Validity, realism, and normativity
Dominic Murphy
Intuitively, validity is obvious. Whereas reliability is a gauge of agreement across measurements, validity is supposed to be about what is really there. It looks obvious that even the most expert observers could all agree but nonetheless be wrong, whereas proper validation reassures us that we are measuring something that is really there. The judgment that the Diagnostic and Statistical Manual of Mental Disorders (DSM) has exactly this shortcoming—reliability but insufficient validity—has led the National Institute of Mental Health in the U.S. to encourage the use of the Research Domain Criteria (RDoC) in grant proposals (Insel et al. 2010; NIMH 2011). The originators of the RDoC acknowledge that the system entrenched by previous versions of the DSM has increased diagnostic reliability. But they worry that it is too detached from the nature of mental illness, conceived of as disorders of brain circuits. These disorders could be studied at many levels and need not be identified with simple lesions. But future models of mental illness are expected, on this vision, to draw on psychological, neurological, and genetic mechanisms, and diagnosis will have to be based on these models in order to be properly based on the underlying facts about mental illness, rather than on clinical signs and symptoms.
Clearly, in this case, validating a diagnosis is thought of as understanding its underlying causal structure: a diagnosis is valid if it rests on a biological process that can be identified by experiment and observation using the methods of the biological and cognitive sciences.
Any approach to psychiatry that looks to science to validate its categories in this way must meet at least two conceptual challenges, which I will discuss in this chapter. First, there is a metaphysical challenge, which is that a concept of validity tied to the uncovering of neurobiological processes commits realism, the sin of thinking that science can tell us how the world is really put together. Many philosophers think that realism adds a wholly unnecessary and unwarranted metaphysical commitment, and that all science can really tell us about is a set of relationships among the data. These relationships let us make predictions and exert some control over nature, but do not tell us what is really out there. I shall argue that biologically based psychiatry does not make any needless metaphysical commitments, and if it is realist, it is realist in an entirely harmless way.
The second conceptual challenge to the realist interpretation of validity I will consider is normative. The challenge is that there is an important sense in which diagnoses cannot be validated at all, if by “validation” we mean “shown to be a real disorder.” All validation can do is show that a pattern of behavior deemed to be clinically significant depends on a physical process. Whether that pattern is really pathological—rather than immoral or harmlessly odd—is another matter. The issues here are tricky, but I think this second challenge probably cannot be met. Suppose we think that judgments of pathology are like judgments of positive charge, i.e., scientifically grounded, rather than judgments of bad taste, i.e., human responses. If so, there has to be some natural fact of the matter about whether some physical system—at whatever level of description—is dysfunctional. I will review some attempts at doing this and conclude that they fail. Predictions about physical states can be validated, but disorders cannot be.
Before I discuss the metaphysical and normative questions I just raised, I will say a little more about validity in general. Then I will look at the idea of validation as the uncovering of the structure of the world. Ken Schaffner (2012) has recently argued for a pragmatic account of validity which disputes the whole notion of some phenomenon really being there. He insists that it is sufficient to think in terms of predictability and utility. I will argue that there is nothing in his overall position to worry the most uncompromising realist. Then I will try to put the normative point in the context of recent disputes about the proper analysis of the concept of mental disorder.
As Zachar (2012) points out, the intuitive sense of validity I introduced—telling you what’s really there—does not map neatly onto any of the many concepts of validity that psychiatrists argue about. There are, as he puts it, numerous small-v senses of validity that don’t always fit together, nor constitute a coherent big-V concept. Zachar advocates validity pluralism, arguing that psychiatry can employ many distinct small-v concepts of validity. These small-v concepts enable us to answer specific questions about our constructs and their relation to sundry tests and statistical measures. But how are these concepts all related—what do they have in common, and what makes them concepts of validity rather than something else? Zachar sees validity pluralism as providing a means “to construe validity as a matter of degree” (p. 22). That does suggest that the many concepts used in day-to-day medical and psychometric practice bear a straightforward conceptual relation to big-V validity, however modest and uncertain their clinical application. They appear to be attempts to get at whatever the big-V concept is doing: we might never get at the real metaphysical underpinnings, but we aim at approximating them by measuring whatever is accessible. The epistemic gradient provided by more or less rigorous tests gives us varying degrees of certainty that what we are measuring is what is really there. But what we get, pretty much all the time, is not a representation of underlying reality but a representation of measurable relationships of clinical interest. That is, we get a bunch of results that, we hope, will correlate a diagnosis with some test outcome, or distinguish between populations. On the basis of the results, all manner of treatments and grants are dispensed.
