How should a psychiatric nosology set the boundary between mental disorders and "nondisordered" psychiatric problems that occur commonly in human populations? Should a diagnostic system take a purely descriptive approach, outlining the key inclusion and exclusion criteria for important biobehavioral syndromes? Or should a diagnostic system require that a disorder have a significant impact on functioning (as described by a range of terms including disability, impairment, dysfunction, or disadvantage)? The following case vignettes illustrate the problem.
Mr. A is a 42-year-old single man with a middle-management position in an investment firm. He reports an extensive delusional system in which he receives special vibrations from Jupiter that bounce over a near-by satellite dish and inform him of "how things will go for me that day." However, Mr. A is functioning well at work and has a small circle of friends that he sees regularly. He realizes that most people think that his ideas are "funny," and he usually remembers not to talk about his ideas with others because "it sometimes gets me into trouble."
Ms. B is a 38-year-old schoolteacher who states that since early childhood, she has been irrationally afraid of snakes. She still gets an uncomfortable feeling when seeing pictures of snakes in a magazine or on television. However, aside from refusing to take her children into the snake house at the zoo last year, she is unable to recall any ways in which her fear has had any impact on her life. She lives in a city and has not seen a real snake for years.
I suspect that most clinicians would feel quite comfortable with the conclusion that Mr. A suffers from a psychiatric disorder. However, most would feel uncomfortable reaching the same conclusion about Ms. B. Why? Neither Mr. A nor Ms. B is significantly impaired by their symptoms. However, the nature of the psychiatric dysfunction in these two patients is quite different. Mr. A has lost his ability to accurately assess reality, whereas Ms. B is "just" afraid of snakes.
In DSM-III and DSM-III-R, modest attention was paid to the problem of setting boundaries for psychiatric illness. The descriptive approach was predominant, and specific reference to required dysfunction in the diagnostic criteria was limited to a few disorders. Although rarely read, the introduction to both editions did include statements about the definition of a mental disorder. Spitzer and Wakefield quote the relevant introductory section for DSM-IV, which is essentially unchanged from DSM-III-R. This all changed in DSM-IV. As reviewed by Spitzer and Wakefield in this issue of the Journal, clinical significance criteria were added to nearly half of all DSM-IV diagnostic categories. From the perspective of the criteria-oriented clinician or researcher, clinical significance criteria moved from a distal consideration to center stage. While the wording was variable, these clinical significance criteria most commonly required dysfunction or distress beyond that specified in the diagnostic criteria.
While this shift in focus was likely influenced by several factors, the relatively high rates of disorders reported in the major epidemiologic studies of psychiatric illness (the Epidemiologic Catchment Area Study  and the National Comorbidity Survey ) were particularly influential. Such high rates raised questions about the credibility of the diagnostic criteria and concerns that a substantial proportion of diagnosed cases were false positives. In addition, the clinical significance criteria, it was argued, simply made explicit what had already been implicit in DSM-III and DSM-III-R.
Spitzer and Wakefield review the application of the clinical significance criteria in DSM-IV and conclude that it was likely misguided. I will not repeat their arguments here but, rather, wish to make seven additional comments. First, the empirical basis for the addition of the clinical significance criteria was at the time of the development of DSM-IV and remains now quite limited. Most other changes to DSM-IV were guided both by literature reviews and field trials. This was not the case for the clinical significance criteria. Even their impact on caseness was not studied systematically. This is not to claim that the issues in evaluating empirically the clinical significance criteria were simple. They were not. What external measures (or validators) would be used to determine whether the addition of the clinical significance criteria reduced the rate of false positive diagnoses and at what cost in an increase in false negatives?
Second, Spitzer and Wakefield only briefly address the core conceptual issue behind the clinical significance criteria—is a syndromal diagnosis with no requirements for dysfunction or distress sufficient to define a diagnostic category? On the one hand, this is more the rule than the exception in nonpsychiatric medicine. For Tourette’s syndrome or psychosis not otherwise specified (e.g., Mr. A), it is hard to argue that individuals meeting clinical criteria should not be so diagnosed even if they are neither impaired nor distressed by their condition. On the other hand, some psychiatric symptoms are so common or normative in human populations (e.g., fear of snakes) that it seems hard to justify considering someone with those symptoms alone to be disordered. Is it possible that the requirement for additional dysfunction or distress might be appropriate only for certain psychiatric disorders for which the symptomatic criteria alone are insufficient? By what criteria might we decide whether the symptoms alone are sufficient to indicate a disorder? Our earlier examples suggest that symptoms which reflect a disruption of "fundamental" or "vital" psychological functions (e.g., reality testing) might alone constitute disorder. But if psychosis alone merits a designation of disorder, whereas an irrational fear must be accompanied by significant dysfunction, where, for example, does major depression lie on this continuum? The most strident discussion in the application of clinical significance criteria in the DSM-IV Task Force involved this diagnostic category. Are the symptomatic criteria (e.g., lowered mood, disordered appetite and sleep, loss of energy) alone sufficient to define an illness? (The reader who is interested in this important conceptual issue might wish to review the cogent article by Wakefield , who concludes that the concept of "disorder" implies and probably requires significant impact on the functioning of the organism.)
