The major theorist of identity is Erik Erikson (1, 2), who popularized the term in his discussion of identity crises in adolescence. According to Erikson, identity includes role commitments, a sense of personal sameness or continuity over time and across situations, a sense of inner agency, and some acknowledgment of one’s role commitments and views of oneself by the broader community. Erikson argued that adolescents in many cultures experience a period of identity crisis, from which they emerge with some balance between identity achievement and identity confusion. A healthy identity includes the ability to choose an appropriate avenue for industry, achieve intimacy with another, and find a place in the larger society.
The opposite pole of identity is identity confusion, which Erikson originally called identity diffusion. Identity confusion manifests itself in a number of ways: 1) in a subjective sense of incoherence; 2) in difficulty committing to roles and occupational choices; and 3) in a tendency to confuse one’s own attributes, feelings, and desires with those of another person in intimate relationships and hence to fear a loss of personal identity when a relationship dissolves. Some individuals escape this state by choosing a "negative identity" (i.e., a role that is inappropriate or unusual given the individual’s attributes, such as race or socioeconomic status) that often constitutes a role or group identification negatively viewed by the broader culture.
Reviewing both the empirical and theoretical literature on self and identity, Westen (3, 4) summarized the major components of identity as being a sense of continuity over time; emotional commitment to a set of self-defining representations of self, role relationships, and core values and ideal self-standards; development or acceptance of a world view that gives life meaning; and some recognition of one’s place in the world by significant others.
The major research on identity reflects the work of James Marcia (5, 6), who operationalized Erikson’s theories into "identity statuses" or types. Marcia distinguished four identity statuses: identity achievement, moratorium, foreclosure, and identity diffusion. For identity achievement, a person must have struggled with issues of family, profession, religion, and values and have come to some kind of committed resolution. A person who falls into the moratorium category has put off resolution of identity issues and remains in an extended state of identity search. Foreclosure is the label assigned to people who have made major role commitments without ever seriously considering alternative possibilities or experiencing any period of struggle; people with a foreclosed identity have chosen a kind of de facto identity. Finally, the most severe identity problems are found in people with identity diffusion, who may have had multiple identity crises, chosen a succession of careers or religions, or may not even be aware of their lack of a cohesive identity.
Several clinical theorists have attempted to describe the nature of identity disturbance in borderline personality disorder. According to Kernberg (7, 8), identity diffusion in patients with borderline personality organization reflects an inability to integrate positive and negative representations of the self, much as the patient has difficulty integrating positive and negative representations of others. The result is a shifting view of the self, with sharp discontinuities, rapidly shifting roles (e.g., victim and victimizer, dominant and submissive), and a sense of inner emptiness. Kernberg also emphasized the way defenses that allow patients with borderline personality disorder to remain comfortable with remarkable inconsistencies inhibit the capacity to form a coherent view of themselves.
Adler and Buie (9, 10) described patients with borderline personality disorder as suffering from a sense of incoherence and disjointed thinking, feelings of loss of integration, concerns about "falling apart," and a subjective sense of losing functional control over the self and other forms of "self-fragmentation." From a self-psychological perspective, these patients lack an ability to internalize many aspects of their primary caregivers that would allow them to develop a cohesive sense of self. Fonagy and colleagues (11) drew upon empirical data with both borderline patients and maltreated young children to emphasize the failure of patients with borderline personality disorder to develop the capacity to step inside the mind of another and to imagine the way the other experiences the patient. Historically, social identity theorists such as the symbolic interactionists (notably George Herbert Mead) emphasized the extent to which our views of ourselves result from the reflected appraisals of others—that is, from seeing ourselves in others’ eyes and hence learning about who we are. To the extent that patients with borderline personality disorder have difficulty seeing themselves in the mind’s eye of another, they should have difficulty in developing coherent identities.
Systematizing the clinical and theoretical literature, Westen and Cohen (12) summarized the major attributes of identity disturbance hypothesized to be central to borderline personality disorder. These include a lack of consistently invested goals, values, ideals, and relationships; a tendency to make temporary hyperinvestments in roles, value systems, world views, and relationships that ultimately break down and lead to a sense of emptiness and meaninglessness; gross inconsistencies in behavior over time and across situations that lead to a relatively accurate perception of the self as lacking coherence; difficulty integrating multiple representations of self at any given time; a lack of a coherent life narrative or sense of continuity over time; and a lack of continuity of relationships over time that leaves significant parts of the patient’s past "deposited" with people who are no longer part of the individual’s life, and hence the loss of shared memories that help define the self over time.