The big-V concept of validity, then, may not capture what the sciences of the unsound mind aim at in ordinary practice. Mostly, those sciences care about relations among measurements. The big-V concept seems more like a way of cashing out a philosophical position, to wit, that understanding the world involves grasping its causal structure. Some philosophers, as we will see in a moment, think that the big-V position carries with it objectionable philosophical commitments. They think that embracing the big-V position commits one to a kind of scientific realism that allows inferences about what is really there that the relevant science cannot legitimate.
The NIMH now seems to have thrown its considerable weight behind a particular causal-explanatory account of big-V validity. This too is a paper about big-V Validity—the idea that we can ascertain that the causes of mental illness are genuinely out there in the structure of the world, waiting to be discovered. That doesn’t mean, however, that the diagnoses we currently employ will be validated. Many of our existing diagnoses may get replaced or at least reformed
Following Zachar, although he does not quite put it this way and may cavil at it, I have suggested that the small-v concepts should not be seen as aiming at something completely different from the big-V version. Scientific concepts are part of unfolding epistemic projects, and all the different concepts of validity involve attempts to understand and manipulate relationships among parts of the world. The small-v concepts track measurable relationships, but the existence of the relationships does suggest that something is going on. Even if we don’t know the underlying structure, we can still obtain clues to the way it works. I think that the small-v concepts often aim to establish correlations that can be seen as points on the way to a fuller causal story. However, I do not think that the fuller causal story should be seen as making illicit metaphysical commitments, for reasons I will now outline.
Schaffner also detects a gratuitous commitment to reductionism in much recent talk of psychiatric validation. He suggests that this commitment is biologically ill-informed. On the latter point, he is surely right. A commitment to reductionism often amounts to simply a metaphysical preference for the very small. But sometimes it involves mixing up a metaphysical position—small things good, big things bad—with an epistemic one, which is the position that satisfactory explanation involves seeing how complex phenomena arise from interactions among other (ideally, simpler) ones.
This is analysis, rather than metaphysical reduction, and it is essential to our idea of good explanation. Reductionists often couch their theses in terms of explaining higher-level phenomena in terms of lower-ones, but that is entirely optional. The point of good explanations is that they show how things come about. For that to happen, we don’t need to appeal to lower-level phenomena, although we sometimes can, and scientists often strive to. What we do need is to get a grip on the processes in the world that explain the phenomena of interest. This underlies the objections by RDoC advocates to the descriptive, syndrome-driven approach to mental illness. The DSM and ICD see mental disorders as collections of signs and symptoms. The objection to this descriptive approach is that it is outmoded. In the rest of medicine it was supplanted in the nineteenth century by the concept of diseases as resting on specific pathological processes in the organism. The strong interpretation of the medical model in psychiatry holds that mental illnesses are diseases of this type. They are not just sets of co-occurring systems, but destructive processes taking place in biological systems.
It might seem that there are different conceptions of validity that are naturally congruent with these two interpretations. For instance, one might think that the strong version demands an account of validity in terms of what is really there, whereas the minimal descriptive version would be satisfied with mere predictive utility. After all, the strong interpretation is committed to the idea of specific pathologies, which involve a genuine causal story about processes unfolding within the poor individuals who share a diagnosis. But on second thoughts, this is not correct, because either the strong approach or the minimal one can rest happily with an instrumentalist or pragmatist approach. Schaffner’s (2012) view is that scientific reality is a matter of predictive utility, and one can make predictions about the behavior of brain systems or ion channels just as well as anything else observable.
Kendell and Jablensky (2003, p. 9) clearly distinguished utility and validity, defining the former as the provision of useful information about outcomes and/or testable hypotheses about correlates of diagnoses. Validity they saw metaphysically as the existence of categorically distinct kinds of diagnostic entity; genuine carvings at nature’s joints. Schaffner (2012) argues that such a distinction is unsupportable. He makes a bet that, based on general biological considerations, we should expect to see dimensions and not categories in human populations. But Schaffner’s main arguments are philosophical, based on the pragmatist idea that we cannot separate utility from a genuine representation of nature. For Schaffner, utility is constitutive of reality. Schaffner does not embrace a strong anti-realism about psychiatry and the related fields, but he does reject a thoroughgoing realism. Although I will not attempt a general defense of scientific realism, let me critically analyze some of the arguments Schaffner offers.
