Borderline personality and depressive symptomatology: Common psychosocial predictors and comorbidity

Abstract

Objective

Borderline Personality Disorder (BPD) and the affective disorders, in particular depression, share some overlap in symptomatology and a set of common psychosocial determinants. The purpose of the current study was to provide further evidence informing debates of diagnostic distinction by quantifying (1) self‐reported symptomatic comorbidity of the two disorders and (2) relative contributions of parental bonding and early maladaptive schemas to adult levels of borderline and depressive symptomatology.

Method

A general population sample of 411 participants, ranging from 18 to 65 years (99 males and 312 females), completed the questionnaire package.

Results

Retrospectively reported level of parental care was found to account for significant unique variance in adult borderline and depressive symptomatology. Furthermore, schemas of the Disconnection/Rejection domain and others were found to significantly and concordantly predict both borderline and depressive symptomatology. Participants reporting substantive levels of borderline symptoms also reported depressive symptoms outside the normal range.

Conclusions

These findings support significant overlap in the symptomatology, co‐severity and psychosocial determinants of BPD and depression. These data have implications for the management and early prevention of both depression and BPD.

Key words:

• borderline personality
• comorbidity
• depression
• early maladaptive schemas
• parental bonding

Borderline personality disorder (BPD) is one of the most commonly reported personality disorders within the general population (American Psychiatric Association, 2013). Additionally, individuals diagnosed with BPD are the largest consumers of mental health services, consequent to high levels of suicidal behaviours associated with the disorder (American Psychiatric Association, 2013). Indeed, 75% of individuals with BPD attempt suicide at least once in their lives and 10% complete suicide (Kolla, Eisenberg, & Links, 2008). Most completed suicides occur later in the course of the disorder, suggesting current treatment options are to some extent inadequate (Kolla et al., 2008). A potential contributing factor to the high level of suicide attempts is a lack of diagnostic clarity in the current Diagnostic and Statistical Manual of Mental Disorders (DSM‐5). Research has found substantial overlapping Diagnostic Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM‐IV‐TR) symptomatology between BPD and the affective disorders (Koenigsberg et al., 2002), with potential for frequent misdiagnosis and associated suboptimal management and outcomes (including increased suicidal risk). In addition to sharing some common symptomatology and being highly comorbid (Zanarini et al., 1998), research has also suggested that BPD and the affective disorders may share common developmental determinants (Nordman, Holthe, & Haugum, 2005; Zweig‐Frank & Paris, 1991). These findings have led to substantial debate around the need for revision of the BPD classification to achieve a clearer distinction with affective disorders (Acta Psychiatrica Scandinavica & Akiskal, 2004). To further inform this debate, the current study investigated dimensionally operationalized levels of overlapping borderline and depressive symptomatology in a large, general population sample, and sought to identify common psychosocial predictors, specifically parental bonding and early maladaptive schemas (EMS), for these symptoms.

According to the DSM‐5, BPD is currently defined by a pervasive pattern of impulsivity as well as instability in interpersonal relationships and affect, emerging by early adulthood within a variety of contexts (American Psychiatric Association, 2013). Affective instability is an important predictor of suicidality in both BPD and the affective disorders (Acta Psychiatrica Scandinavica & Akiskal, 2004). Additional overlapping symptomatology for both disorders include chronic emptiness and boredom, intense anger, and recurrent suicide ideation. Additionally, research would also suggest commonalities in the kinds of early psychosocial environments which influence the development of both BPD and the affective disorders (Ingram & Ritter, 2000; Nickell, Waudby, & Trull, 2002). Of particular focus in this study are parental bonding and the development of EMS.

