Changes in Attachment Representation and Personality Organization in Transference-Focused Psychotherapy

Object relations theory has provided an integrated theoretical framework for understanding and assessing personality disorders (PDs). This framework has guided clinical practice and empirical research in key domains that are central to personality functioning, including identity, object relations, primitive defenses, and management of aggression (1–4). In a parallel development, attachment theorists, well grounded in developmental research identifying and assessing parent-child attachment patterns, have investigated the contribution of attachment representations in children and adults to normative or pathological personality development (5–8). Both object relations and attachment theorists have focused on explaining how maladaptive patterns of representing the self in relation to others and a painful sense of incoherence in the self are core features of personality pathology in general and borderline personality disorder in particular (7, 9–11).

From an object relations perspective, borderline personality disorder is characterized by instability in the individual’s sense of identity stemming from a defensive split between positive and negative representations of self and others, difficulties in affect regulation, unstable interpersonal relationships, impulsive aggression, behavioral dysregulation, and impaired mentalization (i.e., the capacity to think about the behavior of self and others in terms of intentional mental states) (2, 10, 12–16). These characteristics have etiologic roots in insecure infant-parent attachment, adverse childhood experiences, and temperamental or constitutional features (17). Central to object relations theory is the idea that internal representations of the self in relation to significant others, linked by an affect state, are the building blocks for higher-level psychological structures, including identity (self-functioning) and object relations (interpersonal functioning) (2, 18). This concept is compatible with the Alternative DSM-5 Model for Personality Disorders (4, 13, 19).

Attachment researchers have widened and sharpened the lens through which normative and pathological personality development are examined (5, 10, 20). Contributing to the understanding of the maladaptive mental representations of those with borderline personality disorder, attachment researchers have found a strong association between borderline pathology and insecure—especially disorganized (unresolved)—internal working models of attachment, in which individuals have failed to integrate or resolve early attachment trauma, such as loss or abuse (10). Attachment researchers have provided a conceptual framework and assessment measures that have increasingly been incorporated into psychotherapy research with patients with PDs (21, 22). However, few studies have systematically explored how attachment- and object relations–based approaches to understanding patients with PDs relate to one another (23).

Identity, Coherence, and Reflective Function

Kernberg (11, 24–26) has defined identity diffusion as the central severity criterion of PDs. From an object relations perspective, identity diffusion involves the lack of integration of concepts of the self and significant others, evident in “a painful sense of incoherence, objective inconsistencies in beliefs and behaviors, [and] difficulties with commitment to jobs, values, and goals” (11), along with a reliance on splitting-based defensive operations to protect positive, idealized representations from negative, persecutory ones. Identity is seen as a higher-order structure that organizes internal object relations in ways that are rigid or flexible, polarized or integrated (3, 27). Recent research has affirmed Kernberg’s view of the centrality of identity diffusion among patients with PDs (28, 29). Higher levels of identity diffusion are associated with more severe symptomatology, including suicide attempts (28–30), suggesting that this aspect of personality pathology should be targeted in both research and treatment.

Kernberg (11) has considered that insecure (especially disorganized) attachment, which involves unresolved loss and trauma and deficits in reflective capacity or mentalization, may also be a risk factor for identity diffusion and borderline personality disorder. The Adult Attachment Interview (AAI) (31) is thought to contribute to an understanding of the concepts of identity integration and identity diffusion by exploring two dimensions that have been found to be associated with attachment security: narrative coherence (32, 33) and reflective function (RF) (14). First, narrative coherence, which represents the capacity to access and reflect on difficult or painful attachment-related memories and relationships with self-cohesion, has been found to be the best predictor of attachment security on the AAI (34). This concept echoes formulations of identity consolidation from object relations theory as involving “a sense of coherence and authenticity across time and situation, the narrative description of the self” (27). Second, RF, as a separate measure that assesses the degree of mentalization in attachment relationships (as rated on the AAI), is also thought to play an essential role in identity, self-development, and self-regulation (14). Low ratings on the RF scale (35) have been linked to insecure and disorganized (unresolved) working models of attachment (14, 22, 36, 37) and to identity diffusion, the core component of personality pathology that has been found to mediate the relationship between mentalizing difficulties and interpersonal problems among those with borderline personality disorder (38). Improvements in the capacity for mentalization are thought to coincide with the capacity to develop fuller, more elaborated representations of the self and others (11, 14). Thus, the concepts of RF and narrative coherence that have emerged from attachment theory and research (7) are complementary to the structural domain of identity as formulated in object relations theory (11).