Philosophical disputes over realism have often centered on the notion of the unobservable. Electrons are a paradigm example, but other posits of fundamental physics play the same role. In psychiatry the situation is different; theories and constructs employ latent variables rather than reference to unobservables, as well as, just like any science, hypotheses that go beyond observed cases to cover the infinitely many more cases that have never been—could never be—observed. Science is a massive data-compression project: rather than enumerate all the individual cases, we apply a general label to them such as “schizophrenic” (or “soluble” or “haplodiploid”). This commits the scientist to a bet, which is that the behavior of phenomena that are unobserved but fall under the label will be enough like the behavior of the observed cases to justify the application of the label (i.e., labels are inference tickets that tell us what to expect from unobserved cases).
The chief anti-realist challenge concerns the unobservable. The charge with respect to the unobservable is that our techniques of observation can carry us only so far: we can perceive lots of things, but those things that cannot be observed should not be presumed to exist. This anti-realist approach originates in reflection on the physical sciences where mentioning unobservable entities might help with understanding the mathematics that expresses the main commitments of the theory. But the unobservables themselves should not be taken to exist. The point about arguing over what is really there is clear enough in these sorts of physical cases, since what is at stake is the existence or otherwise of bits of the world.
It is not clear how to translate this dispute into the context of psychiatry. What are the unobservable things whose existence we are contesting? The answer could be either all of psychiatry, or none of it. In the first case, one could argue that signs and symptoms are conceptual inferences. What is actually observed are bodily movements and vocalizations. The notions of sign and symptom go beyond these observations and amount to theoretical posits.
Objecting that these posits are illicit is absurdly strong, for it would also rule out every other possible psychological state. We have only movements and noises to go on when we see someone as sad or flirtatious, or as believing one thing and desiring another. Seeing someone’s behavior as clinically significant obviously requires training in a special class of concepts, but so does mastering our everyday folk psychology. In short, the mere presence of a special set of concepts does not license anti-realism, otherwise we would have to be anti-realists about everything that we attribute to human beings on the basis of their behavior, including “this person is dancing” and “this person is kicking a football.” Concepts do not just describe the behavior, they interpret it.
Obviously nobody is arguing for anti-realism about human behavior in general (although I am not arguing for an uncritical realism about all our psychological concepts either). But anti-realism does have some force when we start thinking about an empirical approach to psychiatry. For, as Schaffner notes (2012, p. 176), the constructs that we typically seek to validate must be validated indirectly. So we are looking for correlates of constructs, and the constructs put an interpretation on the observed behavior.
Schaffner (2012, p. 177) adopts the view that reality is constituted by scientific utility. He calls this conditionalized realism, as an acknowledgment that our acceptance of any scientific claim is conditional upon 1) not just the evidence, but the acceptance of auxiliary hypotheses (about, for example, the reliability of our instruments) and 2) the absence of plausible alternative hypotheses.
I agree with Schaffner on both of these points, but I dispute what he sees as their significance. Neither point causes any trouble for the scientific realism that Schaffner regards as “too strong.” Schaffner insists (2012, p. 178) that we do not have any “direct intuitive experience of the certitude of scientific hypotheses or theories.” This is true, but whoever said realism requires some direct intuitive certitude, as if doing science were a matter of revelation? What a realist might be inclined to insist on (as Kitcher 2013 does) is that perception provides unmediated access to the world. Our perceptual encounters with the world require us to be in certain mental states, but we do not perceive those states. We just perceive the world.
Certainly to perceive what is there one needs to possess concepts. Most people will not recognize a googly, a sestina, or a nudibranch when they see one, because they don’t have the relevant concepts. But dependence of recognition on concepts does not mean that what is perceived doesn’t exist, even if it requires a background of concepts. (Of course in saying this I am disregarding global skepticism about perception, but that is everybody’s problem, and not a difference between this view and a pragmatist one.) The realist point is that our perception of the world is a matter of causal contact with it. We have to be in a certain psychological state—concept possession—to recognize a nudibranch or a panic attack. But we do not perceive mediating psychological states or entities. This departs from the “way of ideas” of early moderns like Hume and Descartes, which assumed that we perceived the world via perception of mediating mental entities. This direct causal contact with the world, however, is not some strange philosophical power of “intuition” that puts us in touch with the constitution of reality. It is ordinary perception.