The clinical literature on identity disturbance in borderline personality disorder provides a rich conceptual foundation for understanding identity disturbance, but empirical research remains limited (13, 14). A central issue in understanding identity disturbance in patients with borderline personality disorder is the relationship between identity disturbance and a history of sexual abuse. Research suggests that 30%–75% of adult and adolescent patients with borderline personality disorder have reported histories of sexual abuse (15–17). In addition, sexual abuse history and dissociative experiences are either common in, or diagnostic of, both borderline personality disorder and dissociative identity disorder (18–21). Given the association between sexual abuse and dissociation, the high percentage of borderline personality disorder patients with sexual abuse histories raises questions about whether identity disturbance is really characteristic of borderline personality disorder or rather of a history of severe and pervasive sexual abuse.
The present study represents an empirical examination of identity disturbance with two aims: to clarify the construct of identity disturbance and to try to discern the features of identity that distinguish patients with borderline personality disorder from other psychiatric patients.
Respondents for this study were experienced psychologists, psychiatrists, and social workers. The use of clinicians (rather than patients) as respondents is a growing practice in psychiatric research (22–24). Aside from substantially increasing the numbers of patients who can be included in a study (and hence increasing generalizability and power), the use of clinicians has several advantages. Clinicians tend to be sophisticated observers, who see a patient longitudinally and can often offer more informed and potentially less biased judgments than patients themselves or interviewers who see the patient for 90 minutes or less. Clinicians can, of course, be biased by their theoretical preconceptions; however, all observers have theories and hence potential biases, such as the intuitive theories patients hold about themselves (that is, their conscious self-concepts, through which their answers to the standard questionnaires and structured interviews are always filtered). The question, then, is whether, in a given research domain, self-report biases or clinician-report biases are likely to be greater. Four factors led us to prefer trained observers as our informants: 1) the absence of shared theories about the multidimensional nature of identity disturbance that could produce bias; 2) the possibility of drawing from clinicians with diverse training experiences (psychiatrists, psychologists, and social workers) who would not likely share the same biases; 3) prior research that had used this method and demonstrated that clinicians do not tend to rely on diagnostic prototypes in describing their patients but instead tend to describe what they see clinically (23, 24); and 4) the problematic nature of asking patients about phenomena such as their conflicts over ethnicity, gross inconsistencies between what they say and what they do, and the tendency to define themselves in terms of extreme groups or roles.
Clinicians were given a diagnostic/demographic/developmental history form adapted from previous studies (23, 24) and an identity disturbance questionnaire designed expressly for the purpose of this research.
We solicited data from clinicians at The Cambridge Hospital/Cambridge Health Alliance at Harvard Medical School by contacting staff and trainees by mail and internal e-mail, which yielded descriptions of 50 patients. Clinicians were paid a token honorarium of $10. We initially planned to include only female patients (to minimize heterogeneity, given the high percentage of women among patients with borderline personality disorder in the population); to limit the patient group to those who had been in psychotherapy for less than a year (to ensure that major symptoms had not changed with treatment); and to limit the patient group to those between the ages of 18 and 40 (again to limit heterogeneity). After it became clear that this would not yield a large enough study group, we expanded our criteria by 1) eliminating gender and age restrictions, 2) including patients seen up to 2 years in treatment, and 3) not predetermining whether the clinician should describe a patient with borderline personality disorder. In addition, we added a second cohort of clinicians (N=45) who completed questionnaires at a workshop on personality organization in Washington, D.C. With these modifications, we obtained our intended study group size, which we had preselected based on power considerations. Clinician respondents were, on the average, quite experienced, with mean of 18.13 years (SD=11.09) of clinical experience. They also knew the patients well; the median length of treatment was 30 sessions (mean=53.27, SD=78.59). Clinician respondents included psychiatrists (N=17), psychologists (N=41), and social workers (N=32).
Identity Disturbance Questionnaire
This assessment tool included 35 items rated on a 1–7 scale (1=not true at all, 4=somewhat true, and 7=very true). We included a mixture of items that required some inference (e.g., "sense of identity revolves around membership in a stigmatized group") and items that described relatively manifest aspects of the patient’s life. In general, the questions were written to require minimal interpretation on the part of the clinician and hence to minimize room for unreliability as a result of idiosyncratic reading of the items.
The item set assesses multiple possible indices or manifestations of identity disturbance, such as unusual name changes (other than marriage), contradictory beliefs and behaviors, frequently changing values, feelings of inner emptiness, and confusion over sexual orientation. We developed items by examining the relevant theoretical, clinical, and research literatures. For example, Erikson (2) posited the existence of a negative identity, in which a person chooses a label or persona for the self that seems inappropriate given his or her socioeconomic status, gender, or ethnicity; often this identity focuses on being "bad." Thus, we operationalized the concept of a negative identity with items such as "sense of identity revolves around membership in a stigmatized group (e.g., child of an alcoholic, sexual abuse survivor)." Also derived from Erikson’s theoretical work (1, 2) and Marcia’s research on identity (6) were various questions regarding professional commitments, political beliefs, and changes in sexual orientation.