Parental bonding refers to the amount of care and protection provided to an individual during their childhood by a parent or caregiver. Investigations of retrospectively reported parental bonding in individuals diagnosed with either BPD or depression in adulthood have found many experienced inadequate parental bonding during their childhood, including reported low levels of care and/or overprotection (e.g., Neale et al., 1994). Within a large sample of 62 male and female BPD patients, Zweig‐Frank and Paris (1991) also found that adults diagnosed with BPD reported their parents to be more overprotective and less caring compared with reports of parental bonding from 99 non‐psychotic control participants, a finding corroborated by Nickell et al. (2002). With respect to the relationship between parental bonding and depression, Ingram and Ritter (2000) confirmed the findings of previous studies reporting lower care and higher overprotection in parents of individuals diagnosed with depression. However, the sample size for this study was small with a substantial gender imbalance (i.e., 26 woman and 9 men) which, while reflecting the general population epidemiology of depression, may limit the generalizability of their findings. In a larger sample of 248 outpatients (115 males and 133 females) with a primary diagnosis of depression, Carter, Joyce, Mulder, Luty, and Sullivan (1999) found that lower levels of parental care (including both maternal and paternal care) were negatively associated with recurring depression and the development of substance use disorders, psychiatric symptoms, and personality disorders.

Past studies have generally failed to consider the use and value of non‐clinical samples, which is particularly limiting given the clear dimensionality of BPD and depressive symptomatology (Carr & Francis, 2009; Trull, Widiger, & Guthrie, 1990). Carr and Francis (2009) discussed in further detail the notion that BPD represents the extreme end of a dimensionality continuum of normal personality characteristics. To our knowledge, no study has directly assessed levels of parental bonding and schemas relative to borderline or depressive symptomatology in a large, general population sample, representing a significant gap of knowledge in this area.

While many children report experiencing adverse childhood environments, including insufficient parental bonding, not all will develop borderline or depressive symptomatology. It may be that the manner in which childhood experiences are interpreted is critical to the formation of either resilience or maladaptive pathological process. In an attempt to understand why a particular experience has occurred, some children may develop EMS, which are described as broad, pervasive belief themes regarding the self and one's relationship with others (Young & Brown, 2003). Schemas develop during childhood via the interaction between innate dispositions and early family environment (Nordman et al., 2005). Additionally, EMS can contribute to dysfunctional lifestyle patterns that are elaborated throughout one's lifetime and are linked to adult pathology (Nordman et al., 2005; Zweig‐Frank & Paris, 1991). Young and Brown (2003) developed the Young Schema Questionnaire‐SF (YSQ‐SF), which measures 15 EMS categorised into five schema domains: disconnection/rejection, impaired autonomy, impaired limits, other‐directedness, and over vigilance. While they reflect ongoing patterns developed earlier in life, these EMS are indexed on an individual's current cognitions and are therefore, in their reporting, not retrospective in nature. A recent revision by the authors has yielded an additional three EMS (Young, 2012), which were not measured in the current study.

Limited research has explored the specificity of schema domains in personality disorders. A study conducted by Jovev and Jackson (2004) provided evidence that individuals with BPD symptoms reported significantly higher scores on disconnection/rejection domain compared with obsessive compulsive personality disorder; this domain also yields strong significant associations with depressive symptomatology in adults (Halvorsen et al., 2009). The schemas within the disconnection/rejection domain include abandonment/instability (AB), mistrust/abuse (MA), emotional deprivation, defectiveness/shame (DS), and social isolation/alienation. Relationships with BPD or depression symptoms have also been found for the schemas of dependence/incompetence (DI) and vulnerability to harm (VH) and insufficient self‐control (IS) (Butler, Brown, Beck, & Grisham, 2002; Jovev & Jackson, 2004). Hence, research has found that adults with BPD or depressive symptoms experience similar maladaptive schemas across many schema domains, which is suggestive of similar psychosocial developmental origins and course. However, caution must be taken when interpreting the findings from these past studies since, as with respect to the parental bonding literature described, they have relied predominately on clinical samples and are thus limited in their generalisability.

We would contend, once again, that with respect to the development of understandings of early psychosocial influences on symptoms and traits broadly and dimensionally represented in the population, it is important to determine the relative association of these traits with predictor variables in general population samples, and that this represents an important gap of knowledge currently. Furthermore, no empirical study to our knowledge has directly determined, within a single sample, the concordance of EMS associated with borderline and depressive symptomatology. Therefore, there is a need for research examining overlap between depressive and borderline symptomatology, not only with respect to global symptomatology and associated features, but also with respect to the patterns of childhood parental bonding and cognitive schema domains, which may give rise to this common phenomenology.

