Neopositivism and the DSM psychiatric classification. An epistemological history. Part 1:  Theoretical comparison

Introduction

In view of the publication of the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders, researchers were solicited to contribute to the literature basis for the revision process (Kupfer, First and Regier, 2002). Among the new issues proposed in the research agenda for the DSM-5, a notable one was the significant space dedicated to the discussion of theoretical questions. This is prima facie surprising, considering that the need to explore the methodological background of psychiatric diagnosis was not felt to be urgent in previous DSM revisions, due to the general ‘atheoretical’ claim that characterized DSM-III and subsequent versions. However, it is now acknowledged that many controversies on the classification of mental disorders are based on the choice of different theoretical models, and that the current crisis of the DSM is related to paradigmatic anomalies that are the by-product of the way the system itself is organized (Aragona, 2009a). Accordingly, it was suggested that the emergence of anomalies and the way the system tried to handle them were best understood if the fundamental role of the implicit philosophical assumptions in the ‘atheoretical’ DSM was considered (Aragona, 2009a). Further analysis showed that the DSM concept of ‘mental disorder’ was two-faced (Aragona, 2009b). On one side, the DSM explicit definition of ‘mental disorder’ generically talks of a clinically significant behavioural or psychological syndrome or pattern occurring in an individual. On the other side, there is an implicit definition emerging from its concrete use and being characterized as follows: (a) use of a ‘purely’ descriptive approach; (b) use of explicit ‘operative’ diagnostic criteria; (c) lack of a hierarchical ordering of symptoms; (d) use of polythetic criteria with quantitative diagnostic thresholds (Aragona, 2009b). The history of the DSM theoretical shifts, starting from the analysis of the first edition, shows that the most important innovation of the DSM-III was the introduction of a formalized implicit definition of mental disorders, in particular the use of operative, explicit criteria, which prima facie appear to be based on a neopositivist epistemology (Aragona, 2010).

This study explores the epistemological and historical relationship between the construction of the DSM-III and the neopositivist assumptions about scientific classification. Indeed, since the publication of the ‘atheoretical’ DSM-III (APA, 1980), scholars have repeatedly stressed the implicit influence of empiricist and neopositivist epistemologies on the development of its basic ideas about taxonomy (Faust and Miner, 1986; Fulford and Sartorius, 2009; Malmgren, 1993; Marquis and Douthit, 2006; Malt, 1986; Schwartz and Wiggins, 1986). Considering that in critical research, neopositivist ideas are rarely explicated in detail, the first section of the present paper will introduce the reader to the main tenets of the neopositivist movement.

Following this brief introduction to neopositivism in general, the second section will present Carl Gustav Hempel’s ideas about psychiatric taxonomies. His views will be discussed extensively because his 1959 lecture at the Work Conference on Field Studies in the Mental Disorders held in New York, under the auspices of the American Psychopathological Association (Hempel, 1961, 1965), is credited as the fundamental theoretical contribution to later developments in psychiatric taxonomy. In the same year the British psychiatrist Erwin Stengel proposed his recommendation to the World Health Organization (WHO) on how to revise the psychiatric section of the International Classification of Diseases (ICD), with acknowledgments to Hempel’s contribution (Stengel, 1959). Later, another participant in the discussion after Hempel’s speech, Sir Aubrey Lewis, played an important role in the development of ICD-8 (Fulford and Sartorius, 2009). In his influential book on psychiatric classification, R.E. Kendell (1975) explicitly endorsed Hempel when discussing operational definitions in psychiatry, although he concentrated on Hempel’s ‘liberalization’ of operational definitions without considering other relevant suggestions of the neo-empiricist philosopher (Malmgren, 1993; Schwartz and Wiggins, 1986). Finally, in the introduction of the DSM-III (APA, 1980), Spitzer briefly described some scientific and administrative motives that led to the DSM-III and emphasized important methodological innovations like the introduction of explicit diagnostic criteria (the operative diagnostic criteria) and the adoption of a descriptive approach intended to be neutral with respect to theories about aetiology. However, Hempel was not cited as a source of these innovations, raising the question whether Hempel (or any other neopositivist philosopher) had any role at all in the development of the DSM-III. The absence of a direct quotation is not proof that Hempel was not influential, and it may be because there are few explicit acknowledgements of the sources in the DSM-III introduction. Accordingly, Spitzer’s introduction was severely criticized for having ‘failed to acknowledge in any very determinate way its own historical or philosophical underpinnings’ (Sedler, 1994: 220). Regardless of the absence of a direct quotation of Hempel or other neopositivists, the DSM-III is largely credited to be a neopositivist creature (Faust and Miner, 1986; Malmgren, 1993; Marquis and Douthit, 2006; Malt, 1986; Schwartz and Wiggins, 1986).

