Differentiating between intrapsychic symptoms and behavioral expressions of borderline personality disorder in relation to childhood emotional maltreatment and emotion dysregulation: an exploratory investigation

ABSTRACT

Background:

Borderline personality disorder (BPD) is a severe mental disorder, characterized by pronounced instability in emotions, self-image, and interpersonal relationships. Experiences of childhood maltreatment are among the risk factors for BPD. While self-damaging and aggressive acts often occur, not every person with the disorder shows markedly dysregulated behaviour. Internalized symptoms, such as shame, loneliness, and self-disgust tend to be more pervasive and persist after clinical remission.

Objective:

Here we investigated associations between BPD symptom severity, childhood maltreatment, and emotion regulation difficulties. We further explored if the Borderline Symptom List (BSL) could potentially be used to differentiate between internalized symptoms (intrapsychic strain) and externalized symptoms (dysregulated behaviours) in future research.

Method:

187 women with at least mild BPD symptoms (65% having a diagnosis of BPD) completed the BSL 23 including its 11-item supplement (BSL-S), the Childhood Trauma Questionnaire (CTQ), and Difficulties in Emotion Regulation Scale (DERS). Participants further underwent a semi-structured clinical interview to assess BPD criteria (International Personality Disorder Examination, IPDE). Multivariate models and regression-based bootstrapping analyses were performed to test direct and indirect effects.

Results:

Childhood trauma severity, especially emotional abuse, positively predicted BPD symptom severity. A significant indirect effect through emotion regulation difficulties was found (k²=.56). When exploring associations with BPD criteria (IPDE), the BSL-23 mean significantly correlated with separation anxiety, identity and mood problems, chronic emptiness, suicidal ideation, and dissociation, while the BSL-S correlated with self-harming impulsive behaviour and anger outbursts.

Conclusions:

Findings complement previous research, highlighting the role of childhood maltreatment and emotion regulation difficulties in BPD. While our findings need to be seen as preliminary and interpreted with caution, they suggest that the BSL may be used to differentiate between internalized symptoms and behavioural expressions of BPD in future research. Such a distinction might help to deepen the understanding of this complex heterogenous disorder.

Highlights

• Severity of borderline personality disorder was related to experiences of childhood maltreatment, mediated by emotion regulation difficulties.

• The Borderline Symptom List might be useful to differentiate between internalizing symptoms and externalizing symptoms associated with borderline personality disorder.

Antecedentes: El trastorno límite de la personalidad (TLP) es un trastorno mental grave, caracterizado por una marcada inestabilidad en las emociones, la autoimagen y las relaciones interpersonales. Las experiencias de maltrato en la infancia se encuentran entre los factores de riesgo del TLP. Aunque a menudo se producen actos autolesivos y agresivos, no todas las personas con el trastorno muestran un comportamiento marcadamente desregulado. Los síntomas internalizantes, como la vergüenza, la soledad y el autodesprecio, tienden a ser más generalizados y persisten tras la remisión clínica.

Objetivo: Investigamos las asociaciones entre la severidad de los síntomas del TLP, el maltrato infantil y las dificultades en la regulación emocional. Además, exploramos si la Lista de Síntomas Límite (BSL por sus siglas en ingles) podría ser potencialmente utilizada para diferenciar entre síntomas internalizantes (tensión intrapsíquica) y síntomas externalizantes (comportamientos desregulados) en futuras investigaciones.

Método: 187 mujeres con al menos síntomas leves de TLP (65% con diagnóstico de TLP) completaron el BSL 23, incluido su suplemento de 11 ítems (BSL-S), el Cuestionario de Trauma Infantil (CTQ) y la Escala de Dificultades en la Regulación de las Emociones (DERS). Además, los participantes se sometieron a una entrevista clínica semiestructurada para evaluar los criterios de TLP (Examen Internacional de Trastornos de la Personalidad, IPDE). Se realizaron modelos multivariantes y análisis bootstrapping basados en regresión para comprobar los efectos directos e indirectos.

Resultados: La gravedad del trauma infantil, especialmente el abuso emocional, predijo positivamente la gravedad de los síntomas del TLP. Se encontró un efecto indirecto significativo a través de las dificultades de la regulación emocional (k² = .56). Al explorar las asociaciones con los criterios de TLP (IPDE), la media del BSL-23 se correlacionó significativamente con la ansiedad por separación, los problemas de identidad y del estado de ánimo, el vacío crónico, la ideación suicida y la disociación, mientras que el BSL-S se correlacionó con la conducta impulsiva autolesiva y los arrebatos de ira.

Conclusiones: Los hallazgos complementan la investigación previa, destacando el papel del maltrato infantil y las dificultades de regulación emocional en el TLP. Aunque nuestros hallazgos deben considerarse preliminares e interpretarse con cautela, sugieren que el BSL puede utilizarse para diferenciar entre los síntomas internalizantes y las expresiones conductuales del TLP en investigaciones futuras. Esta distinción podría ayudar a profundizar en la comprensión de este complejo y heterogéneo trastorno.