Within the broader context of informing the diagnostic classification debate regarding dimensional overlap of BPD and affective disorders, the current study aimed (in an adult, general population sample) to (1) determine the relative contributions of retrospectively reported parental bonding (i.e., care and overprotection) and current EMS to levels of both borderline and depressive symptomatology, and (2) determine levels of self‐reported comorbidity between borderline and depressive symptomatology.

With respect to parental bonding, it was hypothesised that higher levels of both borderline and depressive symptomatology would be associated with lesser parental ‘care’ and greater parental ‘overprotection’. The unique variance explained by the four parental bonding variables (i.e. maternal and paternal care and overprotection) was expected to be similar for both borderline and depressive symptomatology. It was also predicted that the EMS of the disconnection/rejection domain would yield positive associations with both borderline and depressive symptomatology in this non‐clinical sample, and that other EMS explored would reveal further concordance. Of the five schema domains, it was anticipated that the EMS of the disconnection/rejection domain would consistently explain the largest unique variance in levels of borderline and depressive symptomatology. Finally, it was anticipated that there would be a high level of comorbidity between borderline and depressive symptomatology; that is, it was expected that participants who scored high in depressive symptomatology would also report high levels of borderline symptomatology.

Method

Participants

A total of 411 participants, ranging from 18 to 65 years (M = 29.75 years; standard deviation (SD) = 11.44 years) completed questionnaires relevant to the current study and were recruited from a university in Melbourne and the general public via hardcopy and online versions of the questionnaire, the latter via the Survey Monkey website (SurveyMonkey, 2012). The sample consisted of 99 males (M = 31.16 years; SD = 11.69 years) and 312 females (M = 29.30 years; SD = 11.34 years). The ethnic backgrounds of the participants varied, with 266 participants (64.7%) identified as Anglo‐Saxon, 79 (19.2%) reported themselves as European, 16 participants (3.9%) were of Asian background, 8 (1.9%) were of a Jewish background, 2 participants (.5%) were from an Indigenous American background, 1 participant (.2%) was of African background, and 31 participants (7.5%) were from another background not listed within the questionnaire. In addition, eight participants (1.9%) did not report their ethnic background. Of the total sample, 337 participants (82%) reported being employed compared with 74 participants (18%) being unemployed. This would be contradictory to the low employment rate in BPD patient groups; however, was expected for the non‐clinical sample recruited in the current study. Finally, 227 (55.2%) participants were students while 183 (44.5%) participants were no longer studying.

With respect to relevant psychological history, 97 participants (23.6%) indicated some form of past psychological illness: depression (18.2%) and anxiety (10.5%) being the most common. Other issues reported were obsessive compulsive disorder (2.9%), bipolar disorder (2.7%), post‐traumatic stress disorder (2.7%), eating disorder (3.4%), schizophrenia (1.2%), personality disorder (1.2%), and 2.9% of participants reported a psychological illness not listed within the questionnaire.

Measures

The current study is part of a long‐term, large‐scale project investigating more broadly how childhood psychosocial environments influence the development of personality disorders. Only the measures used for the current study are described.

Demographic information

Participants were asked to provide basic demographic information such as their age, sex, occupation, and ethnicity.

Parental Bonding Instrument (PBI)

The PBI is a 25‐item questionnaire designed to measure parental behaviour during the first 16 years of life (Parker, Tupling, & Brown, 1979). There are four subscales within the PBI, which measures the level of care and overprotection exhibited from both parents. Each item within the PBI is rated on a 4‐point Likert scale (0 = very unlikely to 3 = very likely), with a high score on care and a low score on overprotection expressing optimal parenting. The PBI holds strong internal consistency, with Cronbach's alphas ranging from .87 to .90 for the overprotection scale (Willinger, Diendorfer‐Radner, Willnauer, Jorgl, & Hager, 2005), and Cronbach's alphas ranging from .86 to .93 for the care scale (Willinger et al., 2005). The PBI also displays adequate reliability; test–retest reliability shows that the coefficients for care and overprotection are r = .76 and r = .63 respectively (Parker et al., 1979).