Previous scholars focused only on specific aspects of the neopositivist influence on the DSM-III, selecting those parts of Hempel’s lecture which were more appropriate to show that their application in the DSM-III was mistaken. In the present paper Hempel’s lecture will be discussed more comprehensively than in previous studies, focusing on what Hempel really said at the 1959 research meeting and analysing the structure of his talk in its proper historical context.

Following a detailed discussion of Hempel’s general neopositivist ideas and his suggestions for psychiatric taxonomy, this paper will deal with the following question: can a theoretical comparison between the DSM-III and the basic neopositivist assumptions show a clear neopositivist structure of the DSM-III?

Basic neopositivist ideas

The term neopositivism broadly refers to a philosophical movement which began in the early twentieth century in Europe and later spread to the UK and the USA. There, neopositivist ideas deeply influenced not only the debate on the philosophy of science but philosophy in general, being the fountainhead of the tendency to see philosophy as an analytical inquiry.

Depending on the place and on the period, philosophers holding neopositivist views used different names to define their epistemological orientation, so neopositivism is also called logical positivism, logical empiricism and neo-empiricism. The change of name often reflects theoretical changes on some points, but the unity of these reflections under the neopositivist stance still remains.

Neopositivism in German-speaking countries arose as a reaction to Hegelian and neo-Hegelian metaphysics, which was strongly influential at that time. The neopositivist movement was inspired by scientifically-oriented writings published around the turn of the century in Continental Europe as well as in England. The phenomenalist positivism of Ernst Mach, the logistic of Gottlob Frege (with his emphasis on the logical-linguistic distinction between sense and reference of sentences) and of Bertrand Russell (with his theory of types and logical atomism), the mathematical logic of Russell’s and Whitehead’s Principia Mathematica, and the young Ludwig Wittgenstein’s Tractatus Logico-Philosophicus were all very influential among members of neopositivist circles.

Although a first nucleus can be found in the so-called ‘First Vienna Circle’, which was a discussion group meeting at the Café Central before World War I, the neopositivist school was formally instituted with the arrival of Moritz Schlick at the University of Vienna in 1922. Schlick’s Vienna Circle, along with Reichenbach’s Berlin Circle, propagated neopositivist doctrines all around Europe in the 1920s and early 1930s. The philosophies of A.J. Ayer and Gilbert Ryle are examples of the deep influence of the neopositivist circles on British philosophy.

The original Vienna Circle dispersed in 1938 after the Nazis took control of Austria and the deaths of Hans Hahn and Moritz Schlick. The most prominent members of neopositivism emigrated to the USA and the UK, significantly influencing the cultural and scientific debate in those countries. Among them there were Rudolf Carnap and Carl Gustav Hempel, whose neo-empiricist contributions became so influential in America that some of their views began to be considered as the ‘received views’.

Neopositivism has sometimes been seen as a school of philosophy with unanimity of purpose and creed. However, it would be better considered as a general stance, as an anti-metaphysical and scientifically-oriented way of doing philosophy, with a methodological emphasis on the logical analysis of language. Accordingly, some basic ideas were asserted by many (but not necessarily all) members of the neopositivist circles, and after the diaspora of the late 1930s each generation of philosophers with a neopositivist frame of thought gave a new form to those ideas by selecting and/or criticizing some of them in particular. The following is an incomplete list of the main assumptions of neopositivist-oriented scholars.

To sum up, the neopositivist’s image of scientific theories and activity is strictly intertwined with an emphasis on logic, meaning and the analysis of language, and it traces a sharp distinction between the theoretical and the observational levels. It is only the last of these, the reign of directly observable facts, which guarantees the testability of the experimental empirical evidences from which theoretical generalizations derive their meaningfulness. In the neopositivist mainstream the procedure is bottom-up, all meaningful knowledge being based on logical inference from ‘simple’ protocol sentences grounded in observable facts (theories emerge as inductive generalizations from observations). Popper (1934) reversed this direction, considering the creation of theories a non-analysable procedure and focusing on a top-down activity, deriving testable consequences from theories that at the observational level should prove to resist falsification. It is noteworthy that although they are quite opposite, both classic neopositivist views and Popper’s falsificationism share the basic tenet of ‘a pretty sharp distinction between observation and theory’ (Hacking, 1983: 5), ¹ the observational level being the site where the scientific nature of theories is guaranteed. ²