Schaffner argues that the most one can hope for is that a scientific claim has relatively direct evidence in its favor. This is all theory or construct can hope for because of its inevitable latent aspects. Like Kitcher (2013) I propose that we see scientific access to theoretical constructs as akin to epistemic access to ordinary folk “constructs” like flirtatiousness or cynicism. Ordinary perception, like science, provides relatively direct access to the world assuming that we accept auxiliary hypotheses (the light is good, I’m wearing my glasses) and can discount plausible alternative hypotheses. The same considerations that support truth claims in ordinary life support truth claims in science, or at least in psychiatry. Schaffner does not actually dispute this, but he regards it as a strike against realism. In contrast, I think it is decisive evidence in favor of realism, because I reject the idea that realism requires some direct intuitive access to the world.
There is another issue, though, concerning the relations between truth and utility, where Schaffner and the realist may part company. Schaffner thinks that the point of a psychiatric or psychological construct is essentially practical: to gather useful information about a population through tests that are as rigorous as we can devise. Again, I agree, but would emphasize something different, which is that these aims are more likely to be fulfilled the more we know about the underlying structure of the categories involved.
The realist claim would be that schizophrenia, for example, is not just a label that gives us a convenient way of grouping people who seem worth grouping together for predictive and other purposes. We want to think that there is a genuine natural phenomenon that they all share, as there might be if we grouped organisms together on the grounds that they all share the same viral infection. The realist bet is that coming up with the construct is the first step toward further investigation. To begin with we want to find biological markers that correlate with schizophrenia. Much of biological psychiatry’s commitment to big-V validity comes in here as we look for the biologically relevant stuff going on inside people. Schaffner’s point appears to be that these further questions can be asked, but that they should not be seen as leading us toward the truth about how the world really is; they will just uncover more facts about the relations between measurements and outcomes. It is correlations all the way down, and the predictive power and utility gathered in this way constitute reality.
But the schizophrenia construct, unlike electrons, is not a further type of stuff that lies behind the appearances and explains the appearances: it is constituted by visible phenomena. We are not looking for an unobservable, but asking which observable phenomena explain the symptoms. Latent constructs and unobservability are not the same, and when we push further after the sorts of markers and causal processes that the RDoC envisages, we are uncovering causal structures and generating useful predictions. Of course we may refine our categories in such a way as to decide that the original construct is of limited use, and schizophrenia may disappear in favor of one or more alternatives, but the point remains: the generation of new constructs using biological, psychological, or genetic markers is a step toward both causal knowledge and utility. The purpose of a psychiatric construct might be utility, but the investigation of it can uncover the causal structure of the world. This is just part of our normal drive to find out why things happen.
We often explain an entity’s behavior in terms of the kind of object it is, as when we say that Miffy is afraid of dogs because she is a rabbit. Cooper (2007) calls this natural history explanation. We might prefer to think of a natural history explanation as a placeholder for a more complete explanation, but even in the absence of a causal account of why something behaves as it does, we may obtain useful information just by noting the characteristic relations it enters into. Different types of plant may need to be put in the ground at different times, or in different seasons, in order to maximize crop yield, for instance, and different patients may respond to different drugs even if the causal basis of these differences remains unknown. And although we might be ignorant about whatever it is that explains those relations, the natural inference to make is that there is something about the world that explains why the relations hold reliably. Miffy might do all sorts of things occasionally, but she is reliably and predictably afraid of dogs, and everyone will conclude there is something in Miffy’s nature that makes her behave like that. Predictable and repeatable phenomena get measured and serve as the basis for further inquiry.
Zachar’s idea that small-v validity is on a continuum with the big-V concepts expresses this thought: understanding reliable and predictable relations among measurements is the first step toward understanding the hidden structure that accounts for surface phenomena. Small-v validity lets us make use of descriptive and statistical reasoning and offers the hope of accurate prediction and effective control. To know more about why Miffy is so craven we look inside her and her conspecifics to find the perceptual, endocrinological, and physiological mechanisms that relate, in her brain, the representation of a canid to a particular suite of evasive behaviors. It is also a good bet that closely related species will share such systems. And we can then go on to build a picture of the evolution of those systems by seeing how they differ in less related species. It is useful to know that x system often produces y outcome in z conditions, so that we learn about new interventions and points of manipulation. But such knowledge also allows us to dig deeper, and look for the underlying causal picture (and then the causal picture that underlies that, and so on). All this uncovering of further connections gives us both useful predictive knowledge and knowledge of causal structure.