YSQ‐SF

There are 75‐items within the latest edition of the YSQ‐SF. This questionnaire examines the EMS proposed by Young and Brown (2003). Each item within the YSQ‐SF is rated on a 6‐point Likert scale (1 = completely untrue of me to 6 = describes me perfectly). The YSQ‐SF demonstrates strong internal consistency with Cronbach's alphas for each of the subscales ranging from .71 to .93, with a mean of approximately .83 in clinical samples (Glaser, Campbell, Calhoun, Bates, & Petrocelli, 2002). Temporal consistency seems adequate with a test–retest coefficient of .76 (Schmidt, Joiner, Young, & Telch, 1995).

The Structured Clinical Interview of the DSM Axis II, Personality Questionnaire (SCID‐II‐PQ)

There are 119 items within the SCID‐II‐PQ, where participants are asked true or false questions. Fifteen items in the SCID‐II‐PQ assess BPD symptomatology. This questionnaire is used to assess personality disorder symptoms and has been validated for non‐clinical samples. The SCID‐II‐PQ covers all 11 Axis II personality disorders together with self‐defeating personality. It holds strong test–retest reliability with kappa values of .70 to .85 (Modestin, Oberson, & Erni, 1997). The SCID‐II‐PQ was chosen in preference to other personality questionnaires because of its strong psychometric properties, and because of its clear and concise relationship to the DSM‐IV criteria for personality disorders (Walters, Moran, Choudhury, Lee, & Mann, 2006). It is important to note however that there is limited consistency between the self‐report version of the SCID‐II‐PQ and the clinician‐rated version (Mulder, Joyce, & Cloninger, 1994). Therefore, when recruiting clinical samples, both self‐report and clinician‐rated scales should be utilised.

Depression, Anxiety, and Stress Scales 21 (DASS‐21)

The DASS‐21 is designed to measure the three negative affect states of depression (seven items), anxiety (seven items), and stress (seven items), and is the short form of the DASS‐42. Participants were given a timeframe of the past week to report their depression, anxiety, and stress states. Each item was rated on a 4‐point Likert scale (0 = did not apply to me at all to 3 = applied to me very much, or most of the time). The DASS‐21 demonstrates strong internal consistency with Cronbach's alphas for the depression scale ranging from .87 to .89, anxiety scale ranging from .80 to .83, and stress scale ranging from .89 to .91 (Henry & Crawford, 2005). The DASS‐21 was used in preference to measures such as the Beck Depression Inventory and the Beck Anxiety Inventory because it is briefer while still retaining excellent reliability, and is well normed for Australian populations. For the current analysis, only depression scores were of interest. As per manual instructions, depression scores were doubled to interpret the different symptom severity ratings, according to the DASS‐42 (Lovibond & Lovibond, 1995).

Procedure

Following Human Research Ethics Committee approval, participants over the age of 18 years were recruited from the general community, via multiple media advertisements, and among students enrolled at a university in Melbourne. Responses were then gathered either online or from a hard copy version of the survey. Completion of the survey took approximately 30-min. Each participant was provided with a plain language statement about the study prior to completing the survey, assuring participants that their individual responses would remain anonymous and that they had the right to withdraw from the study at any time. Participants were not offered any incentive for participation.

Results

Data screening

Missing values

All variables had missing data. Little's missing completely at random test indicated that missing values were randomly distributed, χ² (1413) = 1442.62, p = .29. The missing value analysis procedure was used to replace the missing data using the expectation maximization method (SPSS Inc, 2007).

Detection of outliers and examination of normality

To screen for data entry errors, the maximum and minimum values for each item from the questionnaire were examined; any item which fell outside the range was corrected by referring back to the questionnaire. Histograms for each variable were then examined for possible outliers. Although some variables appeared to display outliers, these were valid values and thus retained in analysis.

Most variables were strongly skewed and log transformations did not improve the distributions; therefore, the untransformed data were analysed. According to Norman (2010), parametric tests, including correlations and regression analyses, are robust to violations of normality assumptions.

Reliability estimates and descriptive statistics

Table 1 presents descriptive statistics for the study variables. Cronbach's alphas fell above the .70 cut‐off reflecting adequate internal consistency (Murphy & Davidshofer, 2001).