Hempel’s suggestions to implement scientific classifications in psychiatry

In his influential lecture to the American Psychopathological Association, Hempel states that his paper will attempt to provide a systematic discussion of the taxonomy of mental disorders, starting from the basic logical and methodological aspects of the classificatory procedures used in various branches of the empirical science and then suggesting some implications ‘for the taxonomic problems of psychiatry’ (Hempel, 1965: 137). The organization of the five sections of the paper follows coherently. In ‘Classes and concepts’, Hempel describes the general requirements for any classification, whatever its objects may be (‘concrete things such as stars, crystals, organisms, books, and so on; or they may be abstract entities such as numbers, kinship systems, political ideologies, religions, or philosophical doctrines’; pp. 137–8). According to Hempel, a classification divides a given set of objects (the ‘universe of discourse’) into subclasses containing ‘the elements or members of the given set’ (p. 137). He traces a distinction ‘between concepts and the terms that stand for them; for example, the term “soluble in alcohol” which is a linguistic expression, stands for the concept of solubility in alcohol, which is a property of certain substances’, and underlines that what really matters in a classificatory system are concepts: ‘the establishment of a suitable system of classification in a given domain of investigation may be considered as a special kind of scientific concept formation’. Concretely, each class in the system ‘consists of just those objects in the universe of discourse which possess the specific characteristics which the concept represents’ (p. 139), and these characteristics are the ‘necessary and sufficient conditions of membership in it, i.e., by stating certain characteristics which all and only the members of this class possess. Each subclass is thus defined by means of (more precisely, as the extension of) a certain concept, which represents the complex of characteristics essential for membership in that subclass’. Psychiatric classification is mentioned early in this paper; Hempel notes that the objects of psychiatric classification ‘are not the various kinds of mental disorders, but individual cases, which are assigned to various classes according to the kinds of mental disorder they exemplify’. Moreover, he remarks that albeit similar to the other biological classifications, the diagnostic classifications differ in the time span considered; while the elements of biological taxonomies are ‘best considered to be individual organisms during their total life span’, the individual case classified in psychiatric taxonomies ‘is best understood to be a particular human being at a given time, or during a given time span, in his life history’(p. 138). It is worth noting that, for Hempel, the different kinds of mental illness are ‘conceptually’ distinguished in the system; that is, the same universe of discourse can be divided in classes (the mental disorders) in ‘alternative ways’, depending on ‘the use of alternative sets of concepts in singling out similarities and differences among the objects under consideration’ (p. 139).

After this discussion of what should be intended for classification and the role of classificatory concepts, Hempel’s next section, ‘Description and theoretical systematization as two basic functions of scientific concepts’, borrows from Northrop (1947) the idea that science progresses:

from an initial ‘natural history’ stage, which primarily seeks to describe the phenomena under study and to establish simple empirical generalizations concerning them, to subsequent more and more ‘theoretical’ stages, in which increasing emphasis is placed upon the attainment of comprehensive theoretical accounts of the empirical subject matter under investigation. (Hempel, 1965: 139–40)

Hempel calls ‘empirical import’ the property of scientific terms to have clear and objective criteria of application, and ‘systematic import’ their contribution to the systematization of knowledge in a given field by means of laws or theories: ‘to be scientifically useful a concept must lend itself to the formulation of general laws or theoretical principles which reflect uniformities in the subject matter under study, and which thus provide a basis for explanation, prediction, and generally scientific understanding’ (p. 146). Accordingly, in Hempel’s view taxonomy progresses from an initial, descriptive stage to subsequent systems based on theoretical concepts: ‘This process is illustrated, for example, by the shift from an observational-phenomenal characterization and classification of chemical elements and compounds to theoretical modes of defining and differentiating them by reference to their atomic and molecular structures’ (pp. 148–9). He clearly stresses that such a progression had already taken place in biology, where early taxonomic systems based on morphological characteristics came to be replaced by systems with a phylogenetic basis, based on the theory of evolution (p. 147). Interestingly, in his opinion the psychiatric classification of the time (the first edition of the DSM) showed the same trend, the ‘concepts determining the various classes or categories’ being ‘no longer defined just in terms of symptoms, but rather in terms of the key concepts of theories which are intended to explain the observable behavior, including the symptoms in question’ (p. 149). However, from the subsequent part of the lecture we understand that being at the theoretical stage is not a guarantee unless other general requirements of a scientific classification are satisfied. Hempel stresses that ‘Science aims at knowledge that is objective in the sense of being intersubjectively certifiable, independently of individual opinion or preference, on the basis of data obtainable by suitable experiments or observations’. Accordingly, a necessary requirement is ‘that the terms used in formulating scientific statements have clearly specified meanings and be understood in the same sense by all those who use them’. It is on this basis that a crucial critique to the psychodynamically oriented DSM-I classification is formulated. If, in order to be scientific, the used terms must be clearly specified and intersubjectively uniformly understood, the central concepts of ‘various types of contemporary psychodynamic theories’ like those used in the DSM-I ‘lack clear and uniform criteria of application, and […], as a consequence, there are no definite and unequivocal ways of putting the theories to a test by applying them to concrete cases’ (p. 141). This same critique will be later applied to a classical concept of the European psychopathology, such as the ‘praecox feeling’ (p. 142), and to contemporary operational reformulations of the psychoanalytic concepts (Ellis, 1956), which provide ‘no clear criteria of application for the terms of psychoanalysis and no objective ways of testing psychoanalytic hypotheses’ (Hempel, 1965: 144).