背景：边缘性人格障碍（BPD）是一种严重的精神障碍，其特征是情绪、自我形象和人际关系的明显不稳定。 童年期受虐待经历是边缘性人格障碍的风险因素之一。虽然自我伤害和攻击性行为经常发生，但并不是每个患有这种疾病的人都表现出明显的失调行为。羞耻、孤独和自我厌恶等内化症状往往更加普遍，并在临床缓解后持续存在。

目的：在这里，我们考查了 BPD 症状严重程度、童年期虐待和情绪调节困难之间的关联。 我们进一步探讨了边缘症状列表（BSL）是否可以在未来的研究中用于区分内化症状（内心紧张）和外在症状（失调行为）。

方法：187 名至少有轻度 BPD 症状的女性（65% 诊断为 BPD）完成了 BSL 23，包括其 11 条目补充 (BSL-S)、儿童创伤问卷 (CTQ) 和情绪调节困难量表 (DERS）。 参与者进一步接受了半结构化临床访谈，以评估 BPD 标准（国际人格障碍检查，IPDE）。进行多变量模型和基于回归的重新抽样分析来检验直接和间接效应。

结果：童年创伤的严重程度，尤其是情感虐待，可以正向预测 BPD 症状的严重程度。发现情绪调节困难有显著的间接效应（k² = .56）。在探索与 BPD 标准 (IPDE) 的关联时，BSL-23 平均值与分离焦虑、身份和情绪问题、慢性空虚、自杀意念和解离显著相关，而 BSL-S 与自残冲动行为和愤怒爆发相关 。

结论：研究结果补充了之前的研究，强调了童年虐待和情绪调节困难在边缘性人格障碍中的作用。虽然我们的研究结果需要被视为初步的并谨慎解释，表明在未来的研究中，BSL 可用于区分 BPD 的内化症状和行为表达。 这种区别可能有助于加深对这种复杂的异质性疾病的理解。

KEYWORDS:

• Borderline personality disorer
• childhood trauma
• emotion regulation
• trauma
• borderline symptom list

PALABRAS CLAVE:

• Maltrato emocional infantil
• trauma complejo
• trastorno límite de la personalidad

关键词:

• 童年情感虐待
• 复杂性创伤
• 边缘性人格障碍

1. Introduction

Borderline personality disorder (BPD) is a severe mental disorder, characterized by instability in affect, self-image, and interpersonal relationships, with a lifetime prevalence of 1–3% (Bohus et al., 2021). BPD usually emerges in adolescence, reaches its peak in early adulthood and then declines over time (Winsper, 2021). While around 60% of people reach symptomatic remission (Álvarez-Tomás et al., 2019; Ng et al., 2016), full recovery is rare and impairments in the vocational and psychosocial domain are still present (Zanarini et al., 2018).

Problems in emotion regulation are a core feature of BPD (Bohus et al., 2021) and often persist even after successful treatment (Gunderson et al., 2018; Zanarini et al., 2016). Individuals with BPD usually show heightened emotional sensitivity and reactivity (Fossati et al., 2016; Glenn & Klonsky, 2009; Hope & Chapman, 2019; Iverson et al., 2012; Salsman & Linehan, 2012; Sinclair & Feigenbaum, 2012) and report more problems in emotional domains, such as a lack of clarity and awareness or acceptance of emotions (Barker et al., 2015; Daros & Williams, 2019; Derks et al., 2017; Krause-Utz, Erol, et al., 2019; McHugh & Balaratnasingam, 2018; Turner et al., 2017). At the same time, there is growing evidence that emotion dysregulation in BPD is a contextual problem, strongly influenced by social interactions (Fitzpatrick et al., 2023). Main problems seem to be an inflexible selection and implementation of emotion regulation strategies, especially in interpersonal contexts.

Emotion regulation develops in close interactions with primary caregivers (Barthel et al., 2018). Developmental models of BPD, such as the biosocial theory (Linehan, 1993), highlight the role of stressful interpersonal experiences (e.g. emotional, sexual and physical abuse and neglect), which interact with genetic neurobiological vulnerabilities. An emotionally invalidating environment (e.g. constant devaluation, unreliable communication patterns, and inadequate coregulation of emotions by primary caregivers) may result in a pervasive pattern of emotional dysregulation, which contributes to maladaptive coping behaviours (Crowell et al., 2009; Fonagy et al., 2017; Hughes et al., 2012; Luyten et al., 2020).

According to a meta-analysis of case–control studies, people with BPD are 13.9 times more likely to report adverse childhood experiences than healthy controls and 3.2 times more likely than people with other psychiatric disorders; emotional maltreatment (abuse and neglect) had the strongest relation to BPD psychopathology (Porter et al., 2020). When controlling for other types of maltreatment, emotional abuse was the only form that consistently predicted BPD symptoms in previous studies (Bornovalova et al., 2006; Fossati et al., 2016; Frias et al., 2016; Hengartner et al., 2013; Hernandez et al., 2012; Kuo et al., 2015; Lobbestael et al., 2010; Xie et al., 2021). In a recent network analysis, emotional abuse was the most central node and a bridge symptom between other types of childhood maltreatment and BPD features (Schulze et al., 2022).

Numerous studies suggest that childhood emotional maltreatment is associated with more emotion dysregulation, which in turn is predictive of more severe BPD symptoms (Carvalho Fernando et al., 2014; Gratz et al., 2008; Peng et al., 2021; van Dijke et al., 2013, 2018). Such associations were found in a wide variety of samples, including university students (Kuo et al., 2015), a community sample (Fossati et al., 2016), psychiatric outpatients (Rosenstein et al., 2018; Schaich et al., 2021) and inpatients (Bertele et al. 2022), and substance abusers (Gratz et al., 2008). Two studies found a prominent role of impulse control difficulties (Oshri et al., 2015; Schaich et al., 2021) and limited access to regulation strategies (Guérin-Marion et al., 2020).