Table 1. Descriptive statistics for important study variables (n = 411)

Variable	M	SD	Minimum	Maximum	α
SCID‐II‐PQ—BPD subscale	3.11	2.81	0	13	.80
DASS‐42—depression scale	8.07	7.75	0	42	.87
PBI—maternal care	27.46	7.75	2	36	.93
PBI—maternal overprotection	13.12	6.56	0	37	.84
PBI—paternal care	23.47	8.42	0	36	.93
PBI—paternal overprotection	11.91	6.77	0	36	.86
YSQ‐SF emotional deprivation	10.36	5.71	5	30	.88
YSQ‐SF abandonment/instability	10.28	5.57	5	30	.93
YSQ‐SF mistrust/abuse	10.73	5.36	5	30	.90
YSQ‐SF defectiveness/shame	8.06	4.46	5	30	.92
YSQ‐SF social isolation	10.89	5.96	5	30	.94
YSQ‐SF dependence/incompetence	8.33	3.84	5	30	.77
YSQ‐SF vulnerability to harm	8.91	4.05	5	25	.81
YSQ‐SF enmeshment	7.33	3.36	5	30	.79
YSQ‐SF failure	9.42	4.56	5	30	.92
YSQ‐SF entitlement/grandiosity	11.73	4.25	5	26	.74
YSQ‐SF insufficient self‐control	12.30	5.03	5	30	.86
YSQ‐SF subjugation	9.55	4.47	5	30	.85
YSQ‐SF self‐sacrifice	16.33	5.15	5	30	.85
YSQ‐SF emotional inhibition	10.74	5.09	5	28	.88
YSQ‐SF unrelenting standards	17.25	6.15	5	30	.89

Note. SCID‐II‐PQ = The Structured Clinical Interview of the DSM Axis II, Personality Questionnaire; DASS‐21 = Depression, Anxiety, and Stress Scale 21; PBI = Parental Bonding Index; YSQ‐SF = Young Schema Questionnaire‐Short Form.

Correlations

Pearson's correlations examined associations between study variables (refer to Table 2). Almost all correlations were significant or highly significant, and were assessed for strength according to Cohen's (1988) conventions. Significant weak associations were found between age and BPD and between working status and depressive symptomatology.

Table 2. Pearson's correlations between borderline and depressive symptomatology and other study variables

	1	2
1. SCID‐II‐PQ‐ BPD subscale	–	.44**
2. DASS‐21– Depression Scale	.44**	–
3. Age	−.27**	−.09
4. Working status	.09	.14**
5. PBI—maternal care	−.23**	−.31**
6. PBI—maternal overprotection	.17**	.15**
7. PBI—paternal care	−.24**	−.22**
8. PBI—paternal overprotection	.11*	.15**
Disconnection/rejection domain
9. YSQ‐SF emotional deprivation	.31**	.38**
10. YSQ‐SF abandonment/instability	.54**	.39**
11. YSQ‐SF mistrust/abuse	.47**	.43**
12. YSQ‐ SF defectiveness/shame	.41**	.56**
13. YSQ‐ SF social isolation	.42**	.54**
Mean correlation value	.43	.46
Impaired autonomy and performance domain
14. YSQ‐SF dependence/incompetence	.37**	.40**
15. YSQ‐ SF vulnerability to harm	.37**	.38**
16. YSQ‐ SF enmeshment	.23**	.31**
17. YSQ‐ SF failure	.34**	.49**
Mean correlation value	.33	.40
Impaired limits domain
18. YSQ‐ SF entitlement/grandiosity	.20**	.05
19. YSQ‐ SF insufficient self‐control	.39**	.45**
Mean correlation value	.30	.25
Other‐directedness domain
20. YSQ‐ SF subjugation	.44**	.46**
21. YSQ‐ SF self‐sacrifice	.21**	.20**
Mean correlation value	.33	.33
Over vigilance and inhibition domain
22. YSQ ‐ SF emotional inhibition	.33**	.44**
23. YSQ‐ SF unrelenting standards	.11*	.08
Mean correlation value	.22	.26

Note. SCID‐II‐PQ = The Structured Clinical Interview of the DSM Axis II, Personality Questionnaire; DASS‐21 = Depression, Anxiety, and Stress Scale 21; PBI = Parental Bonding Instrument; YSQ‐SF = Young Schema Questionnaire‐Short Form.

*p < .05 level, two‐tailed; **p < .01 level, two‐tailed.