Measures of security of attachment, narrative coherence, and RF assess for problematic or traumatic aspects of early attachment relationships and experiences (7, 39). Kernberg spoke to this point when he stated that “a history of trauma can contribute substantially to the sense of painful incoherence associated with dissociative tendencies” (11). However, Kernberg questioned whether such measures adequately assess the splitting of the idealized, positive and the persecutory, aggression-laden segments of experience that interfere with identity integration in borderline personality disorder. In other words, when the self has not cohered, intense affects of love and aggression remain exaggerated, unintegrated, and split off, contributing to identity diffusion.

In the current study, we attempted to construct an empirical bridge between object relations theory and attachment theory by investigating the association between change in attachment representations, narrative coherence, and RF, as assessed with the AAI (31), and structural change in personality organization, including identity, object relations, and aggression, as assessed with the Structured Interview of Personality Organization (STIPO) (1). We used a subsample of patients with borderline personality disorder from the Vienna-Munich randomized controlled trial (RCT) (40).

The current study is a secondary analysis of data, which have been previously collected and published (21, 36), on the effect of 1 year of transference-focused psychotherapy (TFP) compared with 1 year of treatment by experienced community psychotherapists (ECPs) on attachment representations, personality organization, symptomatology, and global functioning among a sample of women with borderline personality disorder. Although patients’ depression and anxiety symptoms improved in both treatment groups, the group randomly assigned to receive TFP evidenced a lower dropout rate and showed significantly greater decreases in the number of patients attempting suicide, number of inpatient admissions, and borderline personality disorder symptoms and significantly greater improvements in personality organization and psychosocial functioning after 1 year of treatment (40). In other words, only those receiving TFP showed markers of structural change.

Buchheim et al. (21) found a significant change from disorganized (unresolved) to organized attachment among patients with borderline personality disorder and in overall personality organization among those in the TFP treatment group but not in the ECP group (40). In addition, Fischer-Kern et al. (36) reported a greater improvement in RF in the TFP group compared with the ECP group. Although significant improvements in overall global ratings of personality organization on the STIPO were found in conjunction with increased RF and security of attachment in the original RCT (21, 40), the specific STIPO domains of identity integration and aggression were not examined. These and other STIPO domains are consistent with the core processes targeted in TFP.

The primary aims of this study were to analyze whether changes in attachment representations, narrative coherence, and RF among patients with borderline personality disorder were associated with changes in domains of personality organization, particularly identity diffusion and aggression, and to compare the associations by treatment group (TFP vs. ECP). We also investigated whether changes in identity and aggression varied by treatment type.

Methods

Research Design

Our data came from an RCT comparing TFP with ECP for women with borderline personality disorder (40). Here, we focus on the data on the assessments of personality organization and attachment conducted before and after 1 year of treatment.

Participants

The RCT procedure, study setting, therapists, and participants have been previously described in detail by Doering et al. (40). Briefly, 104 female patients with borderline personality disorder, ages 18–45, were recruited between October 2004 and August 2006 and randomly assigned to receive either ECP (N=52) or TFP (N=52). The AAI, the STIPO, and the Global Assessment of Functioning (GAF) were administered to 92 patients before and after 1 year of treatment. The study was approved by the ethics commission of the Medical University of Innsbruck (ID UN1950) and was registered at Clinicaltrials.gov (NCT00714311). Participants who fulfilled the inclusion criteria were given a complete description of the study, gave written informed consent, and were assessed by trained research assistants.

Treatments

TFP is a manualized, structured, twice-weekly psychodynamic treatment for borderline personality disorder that has a treatment duration of 1 year or more and that focuses on the immediate interaction and transference between the patient and therapist (41–43). The treatment targets the affectively charged, unintegrated, and polarized representations of self and significant others that underlie the affective dysregulation, symptomatology, and deficits in self- and interpersonal functioning that characterize PDs. As these polarized extreme perceptions of self and others emerge in the therapeutic relationship, they are gradually modified and integrated, leading to identity consolidation.

ECP therapists had different theoretical orientations, including psychodynamic, supportive, and cognitive-behavioral therapy, but did not have specific training in manualized psychotherapy for borderline personality disorder. The mean±SD length of professional experience after completion of psychotherapy training was 8.9±9.8 years. No significant differences were found in the level of experience between the two groups of therapists (TFP or ECP) (40). The number of sessions conducted within the 1-year study period differed between the treatments (TFP=48.5±34.2, ECP=18.6±24.0, t=5.16, df=101, p<0.001).