In this study, we examined associations between childhood maltreatment experiences, emotion regulation difficulties, and BPD symptom severity. Based on the afore-mentioned theoretical models and empirical evidence, we expected childhood maltreatment, especially emotional abuse and neglect, to positively predict BPD symptoms. We further hypothesized that emotion regulation difficulties would statistically mediate the link between childhood emotional maltreatment and BPD symptom severity.

In this context, we further explored if the Borderline Symptom List (BSL) (Bohus et al., 2009) could potentially be used to differentiate between internalized symptoms (internal expressions of distress or intrapsychic strain), and externalized symptoms of BPD (e.g. risky impulsive behaviours and angry outbursts). The BSL is a well-established self-report scale that assesses a wide range of BPD symptoms (e.g. low self-worth, guilt and shame, self-defeat, dysphoria, loneliness, intrusions, dissociation, and mistrust). A supplement of 11 items was added to specifically assess behavioural expressions, such as non-suicidal self-injury, suicidal behaviour, and angry outbursts.

BPD co-occurs with symptoms of both internalizing (e.g. high levels of depression, shame, loneliness, and self-disgust, a profound negative self-image) and externalizing (dysregulated behaviours, which are directed outwardly), even when psychiatric comorbidity is accounted for, (Bohus et al., 2021; Sharp et al., 2018; Wertz et al., 2020). Both have negative outcomes for those affected, their close ones, and society at large (Bailey & Grenyer, 2013; El-Gabalawy et al., 2010; Quirk et al., 2015; van Asselt et al., 2007; Wertz et al., 2020). It is important, however, to distinguish between internalizing and externalizing symptoms, because there is evidence that internalized symptoms are more persistent than behavioural problems (Gunderson et al., 2018; Zanarini et al., 2016) and critically interfere with psychosocial functioning. Moreover, many individuals with BPD, especially older ones, experience high intra-psychic strain, even in the absence of markedly dysregulated behaviours (Beatson et al., 2016; Bohus et al., 2021; Morgan et al., 2013; Winsper, 2021). Persons with this ‘quieter, hidden’ variant of BPD may attract less professional attention and may run the risk of being misdiagnosed or not getting the appropriate care (Bohus et al., 2021).

Differentiating between symptoms of internalizing and externalizing might contribute to a more nuanced understanding of the heterogeneity of BPD, especially in light of the current shift towards a more dimensional and global personality disorder assessment (Bach & Fjeldsted, 2017). Over the past decade, there have been fundamental changes in the classification of personality disorders. The Alternative Model of Personality Disorders (AMPD) of the DSM (APA, 2013) involves a three-step diagnostic process. After determining the overall severity of the patient’s general level of personality function, the presence and severity of the following BPD trait domains is determined: emotional lability, anxiousness, separation insecurity, depression, impulsivity, risk-taking, and hostility (Krueger & Hobbs, 2020; Widiger & McCabe, 2020). In the ICD-11 (WHO, 2018), a global personality disorder dimension is assessed, which can be specified as a borderline pattern based on the severity of five trait domains: Negative Affectivity, Detachment, Dissociality, Disinhibition, Anankastia (rigidity) (Gutiérrez et al., 2023; Mulder et al., 2020; Sharp et al., 2015; Williams et al., 2018; Wright et al., 2016).

Here, we examined if the BSL and its supplement (BSL-S) are associated with internalized symptoms versus behavioural expressions of BPD, assessed with structured clinical interviews. We further explored how this may be linked to childhood maltreatment severity and emotion regulation difficulties.

2. Methods

2.1. Participants and procedure

Data collection took place at the Central Institute of Mental Healthy in Mannheim, Germany. All procedures were approved by the local Ethics Committee. Participants were recruited through flyers, websites, and internet platforms. Before enrolment in the studies, interested participants were screened to determine their eligibility. Inclusion criteria for the present sample were being female and at least 18 years old. For our research, only female participants were included to make the sample more homogenous and to control for a possibly confounding effect of gender. To have sufficient variance in the outcome, we included participants that showed at least mild BPD-symptoms, defined as a score greater than or equal to 0.30 on the BSL-23 (see Kleindienst et al., 2020). Most participants (65%) were patients with a verified diagnosis of BPD (see below). Participants with severe somatic illness, a diagnosis of substance abuse within the last six months, lifetime history of bipolar-I disorder, psychotic disorder, current suicidal crisis, mental deficiency, and developmental disorder were excluded. Eligible participants were informed about the aims, background, and potential risks of the study, provided informed written consent, and were informed that they could terminate their participation at any time. Then participants underwent a comprehensive interview by trained clinicians. This interview contained the International Personality Disorder Examination for personality disorders (Loranger, 1999) (IPDE, see below) and the Structured Interview for DSM-IV Axis 1 Disorders (SCID-I) (First et al., 1997). Additionally, information on prior treatment history was obtained. After the interview, participants were asked to complete self-report questionnaires in paper-and-pencil form. Participants received 12 euros per hour as reimbursement. Part of the data was analyzed in the context of another study (Bertele et al., 2022). The current study differed from this study by using a different dimensional outcome for BPD psychopathology and by exploring differential associations with DSM-IV criteria.