There was a moderate to large correlation between levels of borderline and depressive symptomatology. Parental bonding variables showed weak associations with both borderline and depressive symptomatology, except for the association between maternal care and depressive symptomatology, which was moderate. Indeed, the overall strength and direction of correlations with parental bonding variables were highly consistent between borderline and depression symptomatology variables.

With only two exceptions, where correlations were non‐significant, moderate to strong associations were observed between EMS from all schema domains and both borderline and depressive symptomatology (refer to Table 2 for the specific coefficient values for each EMS). There was some variability with respect to the average correlation coefficient values across the five schema domains for both borderline and depressive symptomatology; however, they remained within the same strength categories in accordance to Cohen's (1988) conventions (refer to Table 2).

Bivariate regression

Parental bonding as a predictor of both borderline personality and depressive symptomatology

Hierarchical multiple regression was used to assess the ability of the four parental bonding variables (i.e., maternal care and overprotection, and paternal care and overprotection) to predict levels of borderline symptomatology, after controlling for age and depressive symptomatology. Depressive symptomatology and age were entered in step 1, explaining 24% of the variance in borderline symptomatology. After entry of the four parental bonding variables at step 2, the total variance explained by the model as a whole was 28%, F(6, 404) = 27.32, p < .001. The four parental bonding variables explained an additional 4.5% of the variance in borderline symptomatology, after controlling for depression and age, ΔR² = .045, ΔF (4, 404) = 6.42, p < .001. In the final model, only paternal care was a significant predictor of borderline symptomatology (refer to Table 3 for the values of the unique contributors of borderline symptomatology).

Table 3. Unique variance for the significant psychosocial antecedents of borderline and depressive symptomatology

	BPD symptomatology	Depressive symptomatology
	β	β
Maternal care		−.23**
Paternal care	−.15*
Disconnection/rejection domain
Abandonment	.33**
Mistrust/abuse	.13*
Defectiveness shame		.19*
Social isolation		.21**
Impaired autonomy domain
Failure		.21**
Enmeshment	−.12*
Impaired limits domain
Insufficient self‐control	.11*	.18**
Entitlement	.13*

*p < .05 level, two‐tailed; **p < .01 level, two‐tailed.

Another hierarchical multiple regression was used to assess the ability of the four parental bonding variables to predict levels of depressive symptomatology, after controlling for working status and borderline symptomatology. The total variance explained by the model as a whole was 25.5%, F(6, 404) = 23.02, p < .001. The four parental bonding variables explained an additional 5% of the variance in depressive symptomatology, after controlling for working status and borderline symptoms, ΔR² = .05, ΔF (4, 404) = 7.16, p < .001. In the final model, only maternal care was a significant predictor of depressive symptomatology (refer to Table 3 for the values of the unique contributors of depressive symptomatology).

EMS as a predictor of both borderline personality and depressive symptomatology

Hierarchical multiple regression was used to assess the ability of the 15 EMS to predict levels of borderline symptoms, after controlling for age and depressive symptomatology. Age and depressive symptomatology were entered in step 1, explaining 24% of the variance in borderline symptomatology. After entry of the 15 EMS at step 2, the total variance explained by the model as a whole was 43%, F (17, 393) = 17.56, p < .001. The 15 EMS explained an additional 19% of the variance in borderline symptomatology, after controlling for age and depression, ΔR² = .19, ΔF (15, 393) = 8.68, p < .001. In the final model, the following EMS were statistically significant: AB, MA, enmeshment, entitlement/grandiosity, and IS (refer to Table 3 for the values of the unique contributors of borderline symptomatology).

Hierarchical multiple regression was also used to assess the ability of 13 EMS to predict levels of depression symptoms, after controlling for working status and borderline symptomatology. Unrelenting standards and entitlement/grandiosity were removed from the model as both were found to be non‐significantly correlated with depressive symptomatology. Borderline symptomatology and working status was entered in step 1, explaining 20% of the variance in depressive symptomatology. After entry of the 13 EMS at step 2, the total variance explained by the model as a whole was 48%, F(15, 395) = 24.18, p < .001. The 13 EMS explained an additional 28% of the variance in depressive symptomatology, after controlling for borderline symptomatology and working status, ΔR² = .277, ΔF (13, 395) = 16.12, p < .001. In the final model, the following EMS were statistically significant: social isolation, DS, failure, and IS (refer to Table 3 for the values of the unique contributors of depressive symptomatology).

