What does it mean to claim that someone falls into one of the diagnostic categories by which we hope to understand abnormal behavior? The discussion that follows will show how issues in the philosophy of language bear upon the classification enterprise. It will show how a categorical model of diagnosis falls short of providing the understanding we seek in the important field of psychopathology, and will show how a better model would be a dimensional one deriving from the notion that concepts are fuzzy sets.
The upshot of this discussion will be that meanings don't exist in quite the way that the authors of the Diagnostic and Statistical Manual of Mental Disorders, or DSM (American Psychiatric Association, 1994) seem to think they do. The authors of the DSM seem to think that knowing the meaning of a diagnostic category is just a matter of instantiating it (the assumption behind operationism). It is this assumption that is at stake in the DSM's categorical model of diagnosis, and I will argue against it, presupposing a realist philosophy of science in which it is possible to talk about truth and progress.
The concepts of intension and extension can be further illustrated using the phenomenon of comorbidity frequently found in the domain of psychopathology. Comorbidity is sharing by two disorders of the same extension. Assume hypothetically for a moment that all and only persons diagnosed with depression are diagnosed with anxiety as well. This is exactly what we would mean if we were to claim that these two disorders are completely comorbid, or that they are one-for-one concomitants of each other. In that case, the extension of the two terms is exactly the same.
But here is where the ambiguity in the meaning of meaning (as Putnam [1975] calls it) comes in: the terms "person with anxiety" and "person with depression" are not synonyms just because they share the same extension. The two terms have different meanings in exactly this sense: they have different intensions.
According to a tradition that has been widely accepted since Carnap, knowing the concept or intension of a term is just a matter of instantiating the term. Instantiation is similar to what Millon (1991) termed identification--the process of assigning previously unallocated entities to their appropriate categories.
Instantiation is a fundamentally different process depending upon the approach one takes to diagnostic terms. For the scientific realist, instantiation is a matter of discovery, and it is appropriate to speak of there being a truth to the matter whether an item correctly instantiates a category. For the conventionalist, by contrast, there is no truth involved, only a convention: this is how the language is used by speakers in this language community. While truth is rather stable, conventions change over time and among linguistic groups. So do the diagnostic categories in the DSM, now in its fourth edition.
Instantiating a class, for the scientific realist, is what happens when one recognizes something as a member of the class. To speak of recognition implies that there is, a priori, a truth to the matter. However, it does not imply that this truth is sharply defined: something may instantiate a category to a certain degree. Recognizing an object as a member of a category can take place on the basis of something as simple as a dictionary definition. Webster's defines schizophrenia is "a psychotic disorder characterized by loss of contact with environment and by disintegration of personality." It is important to note that one does not need a list of necessary and sufficient conditions in order to instantiate the category "schizophrenia," and that different persons at different times may instantiate this category only to a certain extent. Most terms in the natural language are what Engel (1989) calls matters of degree, terms having no clear-cut boundary between those items that instantiate them and those that do not.
The notion that extension is the set of things of which a term is true is not quite adequate to the explication of matters of degree, because a set in the mathematical sense is itself an all-or-none predicate: any given item either definitely belongs to S or definitely does not belong to S, if S is a set. However, as Putnam (1975, p. 217) says, "If one really wanted to formalize the notion of extension as applied to terms in a natural language, it would be necessary to employ 'fuzzy sets' or something similar rather than sets in the classical sense." A fuzzy set is a set that does not have clear boundaries, such that a given item can be a member of the set to a certain degree and only to a certain degree. For example, should a computer be considered furniture? How about a paperweight or bulletin board? Most would consider these items to be furniture to a certain degree and only to a certain degree.
The category of furniture is a human artifact category. Human artifact categories are to be distinguished from natural kinds. For human artifact categories, agreement among users in the linguistic community is the final criterion of the degree to which an item instantiates the category. For natural kinds, by contrast, the final criterion of the degree to which an item instantiates the category is agreement with nature. While natural kinds, such as the chemical elements, are discovered, human artifacts are invented.