Third, both the framers of the clinical significance criteria and Spitzer and Wakefield appear to assume that dysfunction represents a good "validator" for "true" psychiatric disorders. This key assumption deserves thoughtful scrutiny. For example, several studies have shown that subsyndromal depressive symptoms are associated with substantial disability similar to that seen with common medical disorders (4–6). As Strauss and Carpenter proposed years ago (7), the linkage between symptoms and functioning is probably rather weak.
Fourth, as another possible approach to the problem of setting boundaries around mental illness, Spitzer and Wakefield suggest that the DSM concept of normal grief be expanded to a broader range of depressive and anxiety syndromes that occur in reaction to psychosocial stressors. Several lines of evidence indicate that this suggestion may be problematic. A quite large percentage of individuals with major depression in the general population report that the onset began in the setting of psychosocial adversity (8, 9). This proposal may then produce a marked change in base rates. Even when confronted with the most severe possible events (e.g., assault), a small minority of individuals develop major depression (10), raising critical questions about what is meant by a "normal" reaction to stressors. Finally, this approach suggests a return to an "etiologically" based approach to the diagnosis of depression and anxiety in which mental health professionals would be responsible for distinguishing between understandable reactions to stressors and endogenous disorders. While intellectually appealing, this approach has not been very successful in the past in psychiatry. It is not clear why it would be expected to fare better now.
Fifth, somewhat unexamined during the debate about clinical significance criteria was the assumption that because rates of psychiatric illness in the general population were so high, these data constitute prima facie evidence that the criteria were producing false positives. Might this reaction indicate, rather, the insecurity of our field vis à vis the rest of medicine? Would cardiologists question their definitions of hypertension or coronary artery disease because of their high population prevalence?
Sixth, Spitzer and Wakefield do not review the thorny issue of the attribution of dysfunction. While it is usually straightforward to assess social or occupational dysfunction, it is far more difficult to determine whether any observed dysfunction is "caused" by a given psychiatric disorder. (This is a significant problem because the clinical significance criteria in DSM-IV typically contain the phrase, "The disturbance causes clinically significant distress or impairment.") Given high levels of comorbidity between psychiatric disorders and between psychiatric and medical conditions, is it realistic to expect that we can assess reliably which disorder is responsible for which kind of disability? Many risk factors for psychiatric disorders are themselves associated with disability. How easy is it to determine whether dysfunction stems directly from the clinical diagnosis or results instead from prior existing risk factors such as a disturbed home environment or an abnormal personality? Last, Spitzer and Wakefield do not discuss the possibility that disorder-associated disability should be part of the diagnostic process, but on an axis separate from clinical diagnosis. The multiaxial DSM-IV system captures functioning on axis V, but there is no attempt to determine whether dysfunction is "due to" the diagnosed disorder or disorders. Is this a viable model for DSM-V?
What can we conclude from these discussions? It is unlikely that we will soon reach an internally consistent and comprehensive solution to the problem of defining the boundaries of psychiatric illness. A diverse set of factors affect this question, many of which are not fundamentally "scientific" in the way in which that term is commonly understood (11). Furthermore, many of our psychiatric syndromes may not exist in nature as highly discrete entities easily separable from subsyndromal conditions (12). Our questions about boundaries of psychiatric illness may not be solved definitively until we have a detailed understanding of the pathophysiology of the disorders that we treat.
Spitzer and Wakefield have done us a service in their thoughtful review of these problems. As we begin to move to DSM-V, it is probably best that we, as a field, confront this difficult problem head-on. We should clarify what empirical questions can be answered by further research and strive for a reasonable working definition which recognizes that the role of symptomatic criteria and disability in defining the boundary of mental illness may differ across diagnostic categories.
Address reprint requests to Dr. Kendler, Box 980126, Medical College of Virginia of Virginia Commonwealth University, Richmond, VA 23298-0126.