To reiterate: my talk of underlying systems should not be taken as a commitment to reductionism. In my sense, long-term unemployment or family environment can underlie something just as much as stretches of DNA can. The point is that looking for useful, predictive knowledge is a way station on the road to figuring out how things really work. There is little reason to suppose that psychiatry has got very far along that road, but the objection to big-V validity is not a metaphysical or epistemic objection in principle to the project. Big-V validity requires no commitment to direct intuition of the world, nor is it at loggerheads with a desire for utility. It builds on utility and it uses the methods that we use to establish useful predictive knowledge.
So I think that big-V validation of causal stories about mental illness is possible. The question I turn to now is whether that also amounts to validation of diagnoses, where validation is understood as uncovering the causal structure of the world.
The second topic that the big-V conception of validity leads me to is normativity. Suppose we find something that is really there, perhaps a distinctive pattern of activity in a brain circuit that correlates reliably with thought disorder and makes sense relative to our background knowledge of cognitive neuropsychology. Are we showing that a diagnosis corresponds to an objective fact about human ill-being, and hence supports the idea that psychiatric illnesses are not social constructions in some pejorative sense but rather objective pathological conditions? No, we are not. We are showing that there is some objective fact that explains why one group of humans is different in some respect to some ideal type of normal behavior, but we are not showing that the difference is pathological.
For this to amount to big-V validation—showing that something is really part of the world—biological dysfunction must be objectively established, not just biological markers. According to the RDoC advocates, however, the goal is to discover the underlying dysfunctions that give rise to psychiatric symptoms. I propose that we are entitled only to a weaker goal, the discovery of biological difference, not objective dysfunction. We judge certain ways of life to be pathological. And we can discover objective scientific correlates of those ways of life. This discovery of why people act in ways we judge to be pathological is what the anti-psychiatric movement bet against. It should not have made that bet. On the other hand, the discovery of objective biological difference does not mean that we have found a dysfunction. The attribution of dysfunction requires assumptions that are themselves normative.
The yoking of biological and normative considerations in the validation of mental disorder (or any medical condition) is the hallmark of what I call the two-stage view. It is the most popular account of psychiatric disorder among theorists who deny that ascriptions of mental illness are entirely normative. It was introduced by Wakefield (1992), who adapted earlier ideas of Boorse (1975, 1976). Two-stage theorists hold that there are two individually necessary and jointly sufficient conditions for the attribution of a disorder. First, there is a biological dysfunction. Wakefield’s innovation saw dysfunction as a failure by a bodily system to perform the naturally selected function that explained the system’s replication in past generations, whereas Boorse saw dysfunction in terms of a system’s failure to contribute to the overall systemic capacities of the organism.
Second, the dysfunction must result in harm to the individual concerned, as judged by prevailing social norms. “Harm” is a normative notion. So psychiatric dysfunction is assumed to be a matter for medicine to establish, just as it would establish that an esophagus has become dysfunctional. But, according to this view, whether somebody is harmed is a matter of prevailing social judgments. Establishing whether a specific state of affairs constitutes harm for a person is often going to be controversial—although consensus on many harms is easy. The issue of what counts as harm is not one science can settle, though empirical findings can provide evidence that supports the judgment of harm.
The two-stage view, then, is designed to give both science and social context their due. It aims for a middle ground between (i) a scientism that says psychiatry has no role for values at all and (ii) a constructivist claim that our judgments that a person is disordered depend entirely on their having violated some norm. The view supposedly respects both the role of science in psychiatry and that of social norms. However, the two-stage view faces two sets of conceptual problems. First, there are the difficulties involved in justifying the intuition that science plays a role in the discovery of objective facts about dysfunction. In the rest of this chapter I will discuss what that role is.1
Second, we have an intuition that norms have a role to play in whether an individual is harmed by his or her dysfunction. This intuition is thought to have normative implications with respect to rights and duties to treatment. Now, unpacking the notion of harm is at least as problematic as unpacking the notions of function and dysfunction (e.g., see De Block 2008). But the relevant concept of harm involves judgments about the quality of a life. These judgments need to be sensitive to both the individual’s own needs and goals, and the ideas about well-being that feature in the wider society.