From the clinical literature and our own clinical observations we derived a number of phenomenological items that described the patient’s own feelings and thoughts about his or her identity (e.g., "false self," "unreal") as well as descriptions of problems with identity that have been apparent to clinicians who work with borderline patients. From the research literature on dissociation, we drew items such as "has trouble telling life story; narrative accounts have large gaps or inconsistencies."
Diagnosis, Demographics, and Developmental History
Given potential unreliability of clinician diagnoses and to maximize reliability, we measured diagnosis in a variety of ways. First, we asked if the patient had an axis II diagnosis and, if so, to name it. Later, clinicians rated by means of a 7-point scale (where 1=not at all and 7=very much) the extent to which the patient displayed symptoms of each of the 10 axis II personality disorders in DSM-IV. Because we were most concerned with the validity of borderline personality disorder diagnoses, we also included a list of the nine DSM-IV criteria and asked clinicians to make two determinations: 1) whether each symptom was present or absent and 2) the extent to which each item applied to their patient, again according to a 7-point scale.
We included four additional validity checks. First, we asked clinicians to report the patient’s Global Assessment of Functioning score. We then asked clinicians to report on the patient’s employment history, quality of peer relationships, and amount of social support (the number of close relationships or people in whom the patient feels comfortable confiding that the patient had described to the clinician).
We also asked for basic demographic information, including the patient’s age, sex, race, and socioeconomic status and the clinician’s sex, race, discipline, years of experience, and theoretical orientation. We inquired about clinician characteristics to investigate any possible clinician bias reflecting differential training or demographics. We also inquired about the psychotherapy setting (e.g., outpatient clinic or private practice) and number of sessions the clinician had seen the patient.
Finally, we assessed a number of aspects of the patient’s developmental history. The most important was information about the patient’s history of childhood abuse, particularly sexual abuse. Because sexual abuse history is of particular relevance to questions of identity disturbance, we asked clinicians not only to report no/unsure/yes to questions about abuse history but also to report the basis for the belief about the patient’s abuse history. To be coded as having been sexually abused, the clinician needed to report that the patient had conscious memories when treatment began or that the abuse had been confirmed by an outside source (e.g., Department of Social Services, police, or doctor).
Three questions were central to this investigation: 1) whether identity disturbance is a unitary or multidimensional construct; 2) which aspects of identity disturbance distinguish patients with borderline personality disorder; and 3) to what extent these aspects of identity disturbance continue to predict borderline diagnosis after controlling for history of sexual abuse. To address these questions, we first subjected the identity questionnaire data to factor analysis to see which items aggregated together. We then used t tests to compare the mean scores of each item to determine which, if any, would distinguish patients with borderline personality disorder. We then ran a contrast analysis for each factor to test hypotheses about identity disturbance in patients with borderline personality disorder, patients with personality disorders other than borderline personality disorder, and those with no personality disorder. Multiple regression analyses were performed to predict diagnosis from identity disturbance scores after controlling for sexual abuse history. Finally, on the basis of recent cluster-analytic research that has suggested a distinction between two borderline personality disorder subtypes, characterized either by emotional dysregulation or histrionic features (24, 25), we conducted additional analyses that assessed the relationship of histrionic features to identity disturbance holding borderline features constant, and vice versa.
The patients described by the clinician respondents ranged in age from 18 to 66 years (mean=35.5, SD=10.1) and had Global Assessment of Functioning Scores that ranged from 25 to 85 (mean=62.7, SD=13.8). Additional characteristics of the patient group are presented in t1.
We initially intended simply to compare patients with borderline personality disorder (N=34) to patients without the disorder. However, when we examined the data, we found that the nonborderline group (N=61) included 20 patients diagnosed by their clinicians with personality disorders other than borderline personality disorder, which allowed us to test hypotheses more conservatively. Although the reliability of the "other personality disorders" category is unknown, we were able to assess whether these patients differed in their general social adjustment and level of psychopathology from the subjects with no personality disorder. In any event, diagnostic unreliability and smaller group sizes would foster type II, rather than type I, errors; thus, any findings that emerged with this three-group comparison would be very conservative. Given the preliminary nature of this study, which constitutes the first (to our knowledge) systematic effort to explore the precise nature of identity disturbance in patients with borderline personality disorder, this conservatism seemed warranted. As a secondary analysis, we ran analyses with just two groups (borderline personality disorder versus no borderline personality disorder). The results were the same or stronger in all cases, so here we report the more conservative findings.
As expected, gender ratios differed somewhat among the groups. In the borderline personality disorder group, 82.4% of subjects (N=28) were female, which resembles the gender ratio for this disorder in the population (approximately 75%, according to DSM-IV). Half of the subjects with other personality disorders were female; the corresponding percentage in the no personality disorder group was 65.9% (N=27). Thus, we used sex as a covariate in subsequent analyses.
