Participants and Procedures
Participants were 117 clinicians who constituted a random national sample of experienced psychiatrists and psychologists from the membership registers of the American Psychiatric Association and the American Psychological Association. Initial letters to clinicians described the study, presented them with the DSM-IV diagnostic criteria for borderline personality disorder and dysthymic disorder (which was selected as a comparison condition), and asked them to complete a postcard indicating whether they had at least one borderline personality disorder or dysthymic disorder patient in their practice who met the inclusion and exclusion criteria. Based on their postcard responses, we assigned the clinicians to describe either a borderline personality disorder patient (N=90) or a dysthymic disorder patient (N=27), again presenting them with the DSM-IV diagnostic criteria for borderline personality disorder and/or dysthymic disorder to ensure close attention to the diagnostic criteria. To ensure random selection of patients, we asked the clinicians who reported having more than one appropriate patient to consult their calendars and select the patient they saw most recently who met the study criteria. For the dysthymic disorder group, we asked clinicians to describe a current patient who met the DSM-IV criteria for dysthymic disorder and who had no diagnosable DSM-IV personality disorder and no more than three DSM-IV diagnostic criteria for borderline personality disorder. For patients in both groups, we asked clinicians to select a female patient (to avoid the confounding factor of gender and to maximize power, because 75%–80% of patients who receive a diagnosis of borderline personality disorder are female [20, DSM-IV]) between ages 18 and 55 years (to avoid the confounding factors associated with adolescent and late-life personality disorder diagnosis) whom they had seen for a minimum of eight sessions and a maximum of 2 years (to guarantee that they knew the patient well while minimizing the likelihood of substantial personality change in treatment) and who did not have a psychotic disorder. We asked clinicians to select a current psychotherapy patient to maximize the likelihood of their being able to provide detailed personality assessments. (We selected a comparison group of patients with dysthymic disorder because patients with depression have been the most common comparison group in studies of personality disorders, and patients with dysthymic disorder have enduring, moderate depression that is also common in patients with borderline personality disorder.) To maximize participation, we gave clinicians the option to participate by pen and paper or on our interactive web site (http://www. psychsystems.net). Consistent with the literature on computerized versus paper administration of questionnaires (35), we found no systematic differences between responses with the two methods.

Before analyzing the data, we excluded data on patients who were extreme outliers in age or length in treatment beyond the parameters we requested, data on patients who did not meet the diagnostic criteria for the borderline personality disorder or dysthymic disorder groups, and data suggesting extreme carelessness in responding (e.g., multiple pages not completed). To maximize power, however, we retained patients who exceeded within reasonable bounds the maximum limit for age (two patients whose ages were in the range of 55–61 years) and time in treatment (six patients whose time in treatment ranged from 25 to 48 months). Further, because several dysthymic disorder patients were one criterion short of the diagnostic criteria for the disorder or met the criteria for multiple personality disorders, we were faced with decisions about the “purity” of the dysthymic disorder sample. We ultimately chose to retain patients who had chronic depression if they were within one criterion of the dysthymic disorder diagnosis and to retain dysthymic disorder patients who met the DSM-IV criteria for a non-borderline-personality-disorder diagnosis (mostly avoidant and schizoid personality disorders) to maximize the number of subjects and the generalizability of the sample. The decision to include non-borderline-personality-disorder patients actually rendered findings more conservative and increased external validity, given the high rates of comorbidity (60%) for dysthymic disorder and personality disorders in prior research (36, 37). (In fact, we reran all analyses without the eight patients who did not meet the criteria, and significance values improved in three cases and decreased from 0.01 to 0.05 in one. However, to preserve consistency with other reports of data from this sample, we chose to avoid excluding these subjects for some analyses but not for others.)

Measures
Clinicians completed the following measures, presented in the following order. (We included other instruments for other studies but do not describe them here.)

Clinical Data Form
The Clinical Data Form was used to assess a range of variables relevant to demographics, diagnosis, adaptive functioning, developmental history, and family history of psychopathology. This measure was developed over several years and used in a number of studies (see reference 38). The sections of the Clinical Data Form that were relevant to this study ask clinicians to provide basic demographic data on themselves and the patient, as well as information pertaining to the patient’s diagnosis and adaptive functioning. Prior research found such ratings to correlate strongly with ratings made by independent interviewers (28, 33, 39).

SWAP-200
The SWAP-200 is a 200-item Q-sort designed to assess personality and personality pathology (e.g., references 22, 24, 27, 38). (A Q-sort is a set of statements printed on separate index cards, in this case, statements about personality and personality dysfunction.) An experienced clinical observer sorts the cards into eight piles, thereby assigning each of the 200 descriptive items a numerical score ranging from 0 (for items least descriptive of the patient) to 7 (items most descriptive of the patient). Items for the SWAP-200 were derived from a number of sources, including DSM-III-R and DSM-IV axis II criteria, clinical literature on personality disorders, research on personality disorders, research on normal personality traits and psychological health, pilot interviews, and the feedback of more than 1,000 clinicians. Development of the item set was an iterative process that followed standard psychometric methods, such as eliminating redundant items, items with minimal variance, and so forth. The Q-sort items provide a standardized clinical language that allows for clinicians’ assessments to be quantified, compared with those of other clinicians, and analyzed statistically.

Research thus far has supported the validity and reliability of the SWAP-200 in predicting numerous external criteria, such as suicide attempts and history of psychiatric hospitalizations, adaptive functioning assessed by measures such as the Global Assessment of Functioning Scale (GAF) from the DSM-IV, diagnoses based on interviews, and developmental and family history variables (e.g., references 25, 27, 34). The SWAP-200 has been used for taxonomic purposes in multiple studies (e.g., for empirically deriving personality diagnoses from large samples of adult and adolescent patients [references 27, 40]).