Science Watch

A study in the Journal of Experimental Psychology: General (Vol. 131, No. 4) suggests something that most clinicians probably already suspected: While the Diagnostic and Statistical Manual (DSM) may be atheoretical, clinicians are anything but.

Nancy Kim, PhD, who recently completed her doctorate at Yale University, and Woo-kyoung Ahn, PhD, a professor of psychology at Vanderbilt University, found that even after years of experience with the DSM, clinicians still use their own theories to help them decide whether a patient belongs to a particular diagnostic category.

Specifically, they found that the more central a symptom is to a clinician's theory, the more weight the clinician gives it when deciding whether a hypothetical patient should receive a diagnosis, and the easier it is for the clinician to remember the symptom later.

"It seems to be a very powerful pull for people to use their theories," says Kim. "The fact that this doesn't go away, even in experts, doesn't really surprise me."

Causal centrality

Participants in the study--35 clinicians and 25 clinical trainees--performed four kinds of tasks: drawing the relationships between symptoms of a disorder; judging the importance, or conceptual centrality, of symptoms; diagnosing hypothetical patients; and recalling symptoms several hours after a diagnosis.

Kim and Ahn found that both clinicians and clinical trainees held complex theories about how the symptoms of various disorders--anorexia nervosa, schizophrenia, major depressive episode, antisocial personality disorder and others--are related to each other.

They also found that clinicians and trainees were more likely to diagnose a hypothetical patient with a disorder when the patient's symptoms were causally central to their theories than when they were peripheral. And, several hours after the diagnoses, both groups were more likely to remember causally central symptoms. Diagnoses of highly familiar disorders, such as antisocial personality disorder, were more likely to be influenced by theories than diagnoses of unfamiliar disorders, such as schizotypal personality disorder.

The results suggest that no amount of experience with the DSM can overcome the influence of theories on diagnosis. But because the study was conducted using pen-and-paper questionnaires and hypothetical patients, conclusions about the relevance of the findings for clinicians' real-world behavior must remain tentative, the authors say.

Ahn notes that when she presented the findings to a group of clinicians, including some study participants, they pointed out that they usually make their formal DSM-based diagnoses using checklists. Those checklists might encourage them to follow the DSM guidelines more closely than they would under less formal conditions, such as those used in this study.

On the other hand, the effects of those formal diagnostic procedures could be counterbalanced by biases in memory or attention that might be even stronger in the clinician's office than in the laboratory. After all, says Ahn, "it's human nature to theorize about the world."

Because of these uncertainties, the authors are confident that the study accurately reflects clinicians' thought processes but remain cautious about applying its conclusions to real-world situations. "I think this does reflect some aspect of clinicians' cognition, but I wouldn't say that it reflects the entire process," says Kim.

The study could be used to support the arguments of those who are critical of the DSM's current format, says John Kihlstrom, PhD, of the University of California, Berkeley. He and others have suggested that the DSM should abandon its atheoretical stance and instead try to categorize disorders according to their underlying causes.

But more research must be done if that effort is to be successful, says Kihlstrom. "It's one thing to show, as Kim and Ahn do, that clinicians diagnose based on their theories of mental illness. That's just a fact of professional life, and now we know about it. But before we go off and revise DSM to be more theory oriented, we better be sure we have the right theory. And, for many mental illnesses, we're a long way from that."

Diagnosis as categorization

In addition to shedding light on how clinicians think about diagnoses--at least when they are in the laboratory--the study also makes an important contribution to research on categorization.

In the 1970s, when DSM-III was still on the drawing board, Eleanor Rosch, PhD, and other cognitive psychologists showed that the classical view of categories--as lists of required features--was a poor reflection of human reasoning. They suggested that categories should instead be described as fuzzy sets, or as groups of objects united by "family resemblances."

Then, in 1980, Nancy Cantor, PhD, reported that clinicians' diagnoses actually followed the fuzzy model. DSM-III, the first edition of the DSM to define clinical disorders in terms of fuzzy categories, was published in the same year.

"Diagnosis is an act of categorization, and as our understanding of categorization has evolved, our understanding of the diagnostic process has evolved right along with it," says Kihlstrom.

In the 1980s, a third theory of categorization was proposed: the so-called "theory theory." The theory theory suggests that the categories we use to structure the world are shaped by our theories about how the world works.

The current study provides support for the theory theory by showing that clinicians' theories influence their diagnoses. But it does not rule out the possibility that other kinds of categorization could also be at work. In fact, says Kim, there is a growing consensus that models of categorization that combine theory-based and "fuzzy" similarity-based approaches might best capture the complexities of human reasoning.

Much research remains to be done on when and how different kinds of categorization take place. Until this study, for instance, no one had shown that the theory theory could be applied to real-world experts trained in using explicitly atheoretical categories. Clinical psychology provided a unique opportunity to explore that question.

"What has not been shown up until recently is exactly how background knowledge affects categorization," says Ahn. "We found that these domain theories consist largely of causal relations, and that the causal status of the symptoms determines how important those symptoms are for diagnosis."