To distinguish patients with borderline personality disorder from other patients, we first asked clinicians to fill in an axis II diagnosis. As a validity check, we then asked them to 1) indicate whether each of the nine DSM-IV criteria for borderline personality disorder was present or absent, 2) rate the extent to which the patient showed symptoms of borderline personality disorder as well as symptoms of each of the other axis II personality disorders, and 3) report on level of functioning (Global Assessment of Functioning scores, employment history, and number and quality of peer relationships).
Data relevant to diagnostic reliability for the 95 subjects are presented in t2. To be diagnosed with borderline personality disorder, a patient must manifest at least five of the nine DSM-IV criteria. In this study, patients diagnosed with borderline personality disorder fulfilled an average of 7.4 criteria. Patients with other personality disorders or no personality disorder averaged fewer than five. Differences among these means were statistically significant. Similar findings emerged when we compared the severity ratings of each borderline personality disorder criterion.
To compare personality pathology, we averaged all the personality disorder symptom ratings for the three groups for each of the 10 personality disorders, which produced a composite index of the extent to which patients in each group had personality disorder symptoms of any kind. As expected, the borderline personality disorder group had the highest mean ratings, followed by the subjects with other personality disorders and then those with no personality disorder. The differences among these means were significant. As a specific validity check on the diagnosis of "other personality disorders," we then recomputed these means excluding borderline personality disorder symptom ratings. Post hoc comparisons revealed significant differences but only between the no personality disorder group on the one hand and both personality disorder groups on the other. Thus, both personality disorder groups appeared to score higher on personality disorder characteristics than the no personality disorder group, and the borderline personality disorder group was specifically higher than both other groups on borderline characteristics.
Level of functioning variables also provided support for diagnostic reliability. Mean scores on the Global Assessment of Functioning were significantly different for the three groups as were the quality of peer relationships and the number of close relationships; employment stability, however, did not differ.
One peculiarity of the data did emerge: 19.7% of the patients without borderline personality disorder (N=12, split roughly evenly between the subjects with other personality disorders and those with no personality disorder) fulfilled five borderline criteria. We suspect this reflects both the high comorbidity of borderline personality disorder with nearly all other personality disorders and the tendency of clinicians to prioritize axis II diagnoses, giving such diagnoses as "narcissistic personality disorder with borderline features" (26). We addressed this in two ways. First, in line with our strategy of minimizing type I errors and maximizing conservatism of the findings, we kept patients in the diagnostic categories to which clinicians assigned them. If some subjects with no personality disorder really met borderline personality disorder criteria, that would reduce mean differences between the two groups, not overestimate them. Second, as planned, we supplemented categorical analyses with continuous analyses, with number of borderline criteria as the dependent variable, and used multiple regression to predict the number of borderline personality disorder criteria from identity factor scores. Thus, our findings are applicable not only to categorical but to dimensional analyses of borderline personality disorder symptoms.
The prevalence and confirmation of abuse history for the patients are reported in t1. Of the 24 clinicians who marked "Yes" on the sexual abuse question, all reported confirmation from at least one outside source. Four clinicians who completed this section marked the sexual abuse history as "Unsure." None of these four patients entered treatment with clear, conscious memories or documenting evidence of abuse. The data thus suggest that clinicians were using sensible (although of course not infallible) algorithms in determining degree of likelihood of sexual abuse history.
To ascertain the extent to which our measure of identity disturbance was really capturing the construct it was designed to assess, we used clinicians’ ratings of whether the identity disturbance criterion from DSM-IV was present or absent, which divided patients into two groups. We then compared the mean ratings for each identity disturbance item for these two groups in an effort to assess the extent to which our items were capturing the same construct clinicians classify as identity disturbance. Twenty-eight of the 35 items significantly discriminated between patients with and without identity disturbance at the 0.05 level (conservatively using two-tailed tests, even though predictions were directional), which indicates that our items did in fact address identity disturbance as clinicians conceptualize it.
The items that did not predict identity disturbance tended to describe unusual phenomena with low base rates, such as "patient ‘displays’ identity in ways that appear unusual or deviant (e.g., multiple tattoos, piercings, highly peculiar hair style or coloring)," on which most subjects received a rating of "1." Although deviant appearance may be an indicator of identity disturbance in the general (or psychiatric) population, our study group size may not have been large enough to detect this.