Measures

AAI.

The AAI is a semistructured clinical interview comprising 20 questions asked in a set order with standard probes, intended to elicit thoughts, feelings, and memories about early attachment experiences. Participants are also asked to reflect on their parents’ style of parenting and how it influenced their development and lives as adults. The AAI’s scoring system is designed to quantify an individual’s current state of mind with respect to childhood attachment relationships. Possible scores range from 1 to 9, with higher scores (e.g., in coherence) indicating a better capacity to give an integrated, credible, organized narrative. Main and Goldwyn (44) identified three major organized patterns of adult attachment—secure/autonomous, dismissing, and preoccupied—that can be derived from these responses, as well as two disorganized classifications—unresolved and cannot classify. Whereas individuals with organized patterns of attachment have the capacity to mobilize a consistent strategy for approaching attachment-related memories, those with disorganized (unresolved) attachment patterns are characterized by failure to resolve attachment-related trauma (e.g., loss or abuse) and dissociative defenses for coping with attachment stress. For the current study, we compared organized (secure/autonomous, dismissing, preoccupied) groups with disorganized (cannot classify or unresolved trauma) groups. The AAIs were coded by a certified rater for all five attachment classifications.

RF scale.

The AAI transcripts were also coded independently by certified RF raters (different from the AAI coder) by using the RF scale, a separate measure that employs AAI transcripts to assess the capacity for mentalization. The RF scale has an 11-point range that captures the quality of mentalization in the context of attachment relationships. Scores range from –1, entirely absent or severely distorted reflectiveness, to 9, exceptional reflectiveness (the interviewee demonstrates unusually multifaceted, complex, or elaborated reasoning about mental states).

STIPO.

The STIPO is a semistructured clinical interview designed to evaluate level of personality organization. This measure yields dimensional scores on seven domains of the structure and organization of personality, ranging from 1 (absence of pathology) to 5 (severe pathology): identity (consistency and stability of self and the capacity to invest in one’s primary role in school or work), quality of object relations (shifts in experience of the self or other and level of interpersonal conflict), management of aggression (directed against self or other), defensive functioning (reliance on primitive [e.g., splitting, projection] vs. higher-level defenses [e.g., intellectualization, rationalization]), coping ability, moral values (antisocial behaviors as well as the patient’s capacity for remorse and guilt and concern for others), and reality testing (capacity to distinguish between self and non-self, intrapsychic experience and external reality, and fantasy and reality, or the capacity to understand and adhere to consensual reality). This interview has excellent reliability and validity, as shown by previous research (3).

GAF.

The GAF (45) is used to rate the severity of mental illness. It measures how much a person’s symptoms affect their day-to-day life on a scale of 0–100 (higher scores indicate better psychosocial functioning) and was designed to help mental health providers understand how well a person can manage everyday activities.

Hypotheses and Data Analyses

We aimed to investigate whether changes in attachment representations, narrative coherence, and RF on the AAI among patients with borderline personality disorder are associated with changes in core domains of personality organization, especially identity, object relations, and aggression, on the STIPO after 1 year of TFP.

We formulated three main hypotheses concerning patients with borderline personality disorder. First, we hypothesized that patients with disorganized (unresolved) attachment would have lower levels of personality organization, particularly identity and aggression, at baseline. Patients with less coherent narratives and lower levels of RF were expected to have lower levels of personality organization, including higher levels of identity diffusion and more difficulty with management of aggression. Second, we hypothesized that patients with borderline personality disorder who changed from disorganized (unresolved) to organized attachment on the AAI would also show significant improvement in personality organization on the STIPO, particularly in identity integration, after 1 year of TFP, whereas those who did not change their attachment representation would continue to show indications of identity diffusion. Third, we hypothesized that patients receiving TFP would show greater improvement on the STIPO identity and aggression subscales compared with those receiving ECP.

All analyses were conducted by using R, version 4.2.1. To examine the relationship between attachment and personality organization at baseline, one-way analyses of variance (ANOVAs) were used for categorical comparisons and Pearson’s correlation coefficients were used for continuous comparisons. For within-person comparisons of change in personality organization between patients achieving versus not achieving an organized attachment style after treatment, we used a nonparametric Wilcoxon signed rank test. A repeated-measures ANOVA with two measurements (baseline and treatment termination) was used to compare patients receiving TFP with those receiving ECP. For comparisons of how changes in attachment, personality organization, and other outcomes co-occurred, we used partial correlations of change scores, controlling for baseline values of the two variables being compared (i.e., AAI coherence and RF).