Power analysis: An a-priori sample size calculation was performed using G*Power (version 3.1.9.7) and Monte Carlo simulations (Schoemann et al., 2017). For the Monte Carlo simulations (Schoemann et al., 2017), the number of replications was set at 5,000, the draws per replication at 20,000, the random seed at 1,234 and the confidence level at 95%. Overall, 85 participants were required to detect a moderate direct effect of childhood emotional maltreat on both outcomes (internalized symptoms and dysregulated behaviours), as well as indirect effects through emotion dysregulation.

Sample characteristics: The final sample consisted of 187 women, aged between 18 and 52 (M = 29.84, SD = 8.21), mostly single (n = 152, 81%). About half of the sample (n = 102, 54%) had higher education (primary education: n = 73, apprenticeship: n = 56, high-school: n = 45, other: n = 13). Full sample characteristics can be found in Table 1. One hundred and twenty-one (65%) participants received a diagnosis of BPD according to DSM-IV. Twenty-six participants (14%) did not meet criteria for a mental disorder. The rest met criteria for a mental disorder other than BPD (anxiety disorders). The majority (n = 143, 76%) was currently in treatment and reported previous outpatient treatment (mean number of treatments: 11.94 ± 11.75) and inpatient treatment (mean number: 1.0 ± 0.92).

Table 1. Sample characteristics.

Age (years)	29.84 ± 8.21, [18–52]
Education
Primary school	n = 73 (39%)
Vocational training	n = 56 (30%)
Highschool	n = 45 (24%)
Other	n = 13 (6%)
Relationship Status
Single	n = 152 (81%)
Married	n = 21 (11%)
Divorced	n = 7 (4%)
Separated	n = 1 (0.5%)
Mean number of outpatient treatments	11.94 ± 11.75, [0–60]
Mean number of inpatient treatments	1.0 ± 0.92, [0–5]
Currently in psychiatric treatment	n = 143 (76%)
Severe childhood trauma^ª
Emotional abuse	n = 131 (70%)
Emotional neglect	n = 54 (29%)
Physical abuse	n = 86 (46%)
Physical neglect	n = 124 (66%)
Sexual abuse	n = 94 (50%)
Severity of BPD symptoms
BSL-23 mean	1.63 ± 0.86 [0.30–3.48]
BSL-S mean	0.23 ± 0.29 [0–1.27]
Mean number of DSM criteria for BPD met	3.94 ± 3.10, [0–9]
Current disorders
BPD	n = 121 (65%)
Dysthymia	n = 22 (12%)
Substance misuse	n = 8 (4%)
Obsessive compulsive disorder	n = 7 (4%)
Panic disorder	n = 17 (9%)
Social phobia	n = 36 (19%)
Posttraumatic stress disorder	n = 65 (35%)

Note: The table shows means ± standard deviation and range or frequencies. BPD = Borderline Personality Disorder; ªaccording to cut-offs by Bernstein et al. (2003).

According to recent classifications (Kleindienst et al., 2020), average BPD severity in our sample was moderate, while there was a broad range of symptoms, from none or low to extremely high severity (see below). Majority of participants (n = 131, 70%) reported severe maltreatment, mostly emotional abuse.

2.2. Measures

Borderline Symptom List (Bohus et al., 2009). The Borderline Symptom List 23 (BSL-23) was used to measure BPD symptoms. It includes 23 items that are answered on a 5-point Likert-scale ranging from 0 (not at all) to 4 (very strong). The BSL supplement (BSL) has 11-items that assess behaviours within the past week: non-suicidal self-injury, suicidal attempts, suicidal communication, angry outbursts, disordered eating, risky sexual behaviour, misuse of substances (medication, drugs, alcohol) on a 5-point Likert-scale ranging from 0 (not at all) to 4 (daily). The psychometric properties of the BSL-23 are well-established (Bohus et al., 2001, 2007). Its convergent validity is supported by strong positive correlations with measures of depression and general psychopathology as well as a high negative correlation with measures of global well-being. It previously showed high validity with clinical ratings (Bohus et al., 2001, 2007). Test-retest reliability of .82, internal consistency of α =  .97, and sensitivity towards change of BPD-typical impairment were reported. The BSL-23 is able to discriminate between patients with BPD and patients with other diagnoses (Bohus et al., 2009). In the present sample, Cronbach's alpha for the BSL-23 was α = .94.

A mean score was defined with a theoretical range from 0 to 4, with higher scores reflecting more severe symptoms. Recently, Kleindienst and colleagues proposed a classification of severity levels for the BSL-23 (Kleindienst et al., 2020). Based on data from 1.090 adults and the BSL-23 distribution of 241 individuals with a diagnosis of BPD, severity levels were validated using comparisons with established instruments on psychosocial functioning (e.g. the Global Assessment of Functioning, GAF). Six levels of symptom severity were defined for the BSL-23 mean score: none or low: [0–0.28); mild: [0.28–1.07); moderate: [1.07–1.87); high: [1.87–2.67); very high: [2.67–3.47); and extremely high: [3.47–4]. The BSL-23 score that best distinguished treatment-seeking BPD patients from clinical controls was 1.50, whereas the clearest discrimination of BPD patients and healthy controls was found at a score of 0.64. According to this classification, our sample overall showed moderate severity (mean: 1.63 ± SD: 0.86), while BSL-23 scores ranged from mild to extremely high ([0.30–3.48], see Table 1). For the BSL-S separately, scores were overall low to mild (0.23 ± 0.29 [0–1.27]) and had a restricted range.