Distinguishing Identity Disturbance in Borderline Personality Disorder: Item, Factor, and Construct Analyses
To examine the nature of identity disturbance in borderline personality disorder, as a first step we compared means for each item for patients with borderline personality disorder with means for each of the other two diagnostic groupings (t3). To be conservative, we only considered those items that distinguished borderline personality disorder from both of the other groups as clear markers of borderline identity disturbance, again by using two-tailed t tests despite directional hypotheses. Thirty-two of the 35 items distinguished subjects with borderline personality disorder from those with no personality disorders; of these, 17 items also distinguished subjects with borderline personality disorder from those with other personality disorders. Thus, the data suggest that robust differences do exist between patients with borderline personality disorder and other patients, whether or not they have a personality disorder. It is important to note that the data did not support potential concerns about clinician response bias. Respondents did not simply give high ratings to all 35 indicators of identity disturbance if they had diagnosed the patient with borderline personality disorder and give low ratings otherwise; over half the items that discriminated patients with borderline personality disorder from those with no personality disorder did not discriminate patients with borderline personality disorder from those with other personality disorders, who clinicians clearly identified as nonborderline.
To discern whether identity disturbance is a unitary or multidimensional construct, we first ran a principal-components analysis to identify the factors of the identity disturbance questionnaire that aggregated together, specifying eigenvalues of >1. Examination of the scree plot and explained variance suggested four factors that together accounted for 54.39% of the variance. We then subjected the data to an orthogonal (varimax) rotation, specifying four factors. (Other factor solutions produced similar results.)
The first factor was role absorption, in which patients appeared to absorb themselves in, or define themselves in terms of, a specific role, cause, or unusual group. The second factor, painful incoherence, reflected patients’ subjective experience and concern about a lack of coherence. The third factor, inconsistency, was characterized less by subjective than objective incoherence (i.e., did not imply distress). The fourth factor was lack of commitment (i.e., to jobs or values). The factors all showed high internal consistency, with the following reliabilities (coefficient alpha): factor 1=0.85, factor 2=0.90, factor 3=0.88, factor 4=0.82. t4 describes the items that loaded above 0.50 on each factor.
To see whether patients diagnosed with borderline personality disorder would differ from other patients on these four dimensions of identity disturbance, we compared the means of the three groups by using a one-way analysis of variance (ANOVA) for an omnibus F, and then tested specific hypotheses by using contrast analysis (27). The ANOVA showed a significant difference between the three groups on the first (F=3.87, df=2, 92, p=0.02), second (F=16.14, df=2, 92, p<0.001), and third (F=4.82, df=2, 92, p=0.01) factors and approached significance on the fourth (F=2.65, df=2, 92, p=0.08).
The more important analysis is the contrast analysis, which asks more focused questions than the omnibus F. We tested three competing hypotheses for each factor, specified in advance. 1) Borderline patients would score higher than the other two groups, who would not differ from each other (contrast weights: 2, –1, –1). 2) Scores for the three groups would be linearly related, such that borderline personality disorder patients would have the highest scores, followed by patients with other personality disorders and then those with no personality disorder (contrast weights: 1, 0, –1). 3) Mean scores would follow the same order as in the previous contrast but with a larger mean difference between patients with borderline personality disorder and those with other personality disorders than between patients with other personality disorders and those with no personality disorders (contrast weights: 4, –1, –3). The second, linear model, tended to be the most robust, revealing predicted differences among all three groups. These differences are all the more striking given the limited diagnostic reliability data for the group with other personality disorders. The results of these analyses are detailed in t5. (For simplicity, we only report the first two contrasts in each case, because contrasts two and three were largely redundant.)
Analyzing the data a second way, we used multiple regression to predict borderline pathology, measured dimensionally by the number of DSM-IV borderline symptoms, from patients’ scores on the four identity factors. The four factors together (R=0.71) accounted for 50.2% of the variance, with the first three factors contributing significantly to the variance (p<0.05) and the fourth showing a trend (p=0.10).
Disentangling Borderline Personality Disorder and Sexual Abuse
The data thus far indicate that patients with borderline personality disorder do indeed differ from both subjects with other personality disorders and those with no personality disorder in multiple dimensions of identity disturbance. What is not indicated is whether, or to what extent, those findings reflect the greater incidence of sexual abuse in patients with borderline personality disorder (or gender differences, given the higher prevalence of borderline personality disorder among women).
We thus wanted to examine the extent to which a diagnosis of borderline personality disorder contributed to factor scores independent of a patient’s gender and sexual abuse history. To accomplish this, we used multiple regression to assess the predictive value of gender (dummy coded 0=male, 1=female), sexual abuse history (coded 0=no, 1=unsure, 2=yes), and diagnosis (borderline personality disorder=1, no borderline personality disorder=0) for each factor, entering gender and sexual abuse history in the first step and diagnosis in the second.
Sexual abuse history was correlated to varying degrees with each factor; however, in all cases, the model that included borderline personality disorder diagnosis was significantly more predictive than the model that included only patient gender and sexual abuse history (t6). For the first factor, role absorption, gender and sexual abuse contributed substantially less than the borderline diagnosis to the predictive power of the model. The second factor, characterized by painful feelings of identity incoherence, was highly correlated with sexual abuse history, although inclusion of borderline personality disorder diagnosis significantly improved the model’s predictive power. For the other two factors, patient gender and sexual abuse history did not account for enough variance to reach significance. The fourth factor was best predicted by the model that included borderline personality disorder diagnosis, but this model did not account for a significant amount of the variance.