Results

Relationship Between Attachment and Personality Organization at Baseline

As expected in our first hypothesis, attachment was significantly related to personality organization in this sample, whether viewing attachment categorically (disorganized (unresolved) to organized) or continuously (coherence). Compared with patients with organized attachment classifications, patients with disorganized (unresolved) attachment had significantly more impairment as assessed by overall STIPO rating (β=0.47, 95% CI=0.24–0.70, t=4.03, df=83, p<0.001, d=0.88), identity subscale rating (β=0.44, 95% CI=0.18–0.71, t=−3.29, df=83, p=0.001, d=0.72), and aggression subscale rating (β=0.44, 95% CI=0.15–0.74, t=2.98, df=83, p=0.004, d=0.65) at baseline. Furthermore, patients with disorganized (unresolved) attachment exhibited more borderline personality disorder symptoms on the Structured Clinical Interview for DSM-IV Axis II Disorders (β=0.61, 95% CI=0.07–1.15, t=2.25, df=83, p=0.027, d=0.49) and were rated as having lower functioning on the GAF (β=−3.23, 95% CI=−6.01 to −0.44, t=−2.30, df=83, p=0.024, d=−0.50) (21).

Coherence ratings at baseline were significantly correlated with the STIPO, such that patients with less coherent AAI narratives exhibited lower levels of personality organization (r=−0.34, df=92, p=0.001), including more impairment on the identity (r=−0.24, df=92, p=0.023) and aggression (r=−0.31, df=92, p=0.003) subscales. By comparison, RF scores were not significantly associated with the overall STIPO rating (r=−0.14, df=92, p=0.158) or with the subscale scores for identity (r=−0.05, df=92, p=0.625) and aggression (r=−0.18, df=92, p=0.090). Coherence and RF had a moderate positive correlation (r=0.38, df=92, p<0.001).

Change in Personality Organization Related to Change in Attachment

As expected in our second hypothesis and as indicated in Table 1, patients with disorganized (unresolved) attachment who changed to organized attachment on the AAI after TFP improved in most aspects of their personality organization on the STIPO, including in coherence and continuity, identity, the internal working model, object relations, primitive defenses, aggression, moral values, and reality testing. By contrast, the subgroup of patients who did not change from disorganized (unresolved) to organized attachment improved only in one aspect of their personality organization—aggression.

We also examined the relationship between continuous measures of attachment (i.e., AAI coherence scores) and RF in relation to changes in core borderline personality disorder symptoms, personality organization, and GAF scores (Table 2). As the findings reported in Table 2 indicate, improvements in continuous measures of attachment and RF were correlated with improvements in borderline personality disorder symptoms, domains of personality organization, and global functioning.

Change in Identity and Aggression in TFP and ECP

As hypothesized in our third hypothesis, overall, patients receiving TFP had greater improvements in identity integration as measured on the STIPO, compared with patients receiving ECP (F=6.73, df=1 and 102, p=0.011, d=0.51). However, patients did not differ in improvements on the STIPO identity subscale as a function of their AAI classification (F=0.75, df=3 and 83, p=0.745) or as a function of the interaction between their AAI classification and their treatment group (F=0.31, df=3 and 83, p=0.257). In addition, patients who were treated with TFP, as compared with ECP, showed greater improvements on the STIPO aggression subscale (F=4.95, df=1 and 102, p=0.028, d=0.43). Patients did not differ in improvements on the STIPO aggression subscale as a function of their AAI classification (F=0.54, df=3 and 83, p=0.660) or as a function of the interaction between their AAI classification and their treatment group (F=1.54, df=3 and 83, p=0.210). These findings indicate that patients receiving TFP generally experienced greater improvements in identity integration and aggression, regardless of attachment status. Improvements on the STIPO aggression subscale during treatment were significantly correlated with improvements on the STIPO identity subscale (F=8.16, df=1 and 101, p=0.005, r=0.28).