Having defined the problem in this way, the concrete solution is a method that exactly specifies the meaning of a scientific term in a way that is objective and reliable. This is:

the use of so-called operational definitions for scientific terms. The idea was first set forth very explicitly by the physicist P.W. Bridgman in his book, The Logic of Modern Physics (1927). An operational definition of a given term is conceived as providing objective criteria by means of which any scientific investigator can decide, for any particular case, whether the term does or does not apply. To this end, the operational definition specifies a testing ‘operation’ T that can be performed on any case to which the given term could conceivably apply, and a certain outcome O of the testing operation, whose occurrence is to count as the criterion for the applicability of the term to the given case. (Hempel, 1965: 141)

Hempel immediately continues with three examples of concrete application of operational definitions. The first is taken from mineralogy (the operationalization of ‘the term harder than’); it should show that operationalism is not restricted to quantitative measures but it can also be applied to qualitative measures. The second is the most classical example of operational definition (that of length through ‘appropriate use of measuring rods’). The third example should show how it can be applied to medicine: the operational definition of phenylpyruvic oligophrenia by reference to the operation of ‘chemically testing the urine of the person concerned for the presence of phenylpyruvic acid’ (p. 141). Although this last example has to do with psychiatry, the most typical psychiatric diagnoses cannot be operationalized in this way (aetiology being unknown), and it is probably for this reason that Hempel does not add here a fourth example taken from classic psychiatric syndromes. On the contrary, he shifts the focus from an example of validity like that of phenylpyruvic oligophrenia to the restatement of the importance of reliability, and then to what he calls ‘certain qualifications’ of the customary formulation of operationalism that are required to allow the application of operationalism to psychiatric taxonomy. The first one is that the criteria of application for an operational definition do not cover the full definition: for example, a mercury thermometer is an appropriate instrument for the operative measurement of temperature only in a restricted range, other means being needed for very high or very low temperatures. The second qualification is more relevant for psychiatric nosology: if operationalism has to be applied to clinical activity, the operational definition of symptoms ‘is not to be unduly restrictive’ and so it should ‘not require manipulation of the objects under consideration’; in this case, ‘the idea of operation has to be taken in a very liberal sense’ and ‘the mere observation of an object […] must be allowed to count as an operation’ (p. 143). Here Hempel uses as an example the checklist of the descriptive characteristics of endomorphy (e.g. roundness and softness of the body), stressing that they ‘can be ascertained without any testing procedure more complicated than direct observation’, and restating that this procedure is satisfying because it guarantees ‘a reasonably precise meaning’ and ‘high intersubjective uniformity’ of usage. This is probably the most significant passage of Hempel’s talk, because it is here that he introduces a significant shift from Bridgman’s operationalism to a liberalized form of operationalism in which concrete operations are no more needed and direct observation counts as an operation provided that it is detected reliably.

Immediately after this second qualification, Hempel notes that it would be unreasonable to ask for an operational definition of all the terms used, because this would lead to an infinite regress. For this reason, ‘some terms must be antecedently understood’ (Hempel, 1965: 144) and, pressed in the discussion by Dr Gruenberg who was asking how these terms are defined, he replied by talking about ostension (Stengel, 1961: 35). The picture is now clear enough: in Hempel’s model, directly observable behaviours, those that can be reliably ascertained ostensively, are the basic stones of the operational definition of diagnostic entities. He admits that researchers may differ in their interpretation of the meanings of the terms, and so it would ‘be unreasonable and self-defeating to insist on the highest standards of precision from the beginning; but it is important to aim at increasingly reliable criteria of application for the various categories distinguished in a classification of mental disorders’ (Hempel, 1965: 145). In doing so, publicly observable behaviours have to be preferred while introspective, subjective and valuational aspects, providing no clear criteria of application, should be avoided. However, it is significant that Hempel concludes this section by stating that the objectivity under discussion is in any case a matter of degree, because:

the results of such ‘operations’ as observing an object by microscope or telescope, or a lung via fluoroscope or indirectly through an X-ray photograph, also show intersubjective variation even among expert observers. What matters is, I think, to be aware of the extent to which subjective factors enter into the application of a given set of concepts, and to aim at a gradual reduction of their influence. (p. 146)