International Personality Disorder Examination (IPDE) (Loranger, 1999). The IPDE is a semi-structured interview, which assesses the presence of BPD symptoms and the nine DSM-IV criteria. Criteria are assessed for their presence (‘negative’, ‘probable’, ‘definite’, ‘not available/unclear’). The IPDE previously showed very good internal validity and prognostic properties (Loranger, 1999). Cronbach's alpha for the IPDE was α = .96. Inter-rater reliability was estimated to be κ = 0.90.

2.2.1. Childhood trauma questionnaire short form (CTQ-SF; Bernstein et al., 2003).

Childhood maltreatment was measured with the subscales of the Childhood Trauma Questionnaire – Short Form (CTQ-SF; Bernstein et al., 2003). This 28-item self-report measure assesses physical abuse, emotional abuse, sexual abuse, physical neglect, and emotional neglect with five items each. All items are measured on a 5-point Likert scale ranging from 1 (never true) to 5 (very often true). The CTQ-SF showed good psychometric properties across different samples, also in its German version (Wingenfeld et al., 2010). Its convergent validity is supported by comparisons with therapist ratings. The internal consistencies (Cronbach’s α) vary between .84 and .89 for emotional abuse and between .85 and .91 for emotional neglect (Bernstein et al., 2003). In the current study, we observed Cronbach’s α = .88 for emotional abuse, Cronbach’s α = .94 for emotional neglect, and Cronbach’s α = .94 for the combination of the scales, which was defined by taking the mean of the means on both scales.

Difficulties in Emotion Regulation Scale (DERS; Gratz & Roemer, 2004).

Emotion regulation difficulties were assessed with the Difficulties in Emotion Regulation Scale (DERS; Gratz & Roemer, 2004). The DERS is a 36-item self-report measure on difficulties in emotion regulation. It consists of the six subscales: emotional awareness, emotional clarity, acceptance of one’s emotions, access to emotion regulation strategies perceived as effective, impulse control when experiencing negative affect, and the ability to perform goal-directed behaviour when experiencing negative affect (Gratz & Roemer, 2004). Items are measured on a 5-point Likert scale ranging from 1 (almost never) to 5 (almost always). The DERS has good psychometric properties. It showed good convergent validity with other measures of emotion regulation and good internal consistency, with Cronbach’s α = .93 for the total DERS-score, and α = .80 to .89 for the subscales (Gratz & Roemer, 2004). In the current study, Cronbach’s α was .95 for the total DERS-score and ranged from .82 to .91 for the subscales. Mean scores for the total DERS as well as for the individual subscales were calculated, with a theoretical range from 1 to 5. With the present scoring, higher scores reflect greater difficulties in emotion regulation.

2.3. Statistical analyses

Statistical analyses were performed with IBM SPSS Statistics (version 28). From the full sample (n = 187) that completed the CTQ, DERS, and BSL-23, data for the BSL supplement was missing in n = 19 participants, resulting in a subsample of n = 168 for this measure. Preliminary analyses were conducted to test the distribution of variables and their bivariate intercorrelations (see Table 2). These preliminary analyses supported our conceptual mediation model. Multicollinearity was examined by VIF and tolerance values (Hair et al., 2010). Linearity was checked with residuals plots and scatterplots, which revealed no signs of deviations from linearity. Due to deviations from homoscedasticity, the heteroscedasticity corrected inference H4 was used in the following analyses (Hayes & Cai, 2007). There were indications for a non-normal distribution of the residuals. Where possible, we applied a non-parametric procedure, which is robust against these violations. Some outliers were present, but Cook’s d values indicated no influential cases. Therefore, these outliers were not excluded to avoid data and information loss. Demographics, such as age, were not significantly correlated with the primary outcome (p > .05).

Table 2. Descriptive statistics and spearman correlations.

Variable	1	2	3	4	5	6	7^a	8	9	Mean	SD
1. Age		.33***	.31***	.32***	.27***	.30***	.03	−.03	−.10	29.84	8.21
2. CTQ EA			.73***	.38***	.58***	.53***	.35***	.25***	.12	3.23	1.17
3. CTQ EN				.37***	.47***	.69***	.20**	.22***	.02	3.29	1.18
4. CTQ SA					.56***	.49***	.06	.14	−.04	9.95	6.41
5. CTQ PA						.50***	.14	.07	.04	8.74	4.89
6. CTQ PN							.03	.18*	.09	9.89	3.88
7. DERS – total								.66***	.39***	18.53	3.36
8.. BSL-23									.46***	1.63	0.86
9. BSL-Suppl										0.23	0.29
										0.09^b	0.34^b

Note: CTQ = Childhood Trauma Questionnaire, EA = Emotional Abuse; EN = Emotional Neglect, SA = Sexual Abuse, PA = Physical Abuse; PN = Physical Neglect DERS = Difficulties in Emotion Regulation Scale, BSL-23 = Borderline Symptom List.

^a As the scores on the BSL-supplement were not normally distributed, Spearman’s ρ was used to determine the correlation between the BSL-supplement and other variables.

^b As the scores on the BSL-supplement were not normally distributed, the median and IQR are reported in addition to the mean and standard deviation.

*p < .05, **p < .01, ***p < .001.

To test if childhood maltreatment severity predicted BPD symptom severity, a multivariate linear model with the five CTQ subscales (emotional, physical, and sexual abuse, emotional and physical neglect) as predictors and the BSL-23 mean as dependent variable was performed. We additionally ran an analysis where the CTQ subscales for emotional abuse and neglect as well as physical abuse and neglect were aggregated to avoid some potential multicollinearity issues.