Natural kinds capture deep underlying similarities among exemplars. Thus, if one knows something is a member of a natural kind, one can summon a wealth of information about it from one's knowledge of the natural kind. This is the case because natural kind terms are theory-laden (Quine, 1977; Putnam, 1975). The knowledge of experts is pertinent to deliberations as to whether a particular item instantiates a natural kind, but not to whether a particular item instantiates a human artifact category. Therefore, it makes sense to say, "According to experts, this is pure gold," while it doesn't make sense to say, "According to experts, this is a pencil" (Rothbart and Taylor, 1992; Schwartz, 1978). Scientific realism pertains to natural kinds, not to human artifacts.
I believe that, on a realistic assessment, we would find that persons instantiate the categories of psychopathology to a cerain degree and only to a certain degree. Consequently, I believe that the DSM taxonomy, which assumes the existence of classical, discrete sets of persons, should be replaced with a dimensional taxonomy based upon the notion that persons are arranged into fuzzy sets.
It is important to distinguish being a matter of degree from being vague. Being a matter of degree is enhanced by increased precision, while vagueness is diminished by increased precision. Something that is typically said to be not a matter of degree at all, an all-or-none predicate, such as being pregnant, has only two degrees. Something that has many degrees, on the other hand (and that is thus more precisely specifiable), such as temperature, is easily recognized as a matter of degree. Most terms in the natural language are both vague predicates and matters of degree, but being a matter of degree is not dependent upon vagueness.
The DSM approach attempts to establish the connection of concepts like anxiety, depression, and schizophrenia with their extension, which is people in the world who have these disorders (or who instantiate the categories). For logical positivists like Carnap, who accept the verifiability theory of meaning, the concept corresponding to a term provides a criterion for belonging to the extension of the term--not just in the sense of 'necessary and sufficient condition,' but in the strong sense of way of recognizing if a thing falls into the extension or not (Putnam, 1975). This assumption underlies what is called operationism. Operationism is defining a concept completely in terms of the operations or measurements used to recognize its instantiation. Operationism goes hand in hand with essentialism, which assumes that the meaning of a term is completely exhausted by a set of severally necessary and jointly sufficient conditions that define it. DSM taxonomy is underwritten by both operationism and essentialism.
Some philosophers and scientists have attempted to eliminate vagueness and make meanings strictly empirical by the use of atheoretical operational definitions. The first to do so was Bridgeman in physics, followed independently by Watson in psychology. Watson held that such mentalistic terms as "thirst" and "intelligence," if they are to be acceptable scientific terms, must be operationally defined by objective indices like time-lapsed-since-drinking and intelligence tests (Hull, 1968). According to this conception, intelligence just is what I.Q. tests measure. A similar approach is taken by the authors of the DSM, for whom schizophrenia just is what the DSM criteria single out as schizophrenia, although this category of persons is famously more heterogeneous than that of persons considered "normal." That the DSM falls into the use of operational definitions can be seen by its use of such minor threshold features as "at least one week" and "two of the following."
If the concept corresponding to a term provides a criterion for belonging to a class, as it was thought to on the traditional conception, then it should be impossible to have multiple extensions corresponding to a single intension. As I will show you, however, this is not the case in psychopathology. Two important concepts in psychopathology, depression and schizophrenia, have multiple extensions corresponding to a single intension. Consequently, in this field a single intension does not determine a single extension.
Consider the concept of depression. Depression is a syndrome (Horowitz, Malle, Knutson, Dryer, Nelson, and Person, in press) that comprises a single intension. However, we might readily identify two subsets of persons, both labeled "depressed" and both having comparable scores on a measure of depression, whose extensions show near-zero overlap. Persons who express a vegetative depression purely in somatic form (poor appetite, hypersomnia, low libido, bodily aches) might show very little overlap with a comparable group of persons who express their depression mainly through emotions (guilt, discouragement) and thoughts (self-criticism, hopelessness, rumination on negative events).1
Similarly, the term schizophrenia applies to a number of widely divergent symptom presentations, from schizoaffective disorder, in which a person shows signs of depressed mood as well as disorganized thought, to catatonic schizophrenia, in which a person maintains a rigid posture like a statue and refuses to move for long periods of time. If such widely divergent symptoms covaried with one another that would be impressive evidence that a valid category of psychopathology had been discovered. However, these divergent symptoms do not covary with one another: they are generally seen in completely different populations.