Hempel’s next section is devoted to the relationship, discussed above, between the empirical and the systematic import. Another point of this section is also relevant for the present discussion. Being related to the general laws that provide a basis for scientific explanation, one would expect Hempel to have based the concept of ‘natural’ classification on causality (in the medical field, on the aetiological basis). Quite the opposite – he suggests that the distinction between natural and artificial classifications is based on pragmatic (‘classifications that are scientifically fruitful and those that are not’; p. 146) and statistical criteria: ‘in a classification of the former kind, those characteristics of the elements which serve as criteria of membership in a given class are associated, universally or with high probability, with more or less extensive clusters of other characteristics’ (pp. 146–7). A few lines later Hempel (p. 147) writes that the characteristics may be associated ‘by general laws or by statistical connection’. However, from his examples it is clear that the direction is not from the already known causal chains to their phenomenal effects (as it would be the case, for example, of an aetiologically based classification), but from the statistical connection of phenomenal characteristics to the induction of general laws that should explain these regularities. The conclusion is that in order to be objective (‘carving nature at the joints’) a classification does not necessarily need to be based on already known causal factors, a large number of statistically interconnected traits being sufficient. On the contrary, it is ‘artificial’ when the defining characteristics are arbitrary and, although it might be useful for special purposes, it has ‘few explanatory or predictive connections with other traits’ (p. 147). This is probably the reason for Hempel’s stress on the reliability of observable characteristics which should form the first basis for the classification of putative natural kinds. Only subsequently, with the advancement of theoretical knowledge, will the classificatory criterion change and the original classification be replaced by ‘systems based on theoretical concepts. This process is illustrated, for example, by the shift from an observational-phenomenal characterization and classification of chemical elements and compounds to theoretical modes of defining them by reference to their atomic and molecular structures’ (pp. 148–9). As mentioned above, in Hempel’s time, the DSM-I was already theory-based. Hempel acknowledges this fact and adds:

It is very likely, I think, that classifications of mental disorders will increasingly reflect theoretical considerations. It is not for me to speculate on the direction that theoretical developments in this field may take and especially on whether the major theories will be couched in biophysiological or biochemical terms or rather in psychodynamic terms that lack an over-all physiological or physiochemical interpretation. Theoretical systems of either kind can satisfy the basic requirements for scientific theories. (p. 150)

If we follow this line of reasoning coherently, we should assert that the transition from the basic, descriptive stage (empirical import) to the higher-level theoretically-based stage (systematic import) had already taken place in psychiatry and further developments were awaited in this direction. However, it is here that Hempel inserts a radical stop, although he does not directly attack the ‘psychodynamically-oriented’ DSM-I, but the neo-vitalistic conception of vital forces and entelechies. At a first glance, in the final part of this section Hempel simply outlines his main ideas about the basic requirement for scientific theories; these are: (a) a clear specification of the basic concepts used to represent the theoretical entities; (b) a set of laws/hypotheses asserting certain interrelations among the theoretical entities; (c) the ‘empirical interpretation’ of the theory, connecting the theoretical level with observable phenomena; (d) the testability of the theories through confirmation/disconfirmation of theory-derived ‘definite assertions about observable phenomena that should be found to occur under specifiable test conditions if the theory is correct’ (p. 151). Here the conclusion is that if a theory has no such implications, it ‘cannot qualify as a significant theory in empirical science (not even as an unsound or false one: for these latter attributes presuppose a conflict between the theory and relevant experimental or observational evidence)’. As stated, Hempel’s example is the neovitalist theory, but it is clear that it applies also to the psychodynamic theories that in his view were at the base of the DSM-I. In fact, the ‘empirical interpretation of the theory’ criterion (set out above) ‘might take the form of operational criteria for the theoretical terms or, more generally, the form of a set of laws, statistical or strictly universal in character, connecting the theoretical traits, states, or processes with observable phenomena’ (p.151). To sum up, Hempel asserts that: (a) the terms used in the DSM-I classification system (e.g. those defining a conversion reaction) ‘refer neither to directly observable phenomena, such as overt behavior, nor to responses that can be elicited by suitable stimuli’ (p. 140), and (b) there are no definite ways of putting psychodynamic theories ‘to a test by applying them to concrete cases’ (p. 141). On this basis he can conclude that the DSM-I mental disorders did not meet either the empirical interpretation criterion or the testability principle, and for this reason the DSM-I taxonomy should be rejected not as false but as meaningless. It is not known why Hempel did not restate it explicitly as the logical conclusion of this section. As this was a talk on psychiatric classification, he should have done so, but he preferred to conclude by attacking the already dead vitalistic biology and not to discuss the sentence, quoted above, about the possibility that either biological and psychodynamic theories may satisfy the basic requirements for scientific theories.

