To the Editor: Using a large representative national sample, Mark Zimmerman, M.D., and Jill I. Mattia, Ph.D. (1), attempted to disconfirm a hypothesis my colleagues and I advanced and corroborated (2). I asserted that clinicians do not rely primarily on the direct-question format used in structured interviews to assess personality disorders (e.g., "Have you ever been told that you seemed like a shallow or superficial kind of person?" to assess histrionic personality disorder). Clinicians of all theoretical orientations have reported that they diagnose personality disorder pathology by listening to their patients’ narratives about significant interpersonal experiences and observing their behavior in the consulting room.

Drs. Zimmerman and Mattia hypothesized that if clinicians hear the results of structured interviews for borderline personality disorder (including information on particular criteria, such as self-mutilation), this influences their diagnoses—a hypothesis with which no one, myself included, would disagree (unless the clinicians have reason to distrust the data provided by the interviewer). In one-half of their sample, clinicians made diagnoses as they normally would, on the basis of an initial intake interview of unspecified length. In this situation, clinicians diagnosed only 0.4% out of 500 outpatients entering their clinic with borderline personality disorder. In the other one-half of the sample, researchers conducted structured interviews, then "presented the case to a psychiatrist who reviewed the findings of the evaluation with the patient," and then made a diagnosis. Under these conditions, 27 out of 59 patients diagnosed as having borderline personality disorder by structured interview were diagnosed with borderline personality disorder by a psychiatrist, and seven more were given a rule-out diagnosis of borderline personality disorder.

The investigators largely replicated a well-known psychological finding: that biasing observers with prior information (particularly a label) affects the way they subsequently view a person. What is perhaps more striking is that fewer than one-half of the psychiatrists who were biased in this way gave the patients the borderline personality disorder diagnosis! In the least, from a methodological point of view, the psychiatrist should have been presented with the results of the structured interviews after spending some time with the patients, preferably at least two or three sessions, so that the psychiatrist would have had time to assess the patients’ personalities.

The use of intake diagnoses recorded in the patients’ charts as an index of the clinician’s diagnostic impressions was also an unfortunate methodological choice. The authors report that only two patients out of 500 received an intake diagnosis of borderline personality disorder in the naturalistic condition. This is so far below the norms reported in any study of which I am aware that I agree with the authors’ casual observation that "clinicians were very reluctant to diagnose borderline personality disorder during their routine intake diagnostic evaluations" (1, p. 1572). And I applaud the clinicians’ reluctance: giving a patient a stigmatizing diagnosis in an official record available to third-party payers after seeing the patient only briefly (presumably spending most of the interview inquiring about axis I diagnoses, suicidality, and other issues relevant to triage) would be inappropriate in most circumstances. In a similar study in which clinicians made anonymous diagnoses (3), Arkowitz-Westen and I found that 14.5% of the patients received a borderline diagnosis.

Finally, the choice to study only one personality disorder, borderline, was unfortunate. As I noted in the article to which the authors were responding (2), borderline personality disorder and antisocial personality disorder are the two diagnoses for which structured interviews have the best validity and reliability data; that is precisely because they are the ones with the most objective behavioral indices (e.g., cutting, suicide attempts, and arrests). A more conservative test of their hypothesis would have been to select a disorder such as narcissistic personality disorder, for which criteria are not so accessible by direct questioning. Whether clinicians would have found diagnoses made after 5–10 minutes of direct questioning by interviewers (some of whom were only bachelor’s-level research assistants) compelling enough to accept one-half of the time, as they apparently did with borderline personality disorder, is not so clear.