There are two broad concepts of function in biology (Godfrey-Smith 1993). We can see the function of a biological system as Wakefield does, as that effect of the mechanism which contributed to the success of the ancestral population and thereby the replication of the system (Millikan 1984). I call this the selectionist view. But we can also understand a system’s function as the contribution it makes to a broader biological system of which it is a part (Cummins 1975). I call this the systemic view.
The life sciences encompass many projects, and trim their accounts of function to suit them. Cummins argued that the basic explanatory use of function talk in the life sciences derives from a particular analytic strategy in which the biologically significant capacities of a whole organism are explained by breaking down the organism’s biology into a number of “systems”—the circulatory system, the digestive system, the nervous system, and so on—each of which has its characteristic capacities. These capacities are in turn analyzed into the capacities of their component organs and structures. We can reiterate this systemic concept of functions through levels of physiology, explaining the workings of the circulatory system, the heart, certain kinds of tissue, certain kinds of cell, and so on. Much mechanistic research in biology exemplifies this approach.
As well as questions of survival value, we can ask questions that simply aim to find out how a system does what it does in the context of the superordinate system. There is an idea almost as old as natural selection that may help to answer these questions. It is Claude Bernard’s (1927 [1865]) suggestion that major systems in the human body seek to maintain stable internal homeostatic states. Bernard argued that organisms can only explore and transform the external environment if they have sufficient internal stability. In this view the answer to the question “What is the function of the major physiological systems?” is “To keep the internal environment stable.” Homeostasis, not survival value, is what guides physiological answers to questions about causal explanations of biological systems in this tradition.
It has become clear that this cannot be quite right, since much behavior (reproductive behavior, for example) seems to disrupt homeostasis and organisms endure lengthy periods of stress in which the system is dysregulated. This has led some theorists to embrace the supplementary idea of allostasis, “to take account of the physiology of change and adaptation to diverse circumstances, and to the behavioural and physiological anticipation of future events” (Schulkin 2011, p. 5). Organisms achieve internal viability by adapting as circumstances change over time. Regulation involves response to, and anticipation of future, social and environmental needs. Physiology and medicine appear to be guided by a homeostatic–allostatic concept of function. They employ the systemic view, and not the selectionist view.
The idea behind the two-stage view is that science can tell us that the disorder is real not just by finding markers or causal processes, but by finding dysfunctions, thereby answering the skeptic who thinks of diagnosis as just labeling.
The two-stage view says that scientific facts play a significant role in determining whether or not a condition is a disorder, by ascertaining that something is a dysfunction. If there is a dysfunction that explains the symptoms, then we have a clear link between the biology of the organism and the behavior that attracts the diagnosis (and more broadly, the attention of one’s fellows, which is what leads to the search for the underlying dysfunction).
The dysfunction criterion was initially introduced to help us determine which individuals are in fact disordered, in a way that avoids subjective, mind-dependent, or culturally relative judgments. The scientific aspect of the two-stage view thus has the job of rebutting the skeptical claim that disorders are only violations of prevailing social norms. To rebut such skepticism using the two-stage view, psychiatry (or the basic sciences it draws on) needs to be able to see processes as objectively dysfunctional, not just as causes of ways of being. The skeptic asks why we should see a pattern of brain activity as a correlate of disorder rather than a correlate of eccentricity or immorality or something else that is none of psychiatry’s business. The two-stage view is supposed to give us a decisive answer: that brain activity is not just different, but disordered, and we can show it scientifically.
An alternative view is that disease concepts have a different structure altogether. Cooper (2007) and Murphy (2006) have drawn an analogy between the concept of mental disorder and that of weeds. Weeds are not a scientifically relevant category of entities. We can perhaps say that a weed is a fast-growing species that negatively impacts on economically valuable crops, usually through competition for nutrients, sunlight, and space. What fixes the extension of “weed” (and similar concepts like “vermin” or “precious metal”) is a set of contingent human interests that can change over time.
There is nothing inherently weedish about a species; weeds are just species that we don’t like because of certain interests that we have. Suppose that determining that a condition is a disorder is like determining that a plant is a weed. The judgment is determined by normative considerations that we have already made. But nonetheless there is real, explanatory mind-independent knowledge to be had about each sort of “weed.” Or take “precious metals”: these are just metals that are valuable. They are valuable because demand far outstrips supply, but the demand is based on their aesthetic qualities and the way they let you show off, not on their chemical properties or mere rarity. Aluminum used to be precious but now it is not. Other metals, like Bismuth, are rare but not precious. Being a precious metal is a matter of quite complex human considerations, but there is still perfectly good scientific knowledge of each one of them to be had.