Hempel’s section ‘From classificatory to comparative quantitative concepts’ should also be mentioned. Here he tries to anticipate and suggest future fruitful changes in the concepts and theories of mental disorders. As far as I know, this section has never been discussed in detail in previous research. For reasons of space I cannot discuss it here in detail, but it should at least be remarked that it is an early critique of the categorical approach in psychiatric nosology ³ and a proposal for dimensional diagnosis. Considering the importance of the debate on ‘categorical vs. dimensional diagnosis’ in recent years, further studies of Hempel’s position on this matter would be useful.

The neopositivist assumptions of the DSM: a theoretical comparison

In order to reply to the question of a possible neopositivist structure of the DSM-III, a theoretical comparison is needed to understand to what extent there is a similarity connecting the basic tenets of the neopositivist school, Hempel’s lecture on psychiatric taxonomy and the concrete way the DSM-III and subsequent revisions are organized.

As seen above, the basic ideas shared by many thinkers of the neopositivist movement were: (a) the need of a demarcation criterion distinguishing scientific sentences from non-scientific, meaningless ones; (b) the view that empirical sentences are true if they correspond to an ascertainable state of facts; (c) a sharp distinction between the theoretical level, which is the plan of theories and theoretical entities (the non-observable entities postulated by the theory), and the observational level. On the latter level the phenomena can be observed and (possibly) measured, and the rightfulness of the theory can be tested (e.g. by experiments); (d) the presence of rules of correspondence and bridges laws connecting the purely theoretical, unobservable parts of the theory to observation of the circumstances under which the testable propositions of the theory are true.

The operational definitions of the scientific terms are not, strictly speaking, part of a neopositivist thesis. However, if one of the main aims of the neopositivist scientist is to test the empirical conditions in which the predictions of the theory can be confirmed, the need for clear definitions of what is to be observed and of how it has to be measured is a key part of the enterprise. Accordingly, it is neither strange nor conflicting that Hempel introduces the operational definition of scientific terms as a basic tool for achieving the aim of a scientific taxonomy.

As already mentioned, Hempel lectured on many themes related to the scientific (in general) and psychiatric (in particular) taxonomies. In general, his legacy to the neopositivist thesis is quite clear if we consider that: (a) classifications of mental disorders have to be grounded on theoretical considerations that must satisfy the basic requirements for scientific theories, otherwise they cannot qualify as significant; (b) definite observable phenomena have to occur under specifiable test conditions if the theory is correct; (c) the theoretical level (basic concepts representing the theoretical entities, basic laws, fundamental hypotheses) and the observational level are distinct; (d) an ‘empirical interpretation’ is needed to connect the theoretical and the observational levels.

Having shown a continuity between the neopositivist main tenets and Hempel’s neo-empiricism, it remains to be considered if, and to what extent, the DSM-III should be considered as a neopositivist system. At first glance the DSM-III’s atheoretical assumption conflicts with Hempel’s description of the progress of scientific classifications ‘from systems defined by reference to observable characteristics to systems based on theoretical concepts’ (Hempel, 1965: 154). This point, first stressed by Schwartz and Wiggins (1986), is indisputably relevant. However, other basic similarities still remain between the DSM-III system and neopositivist requirements, suggesting that a line of continuity is preserved.