Meehl (1992, p. 118) states, "It is widely agreed by historians and philosophers of science that one of the respects in which post-Galilean science was superior to medieval science was the replacement of categorical, 'essentialist' ways of conceptualizing the world by quantitative, dimensional modes of thought" (see also Carson, 1996a, 1996b; Eysenck, 1986; Hull, 1965a, 1965b; Lewin, 1931). The idea that the "categories" of psychopathology have extensions that in fact blend seamlessly with one another follows a progressive shift in understanding and observation within science at large. Frances and Widiger (1986, p. 396) have confidently predicted that "dimensional approaches will gradually supplant the categorical in the classification of personality disorders" (see also Costa and Widiger, 1994; Pincus and Wiggins, 1990). This same idea could be expanded to include the DSM Axis I disorders (such as depression, anxiety, and schizophrenia), in which we again see a seamless blending of disorders one into the other. Schizoaffective disorder is a case in point: it seems to exist midway between schizophrenia and depression. Likewise, depression and anxiety share many overlapping features, leading to the proposal of a new category, mixed anxiety and depression (Clark and Watson, 1991), that might be better accounted for by abandoning the categorical assumption altogether and dealing with these disorders dimensionally.
Without the assumption of operationism, the intensions that comprise the DSM diagnoses could be mistaken. Contrary to what one might at first be predisposed to presume, however, fallibility is not only a good thing, it is also a critical feature of any scientific activity. So long as diagnostic criteria sets are fallible, the theories that incorporate them are still falsifiable. This, according to Popper, is what differentiates science from such other pursuits as logic, metaphysics, religion, ideology, and pseudo-science (Popper, 1959; Lakatos, 1970; Bartley, 1984). When operationism and essentialism preclude the possibility that a diagnostic criterion set can be nonreferential, then we are no longer dealing with a classification scheme susceptible of scientific investigation.
If we assume that extensions are arranged along continuous dimensions with fuzzy boundaries, then perhaps our conceptual framework would work better at reflecting these extensions if it were arranged the same way. This conjecture seems quite reasonable based upon a semantic definition of truth as an agreement between intension and extension that reflects the actual state of the world.
My arguments in this article have been motivated by a particular perspective that I see as an alternative to the DSM approach. The perspective that I uphold over against the DSM's operationism and essentialism is a brand of scientific realism. In particular, I want to defend as literally true the way of speaking in which we say that scientific research or clinical experience has discovered a particular disorder. Such a discovery would entail learning that the intensions of the diagnostic scheme, in the form of certain fallible indicators, mirror the phenomena actually found to be extensions in psychopathology. The DSM's assumption, by contrast, is that each indicator, in the form of a diagnostic criterion, is infallible, leaving no room for discovery.
There is something artificial about the DSM diagnostic criteria that makes DSM disorders less like discoveries and more like inventions. It is possible that in fact what are called out by such purely semantic relations do not exist, or have no extensions such as we would expect natural kind terms to have. Perhaps the category of schizophrenia, for example, when conceived as a single unitary construct, does not have any independent existence beyond its intension (Carson and Sanislow, 1993; Carson, 1991). This possibility can at least be articulated as an empirical hypothesis once we abandon the assumption of operationism. Increasing our ability to articulate falsifiable empirical hypotheses that enable an increasingly greater understanding of the real world beyond our conceptual structure is one of the principal goals of science. As Millon (1991, p. 245) observes, "The language we use, and the assumptions it reflects, are very much a part of our scientific disagreements."
Bartley, W. W., III (1984). The retreat to commitment, second edition. La Salle, Illinois: Open Court Publishing Company.
Carson, R. C. (1991). Dilemmas in the pathway of the DSM-IV. Journal of Abnormal Psychology, 100, 302-307.
Carson, R. C. (1996a). Aristotle, Galileo, and the DSM taxonomy: The case of schizophrenia. Journal of Consulting and Clinical Psychology, 64, 1133-1139.
Carson, R. C. (1996b). Seamlessness in personality and its derangements. Journal of Personality Assessment, 66, 240-247.
Carson, R. C., and Sanislow, C. A., III (1993). The schizophrenias. In P. B. Sutker and H. E. Adams (Eds.), Comprehensive handbook of psychopathology, second edition (pp. 295-333). New York: Plenum Press.
Clark, L. A., and Watson, D. (1991). Tripartite model of anxiety and depression: Psychometric evidence and taxonomic implications. Journal of Abnormal Psychology, 100, 316-336.