Comorbidity between borderline and depressive symptomatology

Raw counts relating to reported depressive and borderline symptomatology were entered for chi‐square analysis and are displayed in Table 4. Additionally, the number and percentages of participants falling under the different severity ratings for depression and borderline symptomatology can be seen in Table 4. A chi‐square test for independence (with Pearson chi‐square) indicated a significant difference between borderline severity and depression severity, χ² (4, n = 408) = 35.61, p < .001, phi = .30. Based on the results from the chi‐square analysis, respondents from the survey who were above the median score for borderline symptoms (responding positively to 3 or more SCID‐II‐PQ borderline items) reported a significantly higher frequency of above normal depressive symptomatology range compared with the respondents who did not score above the median for BPD (19.8% vs 44%; refer to Table 4).

Table 4. Comorbidity between borderline personality and depressive symptomatology

	Depressive symptomatology
	Normal	Mild	Moderate	Severe	Very severe
Borderline
Symptomatology
Below the median	163 (80.3%)	19 (9.4%)	17 (8.4%)	2 (1%)	2 (1%)
Above the median	115 (56.1%)	24 (11.7%)	36 (17.6%)	18 (8.8%)	12 (5.9%)

Note. Above the median refers to respondents scoring 3 or more SCID‐II‐PQ BPD items.

Discussion

Towards the broader goal of informing debate around distinctions between BPD and affective disorders, the current study had two aims: (1) determine the relative contributions of retrospectively reported parental bonding and current EMS to levels of both borderline and depressive symptomatology in a general population sample; and (2) determine levels of self‐reported comorbidity between borderline and depressive symptomatology. The hypothesis that higher levels of both borderline and depressive symptomatology would be associated with lesser parental care and greater parental overprotection was supported. Importantly, the pattern and strength of associations with these parental bonding variables were highly concordant between borderline and depressive symptomatology variables. Regression analysis provided partial further support for this consistency. The hypothesis that EMS of the disconnection/rejection domain would consistently, and most strongly, predict both borderline and depressive symptomatology was also supported by both correlational and regression analyses; further concordances for other EMS domains (insufficient control) were also observed. Finally, our findings revealed a significant level of comorbidity between borderline and depressive symptomatology.

Parental bonding as a predictor of both borderline personality and depressive symptomatology

The hypothesis that there would be a negative relationship between the parental care variables and both borderline and depressive symptomatology was supported. Furthermore, a positive relationship anticipated between parental overprotection and both borderline and depressive symptomatology was also supported. However, in the regression analysis, parental overprotection did not explain significant unique variance in either borderline or depressive symptomatology. Based on these findings, the current study could not provide support for parental overprotection as a significant unique contributor to the development of borderline or depressive symptomatology. Nevertheless, the findings from the current study do provide additional support for the past research findings of a consistent influence of low levels of parental care on the development of both BPD and depression (Ingram & Ritter, 2000; Neale et al., 1994; Nickell et al., 2002; Zweig‐Frank & Paris, 1991), strengthening the argument of a common developmental course for the two disorders. Our study improves on the methodological limitations of most past studies by including a large, gender‐balanced, non‐clinical sample, enabling confident extrapolation of findings to the general population, and with broader developmental implications. The finding that paternal care was the significant unique predictor for BPD symptomatology, whereas maternal care was the unique predictor for depressive symptomatology, was unexpected and requires further corroboration and investigation. A potential explanation for this unexpected finding is the confounding influence of childhood abuse, which was not measured in the current study. Given males are more likely to be perpetrators of childhood abuse, particularly sexual abuse (Peter, 2009), this variable may have influenced retrospective reports of paternal care for some participants. Further research is required to support this claim.