To test direct and indirect effects, non-parametric regression-based bootstrapping was performed using the SPSS Macro Process (version 4.0; Hayes, 2017; Preacher & Hayes, 2004). In a simple mediation model (Process model number 4), the mean of the DERS total score was entered as mediator variable. The mean score of the CTQ subscales emotional abuse and neglect was used as predictor. The mean score of the mean BSL-23 was the outcome variable. Additional analyses were run to test if results remained stable when controlling for the other CTQ subscales (sexual abuse, physical abuse and neglect) as covariates. To evaluate significant indirect effects, k² was computed manually, as this is the recommended effect size for mediation, representing the current indirect effect as a proportion of the maximum possible indirect effects (Preacher & Kelley, 2011). Since the calculated four tests were dependent, no additional correction for multiple testing was applied. For all analyses, a 95% confidence interval and 5,000 bootstrap samples were used.

In the afore-mentioned analyses, we focused on the mean BSL-23 as main outcome variable, since the BSL-S has not been validated for separate use. To explore if the BSL-23 and BSL-S may differentiate between intrapsychic symptoms and behavioural expressions of BPD respectively, a repeated measure analysis of covariance was used. The BSL-23 mean and the BSL-S mean were included as predictors and the nine lifetime criteria for BPD (assessed with the IPDE) as dependent variables.

3. Results

3.1. Associations between childhood maltreatment (CTQ subscales) and BPD symptom severity (BSL-23)

Overall, childhood maltreatment severity significantly positively predicted BPD symptom severity (BSL-23), F_(5,171)= 3.59, p = .004, η² = 0.10, _adjR² = 0.07. Among the five CTQ subscales, emotional abuse had a unique effect. More severe emotional abuse predicted more severe BPD symptoms (see Table 3). Similar results were found for emotional maltreatment, when collapsing emotional abuse and neglect as well as physical abuse and neglect subscales (see Supplemental Table 1).

Table 3. Results of the multivariate linear model with CTQ subscales as predictors.

Dependent Variable	Predictor	F(1.171)	B	Std.-Error	T	Sig.	95%CI Lower Bound	95%CI Upper Bound	Partial Eta²
BPD symptom severity	Emotional abuse	5.17	.296	.130	2.27	.024	.039	.553	.029
	Emotional neglect	0.02	−.005	.034	0.14	.886	−.072	.062	.000
	Physical abuse	3.00	−.027	.016	1.73	.085	−.059	.004	.017
	Physical neglect	0.28	−.012	.022	0.53	.595	−.056	.032	.002
	Sexual abuse	0.64	.009	.012	0.80	.426	−.014	.032	.004

Note: BPD = Borderline Personality Disorder, CTQ = Childhood Trauma Questionnaire; CI = Confidence Interval.

3.2. Regression-based bootstrapping mediation analysis

The analysis revealed a significant total effect of childhood emotional maltreatment (abuse and neglect) on BPD symptoms (BSL 23 mean: c: B = 0.20, SE = 0.06, t = 3.69, p < .001, 95% CI [0.10, 0.31]). When including the mediator variable in the model, this effect was not significant anymore (B = 0.04, SE = .05, t = 0.87, p = .384, 95% CI: [−0.05, 0.13]). There was a significant indirect effect of childhood emotional maltreatment via emotion regulation difficulties on BPD symptoms (B = 0.16, SE = 0.04, 95% CI [0.09, 0.25]), suggesting mediation. An effect size of k² = .56 was found, corresponding to a large effect. As shown in Figure 1A, more severe emotional maltreatment predicted more difficulties in emotion regulation (B = 0.23, SE = .05, t = 4.85, p < .0001, 95% CI: [0.14, 0.32]), which in turn predicted more BPD symptoms (B = 0.71, SE = .07, t = 10.43, p < .0001, 95% CI: [0.58, 0.85]). Results remained stable when controlling for the other CTQ subscales (sexual abuse, physical abuse and neglect) as covariates (see Supplemental Material).

Figure 1. Results of the mediation analysis.

3.3. Exploratory analysis: Associations between BSL 23 and BSL-S with BPD criteria (IPDE)

As shown in Table 4, higher scores on the BSL-23 mean were correlated to fears of abandonment, instable relationships, instable identity, suicidal ideation, mood problems, chronic emptiness, and dissociation. The BSL-S in turn was associated with self-damaging impulsive behaviours and anger outbursts. The BSL-23 and BSL-S were significantly correlated. Correlations with the DERS and CTQ subscales suggest that both BSL-23 and BSL-S were correlated to difficulties in emotion regulation, while reports of emotional childhood maltreatment significantly correlated with BSL-23 (Table 2).

Table 4. Results of the multivariate linear model with IPDE criteria as predictors.