It might be, then, that “mental disorder” or “disease” more generally, works like “weed” or “precious metal,” in that it is a concept that is driven by human concerns. We respond to people in a way that makes us judge them to be pathological. We can uncover facts about them, but the facts do not explain why they are judged to be pathological; they rather supply the causal story about the behaviors that lead us to make the judgment in the first place, and that judgment would be in place even if the facts were different.
I’ll call this the norm-first view. Unlike pure constructivist views it acknowledges the role of science in establishing genuine knowledge of the physiology or psychology underlying a diagnosis and in providing opportunities for remedy and manipulation. The norm-first view contrasts with the two-stage view in holding that science does not uncover dysfunction in a way that is independent of our value judgments; science is directed by those value judgments. We first ascertain that someone is dysfunctional based on our socially generated expectations about what people ought to be like. Science can then investigate those people, but it does not determine whether they are pathological. The order of discovery here is often the same as that which the two-stage view would lead us to expect. But the order of nature is the other way round.
Theorists have often thought that the presence of normativity in psychiatry would undermine psychiatry’s status as a branch of medicine, but if diseases are like weeds, then all of medicine has the same normative status as psychiatry. Consider cancer as an example. It is an uncontroversial instance of a biomedical disorder.
If we want to understand cancerous cell development we have two ways of proceeding. One way is to initially proceed by building an idealized model of cell development in general. We can then model cancerous cell development by explaining what “breaks down” during development in order to explain cancer as a “biological dysfunction” of the normal cell development. An alternative would model cancerous cell development on its own terms. Groups of scientists could proceed differently: one group would understand cancer as a “biological dysfunction” whereas the second group would understand it as a particular natural kind of process. Both models seem capable of capturing precisely the same information with respect to understanding biological change and providing different points at which we can intervene to alter the process we have modeled. We can disrupt the course of cell development and we can disrupt the course of cancerous cell development.
But cancer cannot be both a biological dysfunction and a natural kind of biological development at the same time. The two-stage view bets that there are further scientifically discoverable facts that tell us definitively that cancer is really a dysfunction. But what fact is there that science can discover what discriminates between cancer as a dysfunction and cancer as an unusual developmental pathway? If there is no such fact, then we must reconsider whether science can play a foundational role in determining that conditions are disorders, as the two-stage view says. The skeptic’s alternative is that science discovers important biological facts guided by prior normative judgments that something is a disorder.
At this point in the argument, a proponent of the two-stage view has another option to rebut the skeptic who embraces the norm-first view. This is to ask about the role of the system in the overall economy of the organism. The two-stage theorist who adopts a systemic view of function can say: look at what the system you are studying does for the organism. The reason cancer is a dysfunction is that it drives the organism out of equilibrium and into a new state in which other systems stop being able to act as we usually explain them. This approach also requires a way of differentiating normal from abnormal development; it defines normal development as the set of pathways that lead to the final, functional, adult form.
The systemic theorist, then, can use the idea of a natural hierarchy in the organism to defend the claim that disease perverts the functioning that is normal for an organism. However, this does not solve the problem of finding natural, mind-independent dysfunctions, for the proponent of the norm-first view can now ask what justifies our idealized or assumed “normal” systems? Variation in biological traits is ubiquitous. Therefore, establishing whether or not a mechanism is functioning normally depends on whether an overall picture of normality for the organism can be adumbrated in a way that doesn’t depend on our prior values. The proponent of a norm-first view doubts that can be done.
These arguments might occur anywhere in medicine. But this sort of skepticism that motivates a norm-first view does seem to be especially telling in psychiatry, in so far as it is concerned with how people live, think, and feel. Humans are a set of systems, but what is the overall function of the organism they all comprise? In other words, how should people live? If the two-stage view is to work we must be able to objectively determine when people are living dysfunctionally, because only then can we say that there is a failure of the overall system, and thus establish that a subsystem is dysfunctional in the sense that it is not making the correct contribution to overall functioning. At this point, then, the next move in the dialectic should be for me to tell you how we can find out what the good life for humankind is. I hope you see the problem. The norm-first view says that science cannot answer that question. However, there might doubtless be scientifically relevant findings, such as those that suggest that you might not want to consume three packets of cigarettes and a liter of vodka a day if you plan to thrive in the long term.
