Soc Psychiatry Psychiatr Epidemiol (1998) 33: 356362

Springer-Verlag 1998

ORIGINAL PAPER

S. Schwartz

The role of values in the nature/nurture debate about psychiatric disorders

Accepted: 26 January 1998

Abstract This paper discusses the role that values play in our estimates of the relative importance of nature/ nurture factors in the etiology of psychiatric disorders. It is argued that all parameter estimates, heritability estimates, variance explained, odds ratios, and relative risks, are based, at least in part, on value judgements that are made in the course of our research activities. Four aspects of causal inference where values are particularly inuential are discussed: gene/environment interactions, environmental specicity, the distinction between causes of variation and the causes of a factor, and the reication of our measures. The value component in our research is illustrated in two areas of research: life events and depression and individual factors in violence.

Pair 1 The major mental illnesses (chief among them schizophrenia, manic depression and depression) are diseases of the brain. A series of scientic breakthroughs beginning in the early 1970's...has established this beyond serious question. No pathological fact is better established...than that deviations from healthy structure are generally present in the brains of insane subjects. (Issac and Armat 1990, p 19) versus Despite all of this biopsychiatric propaganda...depression is a readily understandable expression of human despair. (Breggin 1991, p 122). Pair 2 Depressive illness can be insidious because it frequently resembles the kind of unhappiness that is a normal part of human living...One reason is that biological depression can be triggered by life events, which may lead people to discount depressive disease as simply normal psychological responses. (Klein and Wender 1993, p 6) versus The misery, demoralization or distress a person feels are not the problem. They are consequences of the problem. Misery is not only real, it is realistic. (Mirowsky and Ross 1989, p 5). These two sets of statements are totally at odds with one another, representing fairly clear nature/nurture positions on a mental health problem. Depression is biological or environmental, a result of an internal dislocation or a natural response to bad situations. In each case the arguments are intelligent and well grounded. Each seems plausible and conclusions about their relative validity are not obvious, based on analysis of the data presented. This paper is an attempt to try to clarify some underlying issue that can illuminate these dierent perspectives; it takes as its starting point the following parable.

Introduction
Despite the recognition of the importance of both genetic and environmental factors in most current causal models of psychiatric disorders (e.g., Gottesman 1991), the nature/nurture controversy in psychiatry is far from moribund. While much research revolves around the question of relative contributions, conclusions are often stated in rather absolute terms. Nature is often broadly conceived as biology and nurture as psychological and sociological inuences. This paper focuses on one aspect of this debate the role of values in making causal inferences about the relative importance of nature and nurture in psychiatric outcomes. The importance of this issue can be illustrated by contrasting the conclusions reached in four books with very dierent perspectives on the nature/nurture components in the etiology of psychiatric disorders, which can usefully be paired in the following way:

S. Schwartz Columbia University, School of Public Health, PH18 - 332 Epidemiology, 630 West 168th Street, New York, NY 10024, USA

357

There was once a king who decided to go to war against an adjacent kingdom. He knew that to win the war he had to locate the best archer in his kingdom. He and his entourage went throughout the country looking for this archer and were just beginning to despair of nding anyone up to their standards of perfection when they passed through a eld where they saw a fence that went on for miles. A target was painted on the fence every few feet, and smack in the middle of each and every target was an arrow bullseyes as far as the eye could see. The kings heart was lightened he had found his archer. Upon making inquiries the king was directed to the local library where the archer, named Harold, was reported to be doing research. When the king saw Harold, spindly arms and legs dangling from a thin body, crunched over a book that he read through thick glasses, he was skeptical that he had the right person. ``Harold,'' he asked, ``are you the archer who makes bullseye after bullseye on the fence in the outer eld?''. When Harold replied that he was, the king smiled and said, ``Please, Harold, tell me the secret of what makes you such a successful archer.'' Harold replied ``Sire, it is really very simple. I stand in front of the fence and shoot my arrow. Then I paint the target around it with the arrow at the exact center.'' (adapted from Ausubel 1948). This parable provides a key to understanding the basic disagreement underlying these paired quotations. For most people, the target was the given and the question was, ``How do we t the arrow into the target?'' However, Harold saw another option. For him, the target wasn't a given settled thing. He gave precedence to the arrow and t the target around it. In the same way, one could argue that authors' perceptions of whether depression or mental illness is essentially a nature thing (something that is wrong in the person) or a nurture thing (something that is essentially environmental) has to do with what the person sees as the given. Is it the environment that is seen as unchangeable or good and therefore not in need of change, so that the individual needs to adjust to the environment? Or is the individual the given and the environment seen as requiring adjustment to t his/her needs? Under the former assumption, it is logical to focus on the decits of the individual, under the latter, one would focus on environmental decits. This would seem to be an underlying value issue in the nature/nurture controversy, particularly as it is applied to mental health problems. Despite the fact that the role of values in such debates has long been recognized, the books cited above, which contain rather heated attacks on the opposing position, never mention an underlying value conict as part of their disagreement. Rather, they critique each other's work in terms of unreliability of measures, confounding, shaky causal inference, confusing proxy measures for cause, etc. The only time that values are alluded to is when one accuses the other of bias. Of course, each side