Dependent Variable	Predictor	Regression coefficient B	Std.-Error	T	Sig.	95%CI Lower Bound	95%CI Upper Bound	Decentra-lized parameter	Observed discrimination strength
Fears of abandonment
	BSL-23 mean	.231	.094	2.46	.015	.045	.417	2.456	.685
	BSL Supplement	.291	.278	1.05	.297	−.259	.841	1.046	.180
Unstable. intense interpersonal relationships
	BSL-23 mean	.282	.085	3.31	.001	.114	.450	3.318	.909
	BSL Supplement	.116	.251	0.46	.646	−.381	.613	.460	.074
Identity disturbance
	BSL-23 mean	.313	.091	3.45	.001	.134	.493	3.452	.929
	BSL Supplement	.050	.269	.186	.852	−.481	.581	.186	.054
Impulsivity in at least two potentially damaging areas
	BSL-23 mean	.060	.084	0.71	.477	−.107	.227	.712	.109
	BSL Supplement	.647	.250	2.59	.010	.154	1.140	2.591	.731
Suicidal ideation and gestures
	BSL-23 mean	.232	.078	2.95	.004	.077	.387	2.953	.835
	BSL Supplement	.059	.232	0.25	.799	−.399	.518	.256	.057
Affective instability
	BSL-23 mean	.247	.075	3.29	.001	.099	.395	3.298	.906
	BSL Supplement	−.011	.221	−0.05	.959	−.449	.426	.051	.050
Chronic feelings of emptiness
	BSL-23 mean	.235	.085	2.75	.007	.067	.403	2.758	.782
	BSL Supplement	.249	.252	0.98	.325	−.249	.747	.988	.166
Inappropriate. intense anger, outbursts
	BSL-23 mean	.169	.092	1.84	.067	−.012	.351	1.845	.450
	BSL Supplement	.679	.272	2.50	.013	.143	1.216	2.501	.700
Paranoid ideation or dissociative symptoms
	BSL-23 mean	.268	.085	3.16	.002	.101	.435	3.163	.882
	BSL Supplement	−.270	.250	−1.07	.282	−.765	.225	1.079	.188

4. Discussion

The aim of this study was to investigate associations between self-reported childhood maltreatment experiences, emotion regulation difficulties, and BPD symptoms. We further explored if the Borderline Symptom List (BSL-23) may potentially be used to differentiate between internalized symptoms (intrapsychic strain) and externalized symptoms (dysregulated behaviours) of BPD in future research.

Experiences of childhood maltreatment, especially emotional abuse, positively predicted emotion regulation difficulties and BPD symptom severity. These results are in line with our hypothesis and previous findings. Numerous studies have established links between childhood emotional maltreatment, emotional dysregulation, and BPD psychopathology (Carvalho Fernando et al., 2014; Fossati et al., 2016; Gratz et al., 2008; Kuo et al., 2015; Peng et al., 2021; Rosenstein et al., 2018; Schaich et al., 2021; van Dijke et al., 2013, 2018) or BPD diagnosis (Bertele et al., 2022).

Consistent with the biosocial theory (Linehan, 1993) and previous evidence, these findings suggest that early experiences with primary caregivers play an important role in emotion regulation processes. Adaptive and flexible emotion regulation develops in close social interactions and involves complex affective–cognitive processes, such as social modelling and learning (Hughes et al., 2012; Thompson, 2011). Experiences of emotional abuse (e.g. devaluation) can shape the development of emotion dysregulation. Over time, vulnerable individuals learn to express and regulate their emotions in more maladaptive ways (Crowell et al., 2009). Importantly, these associations depend on neurobiological, genetic vulnerabilities, which have not been assessed in our study (Bornovalova et al., 2013). Moreover, it is also important to note that these processes are bidirectional and dynamic. From a transactional perspective, caregivers encounter more difficulties validating and co-regulating the emotional expressions of a child if expressions are too intense and if they lack support (Barthel et al., 2018).

Both childhood maltreatment experiences and emotion regulation difficulties were positively related to BPD symptom severity in our study. In line with earlier research, emotion regulation difficulties mediated the relationship between childhood emotional maltreatment and BPD psychopathology (Carvalho Fernando et al., 2014; Gratz et al., 2008; Peng et al., 2021; van Dijke et al., 2013, 2018). Similar indirect effects were found in a wide variety of samples, including university students (Kuo et al., 2015), a community sample (Fossati et al., 2016), psychiatric outpatients (Rosenstein et al., 2018; Schaich et al., 2021), inpatients (Bertele et al. 2022), and substance abusers (Gratz et al., 2008). Like most of these studies, we conceptualized emotion regulation as a multi-dimensional construct based on the model by Gratz and Roemer (2004). This model conceptualizes six components of emotion regulation: emotional awareness, emotional clarity, acceptance of one’s emotions, access to emotion regulation strategies perceived as effective, impulse control when experiencing negative affect, and the ability to perform goal-directed behaviour when experiencing negative affect (Gratz & Roemer, 2004). While people with BPD consistently reported more difficulties in these domains, experimental and psychophysiological research or studies using ecological momentary paradigms have yielded mixed results (Fitzpatrick et al., 2023). While persons with BPD tend to use more maladaptive emotion regulation strategies in daily life (e.g. more suppression, rumination, and avoidance and less cognitive reappraisal, problem solving, and acceptance), the main problem seems to be an inflexible implementation of emotion regulation strategies rather than the lack of it (Daros & Williams, 2019). Neuroimaging studies suggest that individuals with BPD show high levels of emotion regulatory efforts, associated with increased prefrontal control, possibly to compensate for enhanced amygdala reactivity (see Bohus et al., 2021). When instructed to do so, individuals with BPD use emotion regulation strategies like cognitive reappraisal to a similar extent as healthy controls (Krause-Utz, Walther, et al., 2019). In other words, self-reported emotion regulation may not necessarily reflect the ability to effectively implement adaptive emotion regulation strategies in daily life. Given the complex nature of emotion regulation, it should be assessed on different levels using multiple experimental and neurobiological approaches, such as neuroimaging. Future studies should also account for the interpersonal context in which problems are shown and reported, e.g. using ambulatory assessment.