Discussion

Our main finding was that patients with borderline personality disorder who shifted from disorganized (unresolved) to organized attachment on the AAI after 1 year of TFP also showed marked improvement in most domains of personality organization on the STIPO. In contrast, patients who did not change from disorganized (unresolved) to organized attachment improved in only one domain of personality organization on the STIPO—aggression. The strengthening of identity and diminution of aggression seen among patients in TFP are theorized to result from the integration of polarized affect states, and the representations of self and other with which they are linked, as they are enacted and interpreted in the “here and now” of the transference. This integration gradually enables the patient to behave and think in a more organized and integrated fashion, allowing for increased identity consolidation (41).

These findings go beyond previous studies that have shown change in both attachment security and personality organization over the course of TFP (22, 36, 40, 46) to highlight the key role that identity diffusion plays in determining the severity of borderline personality disorder. That patients with disorganized (unresolved) attachment who do not change to organized attachment status in TFP show changes in aggression, but not identity diffusion, suggests that identity diffusion indexes the severity of personality pathology. These findings also suggest that the AAI may evaluate the level of identity integration or diffusion as well as the level of organization in attachment representations and the way they change over the course of treatment. It is possible that narrative coherence and RF, which in previous research have been shown to be related to identity diffusion (38), may index different aspects of identity pathology. Although RF and coherence are correlated and both were correlated with improvements in identity and aggression after 1 year of TFP, RF was not associated with overall STIPO personality organization or subscales measuring identity and aggression at baseline (47).

This finding suggests that the two measures assess different constructs. Whereas RF represents the individual’s capacity to understand the behaviors of self and others in terms of intentional mental states, personality organization covers a wider range of dimensions, including self- and object perception, defenses, moral values, and reality testing. In addition, assessment of an individual’s mentalizing capacity is derived from attachment narratives, whereas the STIPO focuses on the investigation of important domains of personality functioning both in an individual’s report of their current life and in the way they present during the interview.

In sum, our findings show that working through the maladaptive self-object affect dyads that comprise the internal world in the here and now of the transference in TFP leads to significant changes on two fronts. First, it reformulates the representations of self in relation to early attachment figures, as evidenced by a shift from disorganized (unresolved) to organized attachment models on the AAI and improvement in object relations on the STIPO. Second, it catalyzes identity consolidation and modulates aggression, which may be interlinked. The modulation of aggression found in this study among patients with borderline personality disorder with both organized and disorganized attachment after 1 year of TFP treatment suggests that the identification of polarized affect states and the self-other representations with which they are linked fosters improvements in crucial aspects of personality functioning. In fact, the primary mechanism of change in TFP is hypothesized to be increased affect regulation through reflection on the affectively charged perceptions of self and others as they are experienced in the patient’s interactions with the therapist and significant others. As polarized positive and negative experiences of self in relation to others are integrated and reflected on as mental states, the extreme affects with which they are linked are contained and modified, and identity coalesces. Another way to think about this process in terms of attachment theory is that the treatment modality in TFP, with its focus on interpretation of the here-and-now relationship with the therapist and others, activates internal working models of attachment, with the goal of moving the patient toward increased attachment security. That improvements on the STIPO were found regardless of the patient’s attachment status indicates that TFP successfully improved personality organization across a range of different domains of self- and interpersonal functioning among patients with different attachment representations.

Contrary to expectation, in this study, aggression improved among patients receiving TFP and those receiving ECP after 1 year. This finding should be considered in light of the fact that, in the most severe indices of self-directed aggressive behaviors, suicidality and suicide attempts, change was observed in the TFP group but not in the ECP group (40). It should be noted that the STIPO aggression subscale focuses not only on suicidality but also on hostility, resentment, and enacted aggression against others (4)—all fundamental features of patients with borderline personality disorder that are targeted by experienced therapists of all theoretical persuasions.

Our relatively small sample size limited the generalizability and precision of our findings. Also, because our study included only women, our findings cannot be generalized to men, who might show different patterns, particularly regarding aggression. It should be noted that despite these limitations, one strength of the current study is that we conducted analyses of the relationship between attachment and personality organization as continuous variables and attained convergent results.

Conclusions

Our findings highlight the connections and disparities between object relations theory and attachment theory, thereby deepening the perspective of the internal world and potentially enhancing treatment of patients with PDs. These findings have several clinical implications. First, they highlight the importance of identifying and containing aggression and understanding its dysregulating impact on the process of identity consolidation and self-definition. Second, the results underline that the integration of polarized representations of self and other into a coherent identity can be assessed by measures of attachment and personality organization, which are valuable diagnostic tools to evaluate treatment success.