Nobody doubts that people often suffer horrible mental pain or become detached from reality. There is nothing wrong with responding to these plights in the hope of making things better. And in so far as big-V validation can understand the basis of these plights in our biological nature we should pursue it. But that is not the same as validating a claim about pathology. It is uncovering the causal basis of what we respond to when we respond to the mentally ill. The two-stage view thinks that we can establish a diagnosis with scientifically objective credentials. I have not refuted this claim, but I have suggested that to do so requires a scientific solution to the puzzle of what human beings ought to be like, and I do not like the odds of doing that.
The big-V idea of something really being there brings up two philosophical issues within psychiatry with special resonance: first, a metaphysical one about the relation of validity to scientific realism of some sort; second, a normative one about whether we can validate a diagnosis, or merely validate a scientific conjecture. Roughly, the first issue raises the question whether we can know if something is there, and the second raises the question whether, if we find something, we can call it a disorder. I have discussed these two issues in turn. “Validation” sounds like a justificatory concept, as though we were not simply confirming a hypothesis, but also attesting to the correctness of a normative claim. Underlying accounts of validity is the idea that we do not just find a physical malfunction or imperfection, but something more, a genuine disorder built into the structure of the world. I don’t think we can find that. Whatever biological or psychological phenomena we uncover will still leave the normative issues open. In that sense, we can’t validate a diagnosis. We can just correlate it with part of the world’s structure.
1. This discussion draws heavily on Roe and Murphy 2011.
Bernard, C. (1927/1865). Introduction to the Study of Experimental Medicine. New York: Macmillan.
Boorse, C. (1975). On the distinction between disease and illness. Philosophy and Public Affairs, 5, 49–68.
Boorse, C. (1976). What a theory of mental health should be. Journal for the Theory of Social Behavior, 6, 61–84.
Cooper, R. (2007). Psychiatry and the Philosophy of Science. Stocksfield, UK: Acumen.
Cummins, R. (1975). Functional analysis. Journal of Philosophy, 72, 741–64.
De Block, A. (2008). Why mental disorders are mental dysfunctions (and nothing more). Some Darwinian arguments. Studies in the History and Philosophy of Science (Part C), 39, 338–46.
Godfrey-Smith, P. (1993). Functions: consensus without unity. Pacific Philosophical Quarterly, 74, 196–208.
Insel, T. R., Cuthbert, B., Garvey, M., et al. (2010). Towards a new classification framework for research on mental disorders. American Journal of Psychiatry, 167, 748–51.
Kendell, R., and Jablensky, A. (2003). Distinguishing between the validity and utility of psychiatric diagnoses. American Journal of Psychiatry, 160, 4–12.
Kitcher, P. (2013). Real realism: The Galilean strategy. In Preludes to Pragmatism: Towards a Reconstruction of Philosophy. Oxford, UK: Oxford University Press.
Millikan, R. (1984). Language, Thought and other Biological Categories. Cambridge, MA: MIT Press.
Murphy, D. (2006). Psychiatry in the Scientific Image. Cambridge, MA: MIT Press.
NIMH (National Institutes of Mental Health) 2011: NIMH Research Domain Criteria, Draft 3.1. Available at: <http://www.nimh.nih.gov/research-priorities/rdoc/nimh-research-domain-criteria-rdoc.shtml> (accessed October 2013).
Roe, K., and Murphy, D. (2011). Function, dysfunction, adaptation? In P. Adriaens and A. Block (eds), Maladapted Minds (pp. 216–37). Oxford, UK: Oxford University Press.
Schaffner, K. (2012). A philosophical overview of the problems of validity for psychiatric disorders. In K. Kendler and J. Parnas (eds), Philosophical Issues in Psychiatry II: Nosology (pp. 169–89). Oxford, UK: Oxford University Press.
Schulkin, J. (2011). Adaptation and Well-Being. Cambridge, UK: Cambridge University Press.
Wakefield, J. C. (1992). The concept of mental disorder: On the boundary between biological facts and social values. American Psychologist, 47, 373–88.
Zachar, P. (2012). Progress and the calibration of scientific constructs: the role of comparative validity. In K. Kendler and J. Parnas (eds), Philosophical Issues in Psychiatry II: Nosology (pp. 21–34). Oxford, UK: Oxford University Press.