EMS associated with borderline personality and depressive symptomatology

The hypothesis that the EMS from the disconnection/rejection domain would yield the strongest correlations, and concordantly (with a similar mean magnitude) with borderline and depressive symptomatology, was supported. Additionally, EMS from other schema domains also yielded weaker significant associations, and also of a similar mean magnitude, with both borderline and depressive symptomatology. These included DI, VH, failure, IS, subjugation, and emotional inhibition. Of these associations, EMS of the disconnection/rejection domain predicted significant unique variance in both borderline and depressive symptomatology. This finding was consistent with previous studies (Halvorsen et al., 2009; Specht, Chapman, & Cellucci, 2009). Notwithstanding these consistencies, results indicated that different individual schemas from the disconnection/rejection domain predicted depressive (DS and social isolation) and borderline (AB and MA) symptoms, and may reflect disorder‐specific differences within the broader cognitive domain. Further phenomenological analysis and empirical investigation should corroborate and explore these variations, since they may importantly inform differential therapeutic approaches for the two disorders.

In addition to EMS from the disconnection/rejection domain, one schema from the impaired limits domain consistently predicted significant unique variance in both borderline and depressive symptomatology (IS). This finding supports past research which also found IS to be significantly associated with both BPD and depression (Butler et al., 2002; Jovev & Jackson, 2004).

Comorbidity between borderline and depressive symptomatology

The prediction of consistent comorbidity levels across the full dimensions of borderline and depressive symptomatology, in a nonclinical sample, was supported. Participants who reported highest levels of borderline symptomatology also reported depressive symptomatology well above the normal range, and there was a similar correspondence in symptom severity across other parts of the distributions. These findings support past research reporting frequent formal co‐diagnosis of BPD and depression (Zanarini et al., 1998) and extend this to the full distributions of these symptom variables in the general population.

Limitations

The use of self‐report instruments, though commonly utilised and considered reliable, may reduce the validity of some measurements. More particularly, the eliciting of retrospective reports permits two main limitations: (1) that the memories of past events could be unavailable as a result of lapses in memory or from difficulties in retrieving and accessing memories, and (2) that recollections of memories could also be biased by the influence of current experiences and events (Brigham et al., 2010). Furthermore, the yes/no format of the SCID‐II‐PQ is problematic for assessment validity because it fails to differentiate between infrequent symptoms versus enduring patterns of behaviour, which is the underlying criterion for personality disorders. Therefore, caution must be taken when interpreting the current findings given the absence of a follow‐up assessment, which may have resulted in artificial reports of inflated borderline traits by some participants. Additionally, as mentioned earlier, given there is discrepancy between the self‐report and clinician versions of the SCID, it is recommended that both versions be implemented in future clinical studies. Although the current study did not focus on a clinical sample, it is also important to note that relationships between depression and BPD in the research literature are often variable, and may depend on both depression indices measured (lifetime vs current diagnosis) as well as distinctive depression features, such as the number of lifetime episodes experienced, their severity, and the co‐occurrence of dysthymia. Future research focusing on depression and BPD must take these factors under consideration (Skodol et al., 1999).

With respect to sample representativeness, although the current study has obtained a balance of male and female participants more representative of the general population compared with some past studies, there were nevertheless substantially more females than males, even though it could also be argued that this gender balance more closely reflects the gender prevalence of both depression and BPD in community samples in any case. Additionally, more than half the sample were undergraduate students (55%), which is likely to skew the sample regarding age. However, similar to the gender imbalance, it could be argued that the sample reflects the age prevalence of both disorders. Nevertheless, these limitations should be addressed within the methodology of future research designs.

Conclusions

The current study investigated potential shared developmental influences for BPD and depression, operationalized dimensionally in a large, general population sample. Concordant unique contributors to the development of both borderline and depressive symptomatology included level of parental care, EMS from the disconnection/rejection domain (DS, social isolation, AB and MA), as well as one EMS from different schema domains (IS). The results also revealed a high level of concordance in severity levels for borderline and depressive symptomatology across the entire distributions for these variables.

Our data therefore support the contention of, at least partially, common developmental influences for BPD and depression, and high dimensional comorbidity, with important implications for the management of these disorders and early preventative intervention. Future research should examine additional psychosocial influences potentially common to BPD and depression, as well as other affective disorders, such as bipolar. This would allow researchers and clinicians to better identify individuals at risk of these debilitating disorders. Given the similar developmental trajectories of the two disorders, it is also suggested that future inoculation programmes target the EMS found to be predictors for both the development of BPD and the affective disorders.