First, one of the major aims of the DSM-III was an attempt to reach consensus on the meaning of necessary diagnostic terms that had been used inconsistently and an avoidance of terms that had outlived their usefulness (APA, 1980). It was on this basis that many psychoanalytic terms were questioned and possibly removed; even the classic distinction between neuroses and psychoses was questioned, in particular the term neurosis. In the DSM-III the group of the DSM-II neurotic disorders was divided into mood, anxiety, dissociative, and somatoform disorders. The term neurotic disorder was not deleted (this was done only later, in the DSM-III-R) but Spitzer tried as much as possible to restrict its meaning. In his view the term ‘neurotic disorder’ had to be used only descriptively to indicate that symptoms were ego-dystonic, that the contact with reality was preserved, that the social norms were respected, and so on. His aim was to distinguish this descriptive, acceptable meaning of the term from what he called the ‘neurotic process’. Spitzer suggested using the latter term to signify an unconscious conflict that, via defence mechanisms, was responsible for psychiatric symptoms, and this psychodynamic aetiological meaning of the term neurosis was deleted from the DSM (APA, 1980). In a later article he admitted that ‘the DSM-III committee believed that for nosologic purposes, the concept of the “neuroses” was no longer useful’ (Spitzer, 2001: 356), and this explains why even the descriptive, ‘survived’ sub-concept of ‘neurotic disorder’ was deleted from the subsequent editions of the DSM. This example shows not only that Spitzer and colleagues were fighting against psychoanalysis (as if it was just a battle of power between two different schools); more interestingly, it also reveals an implicit theoretical assumption: even in the DSM-III, psychoanalytic concepts were criticized, just as Hempel had earlier, because they were based on a non-testable theory about unconscious dynamics using terms which by definition are unobservable. Accordingly, this suggests that the DSM-III adopted (at least implicitly) two basic neopositivist assumptions: (a) scientific terms must be clearly distinguished from non-scientific ones, which should be rejected as nonsensical, and (b) the acceptable scientific terms are those that can be practically ascertained during the clinical evaluation because they are definite observable phenomena. The authors of the DSM-III were probably well aware that this was a sort of regulative idea because it was hard to restrict psychopathological symptoms exclusively on directly observable behaviours. However, they tried as far as possible to describe mental symptoms at ‘the lowest level of abstraction possible’ (Spitzer, 2001: 355) in order to preserve the ideal of a descriptive, non-inferential basis for the classification.

The distinction between the characteristics described with little inference and the terms that deserve a higher level of abstraction introduces to a third neopositivist tenet: the distinction between the observable and the theoretical level. Officially the DSM-III was self-defined as ‘atheoretical’ and thus the sense of our talking of a theoretical level deserves to be specified in detail. First, it should be stressed that in the DSM-III the term ‘atheoretical’ means more exactly ‘a-causal’, in the sense that it does not embrace a specific theory on the aetiology and pathophysiology of mental disorders (Aragona, 2006). The assumption is ‘that the term “disorder” includes conditions with varying levels of understanding of etiology or pathophysiologic process and that in only a small number of mental disorders do we possess the knowledge for the condition to be regarded as a “disease”’ (Spitzer, 2001: 353). Accordingly, the decision was to be wise and to avoid premature aetiological assumptions that might have hampered the use of the DSM-III by psychiatrists embracing different hypotheses on the causes of mental disorders (Spitzer, 1980). For these reasons, the preponderance of mental disorders (with the exception of the mental organic disorders) had to be ‘simply’ described, without any reference to a theoretical level. Second, the agnostic stance on the theoretical hypotheses about causality were not only aimed at increasing the diffusion of the DSM-III in different schools of psychiatry; the atheoretical choice was also strictly related to one of the major goals of the DSM-III: to strive to improve the reliability of psychiatric diagnoses. As a matter of fact, the problem of the unreliability of psychiatric diagnoses (Ash, 1949; Beck, 1962; Kendell et al., 1971) had become one of the most important problems for psychiatry, discrediting its aspiration to be regarded as a genuine scientific activity. This was mainly due to the resulting idea that groups of patients with the same diagnosis but selected in different contexts would be impossible to compare due to the different standpoints of the researchers (Aragona, 2006).

Accordingly, the DSM-III committee shared the ‘commitment to efforts to improve the reliability of psychiatric diagnosis as a goal of value in itself and also of value in raising the credibility of psychiatry as a profession’ (Spitzer, 2001: 354). It is noteworthy that the emphasis on the importance of the reliability of the terms used in order to assure their scientific status is another clear point of contact with Hempel’s lecture (see also Fulford and Sartorius, 2009: 40–2).

The analysis by Spitzer (1980, 2001) concludes that the first two editions of the DSM as well as the ICD-9, with their brief descriptions of each disorder, were responsible for the low levels of reliability found in the empirical studies of the time. The reason was that the brevity and general character of these descriptions were allowing too much flexibility in the subjective interpretation of the diagnostician on the definition of the contents and boundaries of any diagnosis. The solution (the adoption of explicit operative diagnostic criteria) will be discussed below. Here it is relevant that the analysis of the problem suggested that the diagnostic terms of the DSM-II and ICD-9, being abstract concepts, were strongly influenced by the background aetiological theories assumed by the diagnostician and for this reason were unreliable. On the contrary, the DSM-III committee was persuaded that at the merely descriptive level of symptoms (the level with the minimum of inference, that of the ‘simple’ description of what is actually present) the possibility of theoretically-influenced misinterpretation was basically removed. According to Spitzer (1980), the DSM-III system is descriptive because the definitions of the disorders are based on the description of their clinical features, and these features are usually described at an inferential level which is so low that they practically consist in easy-to-identify signs or behavioural symptoms. Such a conviction was supported by the results of the field trials showing the improvement of the diagnostic reliability when it was based on directly observable clinical features, independently from the agreement of the clinicians on the causes determining mental disorders (Spitzer, 1980). To conclude: the DSM-III basically retained the third neopositivist tenet, the distinction between the observable and the theoretical level, although in this case the theoretical level involved is not that of the explicative theory (at this level the DSM-III is atheoretical, at least officially) but that of the identification of the diagnosis (as seen, a level too abstract and full of theory to function as a reliable basis for the classification).

