (Aviram, Brodsky, & Stanley, 2006).

Defensive Process

The current findings regarding centrality could also be interpreted as the result of defensive processes: central individuals may pose a greater risk for individuals with BPD, and people with more BPD symptoms defensively avoid these people despite an opposing desire for closeness and support. Researchers have consistently conceptualized individuals with BPD as having complex difficulties with approach and withdrawal (Melges & Swartz, 1989). Due to low trust, poor self-esteem, and intense interpersonal sensitivity, individuals with BPD may both be aware of high-value social contacts, and nonetheless avoid significant relationships with such figures.

Because this study is cross-sectional, and only a snapshot of the social worlds of those with BPD, the mechanisms driving our findings remain unclear. Longitudinal work examining change in social networks over time, full-network designs, or methods assessing networks among romantic partners will be needed to better answer discern how each of these processes may affect social networks of individuals with BPD.

Strengths and Limitations

Strengths of this study include a large sample size of rigorously assessed participants with varying degrees of personality pathology. The evaluation of social influences on and characterization of the social networks of individuals with BPD is relatively rare and provides novel information that can prompt additional research and, with replication, can lead to treatment improvement, focused on improving the quality of social ties of individuals with BPD. Several limitations are notable. The cross-sectional design, as noted, limits the conclusions that can be drawn from these findings. Further research is needed to determine how specific individual differences may affect the structure of social networks among individuals with BPD. Ego-based social network assessment, while much more efficient for targeting clinical populations, is also a limitation. The limitations of ego-based SNA is that it is difficult to discern how changes in the social network (e.g., less support, moves to greater centrality), affect individuals with BPD. Whole-network approaches over time could provide a more definitive picture of the social position of individuals with BPD and their closest alters, and yet these studies are exceedingly difficult to conduct.

Conclusion

When thinking about the potentially positive, stabilizing and supportive aspects social networks can provide, people with BPD are at a disadvantage. The etiology of BPD is typically conceptualized as a reciprocal interaction between individual differences and a less-supportive social/familial context. The type of social/familial context that encourages the development of BPD has been described as less supportive, more invalidating and characterized by more disorganized caregiving (Stepp, 2011). In the current study, we provide some evidence that this relation between BPD and a less supportive social context is also evident in adulthood. Individuals with BPD reported poorer social support, more conflict in their networks, and less connection to important people. Research has frequently suggested individuals with BPD exhibit individual differences (Berenson et al., 2011; Mischel et al., 2008) that are likely to impact others in their social network, such as rejection sensitivity and increased needs for help, mixed with greater defensive hostility. The current study provides evidence that BPD is associated with differences in social networks, and highlights the possibility that these differences may exacerbate and/or maintain BPD symptoms over time, through the mechanisms of poorer social support, greater conflict, and poorer social integration.