Extending previous research, we explored if the Borderline Symptom List (BSL) could potentially be used to differentiate between internalized symptoms (intrapsychic strain) and externalized symptoms (dysregulated behaviours) of BPD in future research. More specifically, we explored if the BSL-23 and the BSL-S predicted lifetime BPD criteria assessed with a semi-structured clinical interview (IPDE). The BSL-23 was significantly correlated to fears of abandonment, instable relationships, instable identity, suicidal ideation, mood problems, chronic emptiness, and dissociation, while the BSL-S was associated with outwardly directed behaviours, such as self-damaging impulsive acts and anger outbursts. These preliminary findings suggest that the BSL-23 and its supplement may be useful to differentiate between internalized symptoms (intrapsychic strain) and maladaptive behaviours associated with BPD.

Such a distinction may be helpful given the heterogeneity of BPD psychopathology. While behavioural dysregulation is a core feature of BPD, not every person with the disorder shows markedly dysregulated behaviours and people suffering from a more hidden ‘quiet’ variant may run the risk of being misdiagnosed (Beatson et al., 2016; Bohus et al., 2021; Morgan et al., 2013; Winsper, 2021). Moreover, internalized symptoms seem to be more long-lasting (Biskin, 2015; Zanarini et al., 2007, 2016). Future studies may alternatively use the ICD-11 personality disorder severity scale (PDS-ICD-11; Bach et al., 2021) to assess a broader range of internalized and externalized symptoms. Higher levels of severity on this measure correlate with higher BPD criterion-count scores (Bach et al., 2021; Sharp et al., 2015; Wright et al., 2016). A previous study provided evidence that childhood trauma may be particularly related to internalized symptoms of depressivity (guilt, low self-worth, shame, and pessimism) and dissociation-proneness associated with BPD (Bach & Fjeldsted, 2017).

Current findings need to be interpreted in light of several limitations. Apart from the fact that the cross-sectional design of the study cannot establish causality, such analyses may also lead to overestimation of mediation effects (Maxwell & Cole, 2007). The use of self-reports, particularly the retrospective recall of childhood maltreatment, may be sensitive to recall biases. People with BPD may recall more negative memories and make negatively biased interpretations (Baer et al., 2012). At the same time, certain adverse experiences may be denied or minimized and recalling memories of them may be avoided. This might explain why we did not find associations between BPD symptoms and sexual abuse, as previously documented (Porter et al., 2020). Including documented files of childhood abuse may help to reduce the risk of a recall bias. The use of the behavioural supplement of the BSL-23 as a measure of behavioural expressions of BPD still needs to be further validated. In the current sample, BSL-S scores had a restricted range, which enhances the likelihood of a type II-error and may lead to an underestimation of effects. Studies with larger samples are needed to replicate or extend our results. Since we only included female participants, findings cannot be generalized to male samples. Studies that have investigated sex differences with respect to emotion regulation found that males report lower levels of emotional awareness when compared to women (Neumann et al., 2010). Women may be more likely to engage in maladaptive strategies implicated in the development of depression (Zimmermann & Iwanski, 2014).

Despite the above-mentioned limitations, our findings may have important implications. In supporting associations between childhood maltreatment experiences, emotion regulation difficulties, and BPD symptoms, our findings underscore the importance of addressing these features in prevention and intervention. People with a history of emotional maltreatment may be more prone to developing maladaptive ways of emotion regulation. Addressing these problems is part of evidence-based treatments, such as dialectical behavioural therapy (Linehan, 1993) and mentalization-based treatment (Fonagy & Bateman, 2008). A stepped-care model might be helpful to ensure adequate care based on the levels of severity. People displaying emotion regulation problems and symptoms associated with early and/or milder stages of BPD may profit from generalist-provided psychoeducation and emotion regulation or problem-solving skills, while those with more severe and persistent symptoms and substantial psychosocial impairments should be referred to intensive specialized treatments (Bohus et al., 2021). Involving family members and providing caregivers with additional support can be an important part of this treatment (Bailey & Grenyer, 2014).

In conclusion, our study provides evidence for close associations between childhood maltreatment, emotion regulation difficulties, and BPD symptom severity. More research is needed to investigate if these association differ for internalized versus externalized symptoms of BPD, preferably by using longitudinal and multi-method approaches.

Ethics approval and consent to participate

All subjects have given their written informed consent. The study protocol has been approved by the local Ethical committee of the Mannheim Medical Faculty of Heidelberg University.

Consent for publication

All co-authors gave consent for publication.

Authors’ contributions

RG and AKU designed the study and its rationale, designed the structure of the paper and drafted its first version. CN and AP recruited participants, performed diagnostic interviews and collected self-report data. AR provided theoretical input, performed systematic literature search, and helped with formatting the final manuscript. All authors contributed to the final version of the paper.

Acknowledgements

We thank all participants of this study for their crucial contribution. We thank Androniki Nikakis, Behazin Khosravi, and Linnea Ott for their contribution to the pre-processing of the data.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability

According to European law (GDPR), the sharing of data containing potentially identifying or sensitive information is restricted. Our data involving clinical participants are not freely available in the manuscript, supplemental files, or in a public repository. Data access can be requested on reasonable demand via the corresponding author.