The last neopositivist feature to be considered is the use of rules of correspondence connecting the theoretical entities to the observational level. In the context of Hempel’s psychiatric taxonomy, this function was performed by the ‘empirical interpretation of the theory’, which had to connect ‘the theoretical traits, states, or processes with observable phenomena’ (Hempel, 1965: 150–1). In a later paper, Hempel stressed that in the standard neopositivist construal the correspondence rules are assumed:

to specify the intensions or the extensions of the theoretical terms by means of a vocabulary whose terms have definite and fully understood empirical meanings. That vocabulary, the interpretation base, has usually been conceived as a set of observational predicates, each standing for some property or relation of physical objects that is directly observable in the sense that, under suitable conditions, a normal human observer is able to ascertain its presence or absence in a particular case by means of immediate observation, without reliance on instruments or inferences. The assumption of such an observational interpretation base made it possible to present theoretical knowledge in empirical science as grounded on the data of direct observation in ways made explicit by the correspondence rules. More specifically, the scheme provide an objective and public evidential basis for empirical science by construing the basic evidence for scientific theories as expressed in the form of ‘observation sentences’ asserting the presence or absence of a directly observable attribute in a particular instance: and about such sentences, requiring for their formulation only an observational vocabulary, different normal observers would be in agreement. (Hempel, 1973: 371–2)

Here all the main themes of the diagnostic debate in psychiatry are clearly present: the importance of reliability; the need to specify the objective intention and extension of the diagnoses (concepts that are too theoretical to be empirically tested directly); the need for a basic set of directly observable and reliable features (the mental symptoms); the construction of observation sentences attesting their presence and of rules of correspondence connecting the two levels in order to transfer the empirical import of the symptoms to the corresponding diagnosis (the construction of the operative diagnostic criteria). The history of the development of the DSM-III shows that, having realized that the problem was whether patients described by different groups were comparable, the proposed solution was: ‘The use of formal diagnostic criteria by a number of groups, regardless of whether their interests are clinical, psychodynamic, pharmacologic, chemical, neuropsychological, or neurophysiological’ (Feighner et al., 1972: 57, emphasis added). This neokraepelinian strategy was imported into the DSM-III, where the operative diagnostic criteria were the rules of correspondence expressed in verifiable sentences like ‘x of the following y symptoms have to be present’. The diagnostician was thus ‘forced’ to apply these rules strictly. Although the diagnostic concept was theory-laden due to its higher level of abstraction and inference, it was believed that its scientific status was testable through the concrete application of those rules of correspondence (the operative diagnostic criteria) connecting the diagnosis to the observational symptoms.

Conclusion

The last section compared the basic neopositivist tenets to the implicit theoretical assumptions lying beneath the introduction of the operative diagnostic criteria in the ‘atheoretical’ DSM-III. It is shown that, de facto, the DSM-III was shaped in a neopositivist style. The following were discussed: the need of a clear differentiation of scientific and non-scientific diagnoses; the characterization of the last as those non-testable in principle (e.g., the neuroses, based as they were on unobservable unconscious dynamic processes); the strong commitment to focus on descriptive, reliable evidence to support the recommended scientific diagnoses; the introduction of the operative diagnostic criteria as a set of explicit inclusion/exclusion criteria aimed at mechanistically expelling the subjective, unreliable clinical judgement from the diagnostic procedure. All these features clearly conform with a significant portion of Hempel’s recommendations and, in general, with the neopositivist perspective.

It is indisputable that the DSM-III did not follow every suggestion advanced by Hempel, in particular having rejected his emphasis on explanation in the name of the atheoretical approach. However, this was done as a part of a definite strategy. For example, Spitzer stressed explicitly that although validity and the discover of aetiology were the major goals for the new scientific classification, these could not be accomplished unless reliability was first guaranteed (Klerman, Vaillant, Spitzer and Michels, 1984). De facto this strategy and the way it was executed were, without any doubt, neopositivist in their essence.

[Part 2 will be published in History of Psychiatry 24(4).]