In a second set of analyses, we assessed the relationship between sexual abuse and severity of the disorder (total number of symptoms). A history of sexual abuse predicted a higher likelihood of borderline features (r=0.49, df=90, p<0.001). A regression model including patient gender and sexual abuse (R=0.51) accounted for 25.6% of the total variance on number of symptoms present (F=12.75, df=2, 74, p=0.001). In contrast, a model including patient gender, sexual abuse history, and the four identity factors (R=0.73) accounted for 53.1% of the total variance (F=13.20, df=6, 70, p<0.001); this change was highly significant (F=10.24, df=4, 70, p<0.001). Comparable findings emerged when we used gender, sexual abuse, and the four identity factors to predict the presence or absence of the borderline diagnosis. The combined model (R=0.73) explained 53.1% of the variance (F=13.20, df=6, 70, p<0.001). The same was true when we used these variables to predict number of borderline symptoms excluding the identity disturbance criterion. Gender and sexual abuse alone in this analysis explained 25.0% of the variance, whereas adding the identity variables (R=0.67) explained an additional 20.0% (F=15.20, df=4, 75, p<0.001), a highly significant change (F=7.61, df=4, 70, p<0.001).
Identity Disturbance and Borderline Subtypes
As an exploratory analysis, we followed up on findings of two recent studies that isolated two distinct types of patients currently diagnosed with borderline personality disorder (24, 25), one with more dysphoric features and the other with more histrionic features. The first type (emotionally dysregulated) includes patients who have intense, painful, and poorly regulated emotions that they attempt to escape by using various maladaptive affect-regulatory strategies. The second (histrionic) type have emotions that are intense and dramatic but not very troubling to them; for these patients, dramatic emotions may even be self-defining.
To examine identity disturbance in patients who approximate the emotionally dysregulated type, we examined the partial correlations between ratings of borderline personality disorder and each identity disturbance factor, holding constant ratings of the extent to which the patient had histrionic features. (To maximize the comparability of dimensional diagnoses of borderline personality disorder and histrionic personality disorder, we used clinicians’ 1–7 ratings of each.) As predicted, this analysis indicated a very strong relationship (r=0.58, df=70, p<0.001) between severity of borderline personality disorder and the second factor, painful incoherence, after controlling for histrionic features. The only other significant partial correlation was with the fourth factor, lack of commitment (r=0.24, df=70, p=0.05).
Conversely, to examine identity disturbance in patients who approximate the histrionic type, we examined the partial correlations between ratings of histrionic personality disorder and each identity factor, holding constant borderline personality disorder ratings. For the first factor, role absorption, the partial correlation with the histrionic rating was significant (r=0.24, df=82, p=0.03). Strikingly, the second factor, which correlated so strongly with borderline personality disorder, showed a slightly negative correlation with histrionic ratings (r=–0.07, df=82, p=0.50). The last two factors correlated slightly positively with the histrionic rating (r=0.19, df=82, p=0.08; and r=0.13, df=82, p=0.28, respectively. Thus, some elements of identity disturbance appear more closely associated with histrionic than with borderline features, particularly role absorption, and, secondarily, inconsistency.
The aim of this study was to formulate a more precise conception of identity disturbance, particularly in patients with borderline personality disorder. The 35 items on our identity disturbance questionnaire discriminated 1) patients with and without identity disturbance across the entire cohort, and 2) patients with and without borderline personality disorder. Four factors emerged from the factor analysis, each encompassing a distinct facet of identity disturbance. The first factor, role absorption, describes overidentification with a specific role or group membership, such that a limited role or label defines the person’s whole identity. The second factor, painful incoherence, deals with patients’ subjective experience of their own identity. This factor conveys distress or concern about identity incoherence or lack of a coherence sense of self. The third factor, inconsistency, includes items such as "beliefs and actions often seem grossly contradictory." The final factor, lack of commitment, is a fundamental element of Marcia’s conception of identity (5, 6). This factor includes patients’ difficulties in committing to goals or maintaining a constant set of values.
All four factors were associated with clinicians’ present/absent ratings of identity disturbance, further corroborating validity of the measure. Painful incoherence was most highly associated with presence of identity disturbance, followed by the inconsistency factor, and then by the lack of commitment factor. The four factors together (R=0.61) predicted more than a third of the variance in presence/absence of the DSM-IV identity disturbance criterion (36.7%). When a continuous rating of identity disturbance was used, the four factors predicted almost half of the variance (47.3%).