Costa, P. T., Jr., and Widiger, T. A. (Eds., 1994). Personality disorders and the five-factor model of personality. Washington, D.C.: American Psychological Association.
Engel, R. E. (1989). On degrees. The Journal of Philosophy, 86, 23-37.
Eysenck, H. J. (1986). A critique of contemporary classification and diagnosis. In T. Millon and G. L. Klerman (Eds.), Contemporary directions in psychopathology: Toward the DSM-IV (pp. 73-98). New York: Guilford.
Frances, A. J., and Widiger, T. A. (1986). Methodological issues in personality disorder diagnosis. In T. Millon and G. L. Klerman (Eds.), Contemporary directions in psychopathology: Toward the DSM-IV (pp.381-400). New York: Guilford Press.
Horowitz, L. M., French, R. deS., Lapid, J. S., and Weckler, D. (1982). Symptoms and interpersonal problems: The prototype as an integrating concept. In J. Anchin, and D. Kiesler (Eds.), Handbook of interpersonal psychotherapy. Oxford: Pergamon Press.
Horowitz, L. M., and Malle, B. F. (1993). Fuzzy concepts in psychotherapy research. Psychotherapy Research, 3, 131-148.
Horowitz, L. M., Malle, B. F., Knutson, B. D., Dryer, D. C., Nelson, K. L., and Person, E. A. (in press). A theory of the psychiatric syndrome: Evidence and implications for treatment and assessment.
Horowitz, L. M., Post, D. L., French, R. deS., Wallis, K. D., and Siegelman, E. Y. (1981). The prototype as a construct in abnormal psychology: 2. Clarifying disagreement in psychiatric judgments. Journal of Abnormal Psychology, 90, 575-585.
Horowitz, L. M., Wright, J. C., Lowenstein, E., and Parad, H. W. (1981). The prototype as a construct in abnormal psychology: 1. A method for deriving prototypes. Journal of Abnormal Psychology, 90, 568-574.
Hull, D. L. (1965a). The effect of essentialism on taxonomy: Two thousand years of stasis. The British Journal for the Philosophy of Science, 15, 314-326.
Hull, D. L. (1965b). The effect of essentialism on taxonomy: Two thousand years of stasis. The British Journal for the Philosophy of Science, 16, 1-18.
Hull, D. L. (1968). The operational impreative: Sense and nonsense in operationism. Systematic Zoology, 17, 438-457.
Lakatos, I. (1970). Falsification and the methodology of scientific research programmes. In I. Lakatos and A. Musgrave (Eds.), Criticism and the growth of knowledge (pp. 91-196). New York: Cambridge University Press.
Lewin, K. (1931). The conflict between Aristotelian and Galilean modes of thought in contemporary psychology. Journal of General Psychology, 5, 141-177. Reprinted in A dynamic theory of personality: Selected papers (pp.1-42). New York: McGraw-Hill, 1935.
Meehl, P. E. (1992). Factors and taxa, traits and types, differences of degree and differences in kind. Journal of Personality, 60, 117-174.
Millon, T. (1991). Classification in psychopathology: Rationale, alternatives, and standards. Journal of Abnormal Psychology, 100, 245-261.
Morey, L. C., and McNamara, T. P. (1987). On definitions, diagnosis, and DSM-III. Journal of Abnormal Psychology, 96, 283-285.
Pincus, A. L., and Wiggins, J. S. (1990). Interpersonal problems and conceptions of personality disorders. Journal of Personality Disorders, 4, 342-352.
Popper, K. R. (1959) The logic of scientific discovery. New York: Harper and Row.
Putnam, H. (1975). The meaning of 'meaning.' In H. Putnam, Mind, language and reality: Philosophical papers, volume 2 (pp. 215-271). New York: Cambridge University Press.
Quine, W. V. (1977). Natural kinds. In S. P. Schwartz (Ed.), Naming, necessity, and natural kinds (pp. 155-175). Ithaca, New York: Cornell University Press.
Rothbart, M., and Taylor, M. (1992). Category labels and social reality: Do we view social categories as natural kinds? In G. R. Semin and K. Fiedler (Eds.), Language, interaction and social cognition (pp. 11-36). London: Sage.
Schwartz, S. P. (1978). Putnam on artifacts. The Philosophical Review, 87, 566-574.