Identity Disturbance in Borderline Personality Disorder
Each of the identity disturbance factors distinguished patients with borderline personality disorder from those with other personality disorders as well as those with no personality disorder. In general, subjects with borderline personality disorder had higher scores on all of the factors, which suggests that each type of identity disturbance is more severe in patients with borderline personality disorder than is seen in other nonpsychotic psychiatric disorders. Furthermore, the contrast analysis indicated a linear relationship, such that identity disturbance on each factor was greater for patients with borderline personality disorder than for those with other personality disorders, and greater for patients with other personality disorders than for patients without personality disorders.
Patients’ experience of their own identity incoherence is central to identity disturbance in borderline personality disorder; this factor was the most strongly related to borderline personality disorder in every analysis. Some theorists characterize patients with borderline personality disorder as being unaware or unconcerned about their own identity disturbances, whereas others describe these patients as being distressed by their lack of coherence. Our data support the latter point of view, although this may depend on whether the patient is more emotionally dysregulated or histrionic.
The weakest of the four factors in predicting borderline personality disorder was the fourth, lack of commitment. This may be an important finding, given the heavy emphasis most identity research (as well as DSM-IV) has placed on this construct. For example, DSM-IV describes identity disturbance in borderline personality disorder as being "characterized by shifting goals, values, and vocational aspirations" (p. 651). Erikson (2) and Marcia (6) both describe identity diffusion as being most commonly manifested in lack of commitments to career, religion, or values. Our data indicate that while this factor is a central component of identity disturbance and is somewhat elevated in patients with borderline personality disorder, it does not distinguish borderline personality disorder from other types of psychopathology. Lack of commitment may thus be a less specific index of identity disturbance that is related to multiple forms of psychopathology and not specifically to borderline personality disorder. To what degree this reflects peculiarities of our patient group, in which occupational instability did not distinguish patients with borderline personality disorder, is unclear, and hence requires replication.
Sexual Abuse, Borderline Personality Disorder, and Identity Disturbance
One of the goals of this research was to disentangle the role of sexual abuse history in borderline identity disturbance. In the present study, half of the borderline patients had a history of sexual abuse (in comparison with 11.5% of the subjects without borderline personality disorder), which allowed us to examine the relation between identity disturbance and borderline personality disorder while holding sexual abuse constant. The data suggest that sexual abuse contributes to only one aspect of borderline identity disturbance and does not account for all of the variance on even that aspect.
Many researchers have found a strong relationship between a history of sexual abuse and dissociative symptoms (28–30). Sexual abuse history was highly correlated primarily with the painful incoherence factor, and the item content of this factor suggests that sexual abuse history may play a role in the more dissociative aspects of identity disturbance.
Although the painful incoherence factor was strongly associated with a sexual abuse history, the model that best predicted subjects’ scores on this factor included both abuse history and borderline diagnosis. The patient’s painful concern about identity incoherence is not only the result of trauma; it appears to be integral to the nature of borderline personality disorder, whether or not the patient has an abuse history. Sexual abuse was largely uncorrelated with the other three identity factors, all of which are associated with borderline personality disorder.
Identity Disturbance and Borderline Subtypes
Previous research from our laboratory (24, 25) has found two distinct types of patients currently diagnosed with borderline personality disorder, one more distressed and emotionally dysregulated, and the other more histrionic. The secondary analyses in this study provide suggestive data on differences in the types of identity disturbance characteristic of each subtype. Controlling for histrionic features, only the second and fourth factors, painful incoherence and lack of commitment, were significantly associated with borderline features. In fact, each of these factors had a stronger association to borderline personality disorder ratings with histrionic features held constant. These factors appear to be more closely related to the emotionally dysregulated type. In contrast, when we held borderline features constant, the role absorption factor was significantly associated with histrionic ratings. The inconsistency factor appears to be associated with both kinds of patients, but particularly with the histrionic.
This study represents a first empirical attempt to home in on what identity disturbance in borderline personality disorder really means, but several potential objections require discussion. The first is the question of diagnostic reliability, given that we did not use structured interviews. Recent research suggests that even with study group sizes as small as 20 subjects, the central tendency that emerges when clinicians make categorical personality disorder diagnoses tends to be robust (23, 25). In addition, we measured the borderline diagnosis in four different ways. 1) Clinicians supplied a categorical diagnosis. 2) Clinicians indicated which DSM-IV criteria for borderline personality disorder were present in their patients. The list of symptoms was not identified as the criteria for borderline personality disorder and did not appear to deter many clinicians from rating five borderline personality disorder symptoms (including the identity disturbance criterion) as "present" in patients they had categorized as not having borderline personality disorder. 3) Clinicians rated the extent to which the patient showed features of each DSM-IV borderline personality disorder criterion. 4) Clinicians made a global dimensional rating of the extent to which the patient displayed symptoms of each of the 10 DSM-IV axis II disorders, including borderline personality disorder. In all cases, the borderline personality disorder group was clearly distinct from the other groups. No clinician who described a patient with borderline personality disorder endorsed fewer than five of the DSM-IV criteria; the average number exceeded seven. Secondary analyses that used level-of-functioning variables provided further evidence for validity. Finally, lower reliability among clinicians would foster type II, not type I, errors (i.e., null findings where positive findings are warranted). If clinicians were not diagnosing patients accurately, the borderline group would be more heterogeneous and thus less likely to show such robust differences from the other groups. The findings are even more striking given the less-than-optimal diagnostic reliability. Nevertheless, this is just an initial study, and future studies with more reliable diagnostic procedures are clearly warranted; one is currently underway.
A second potential criticism is that since we relied exclusively on clinician reports, we were not testing the nature of identity disturbance in borderline personality disorder but rather clinicians’ implicit assumptions about it. In part, of course, we were attempting to assess what clinicians mean by identity disturbance. We generated a set of 35 highly specific items, of which 28 distinguished patients diagnosed by clinicians with identity disturbance from those without. Four factors derived from this item set accounted for much of the variance in clinician ratings of the presence as well as the severity of identity disturbance, which suggests the construct validity of the instrument.
For several reasons, we believe these data provide meaningful information on the nature of identity disturbance in borderline personality disorder and do not simply reflect clinicians’ beliefs. First, all research relies on observation, and all observers have biases and intuitive theories. Most studies of psychopathology administer self-reports or structured interviews that ask patients to describe themselves and their psychopathology and then examine associations between these self-reported traits or symptoms and other self-reported variables. Our method is no different from this standard method, except that it uses expert informants rather than lay observers, for whom lack of insight into themselves is diagnostic. Given the subject of this study—identity disturbance—patients would likely have difficulty providing accurate information about their tendency to hold contradictory beliefs, their overabsorption in particular roles, and so forth. We thus chose to rely on skilled observers who knew the patients well and used an instrument that asked very specific questions, most of which called for only minimal inference. Respondents were clinicians with an average of 18 years of experience who had seen their patient for an average of 53 sessions; they were thus likely to know the patients well and to be able to recognize clinically significant patterns. Ideally, studies such as this should use a combination of self-reports, interviews, clinician reports, and reports by family members and significant others to triangulate on the findings. Future studies should clearly rely on data from multiple informants.
Second, and more important, shared theories could not have predicted the factor structure that emerged, the factors that correlated more strongly with borderline diagnosis, the factors associated with borderline personality disorder after controlling for sexual abuse, or the factors associated with particular subtypes of borderline personality disorder because there are no shared theories. The construct of clinical identity disturbance has been relatively ill-defined. No theory would have predicted the existence of four orthogonal factors in identity disturbance, or that subjective and objective inconsistency or incoherence would be uncorrelated with each other. These, we believe, may be important discoveries of this study.
Third, we did attempt to assess the effects of clinician bias by examining the relationship between factor scores and clinicians’ theoretical orientation and discipline. Holding borderline diagnosis constant, theory did not predict any of the scores on any factors. Because most of the clinicians who participated in this study reported a primary psychodynamic orientation, however, we also investigated the role of discipline (psychology, psychiatry, social work) in predicting factor scores. Discipline failed to predict scores on any factors in any diagnostic group, despite the fact that clinicians from different disciplines have markedly different training.
Fourth, dimensional and categorical diagnoses in this study produced identical findings. If clinicians were simply rating the 35 identity disturbance items on the basis of their beliefs about borderline personality disorder rather than on their actual knowledge of the patient, dimensional diagnosis would have produced much weaker findings than categorical diagnosis, since clinicians who described patients without borderline personality disorder would have systematically underdiagnosed identity issues. In fact, the regression analysis that predicted the number of borderline personality disorder symptoms from identity factor scores produced stronger findings than the comparable regression analysis that predicted categorical diagnosis.
Finally, as noted earlier, recent research suggests that when clinicians are asked to describe patients with various diagnoses, they do not tend to reproduce DSM-IV criteria or rely primarily on their intuitive prototypes. For example, when asked to rank order a list of 200 personality descriptors (which included DSM criteria) to describe a patient they were currently treating who had borderline personality disorder, clinicians in two studies did not tend to rank order the DSM-IV criteria the highest; rather, they painted a picture of borderline patients that tended to emphasize their subjective distress more than some of the more socially undesirable traits emphasized in DSM-IV (23, 25). Indeed, cluster analysis of these descriptions led to the discovery of two replicable types of patients currently defined as borderline who do not, empirically, appear to fall into a single diagnostic category. Similarly, in the present study, clinicians rated 64% of all patients—including over half of the subjects without borderline personality disorder—as having identity disturbance as defined in DSM-IV.