gives lip service to ``interactions'' between nature and nurture, and yet the environmentalists downplay the importance and/or scope of the nature features, while those on the nature side see the nurture issues as ``triggers'' something very low on the totem of causal esteem. That the argument should be couched in purely scientic terms seems odd. When we examine research relating to the nature/nurture issue from an historical perspective, it is all too clear to us that values about what needs changing and what does not, what is thought of as the given, the good and just, has had enormous impact. For example, it is clear that research on the diagnosis of drapetomania a disorder of slaves who have a tendency to run away from their owner due to an inborn propensity for wanderlust (Cartwright 1981) considered slavery as a given and therefore looked to problems in the slaves as to why they would run away. Similarly, research on hysteria a condition caused by women's unpredictable wombs could only have been done if the second class position of women was seen as something ``natural'', good, or dangerous to alter (Ehrenreich and English 1978). Studies attributing race dierences in economic achievement to brain size, could only be done in a context where ``race'' was seen as an immutable biological factor (Thomas and Sillen 1991). Yet, when we look back at these studies, conducted by the eminent scientists of their day, we attribute their errors to a failure to rigorously apply scientic methods; we tend to think that, although it took a while, science and scientic knowledge has prevailed and the truth won out. Science has laid rest to the claim that drapetomania is a disorder, women's destinies lie in their wombs, and ``race'' is a biological concept. Historical accounts of such research endeavors, however, seem to support a more Kuhnian perspective (1970). These controversies were not resolved or uncovered by scientic discovery, at least not in the main part, but rather by political activism and extra-scientic factors. Gould's The Mismeasure of Man (1993), Degler's In Search of Human Nature (1992), and Ronald Bayer's Homosexuality and American Psychiatry provide informative examples and documentation of these processes. This perspective makes one wonder whether scientists in future generations will ask, ``How could those intelligent, honored scientists of the 1990s, who received millions of dollars from the government to do research, have been so blind as to their [whatever the 21st century equivalent of racism, sexism, heterosexism etc. will be]?'' The point is not that there are no important scientic issues to be resolved that are relevant to the nature/ nurture controversy in mental health research. Clearly there are and they warrant serious and sustained attention. The varied and profound ways in which nature and nurture interact and inuence each other is an important and growing area of current investigation. However, here I will focus on the limits of science in dealing with the nature/nurture conict, because it is not something

358

that we usually consider in our day to day research activities. The main argument advanced in this paper is that, within broad limits, the parameter estimates that we use to gauge the relative importance of nature/nurture factors in the etiology of psychiatric disorders are malleable. That is, whether we are talking about heritability estimates, variance explained, odds ratios, or relative risks, the size of the parameter estimate is based, at least in part, on value judgements that we make along the way. In particular, in the nature/nurture issue in psychiatric research, the value of whether it is the environment or the person that is in need of adjustment is critical. Much of this discussion focuses on the nature side of things. This is because in the current climate the specter of biological determinism seems more problematic than cultural determinism. However, the problem relates to either case, when we try to argue, on scientic grounds, for the ontological priority of either nature or nurture. A brief outline is given below of four issues in causal inference salient to the nature/nurture issue where values would seem to be inuential in our interpretation of cause and in the weighing of causal importance: (1) gene by environment interaction, (2) environmental specicity, (3) causes of variation versus causes of factors, and (4) reication of measures. After discussing each in turn, I will try to illustrate the value component in two research problems life events and depression and individual factors in violence.

Areas in causal inference where values play a role

Gene/environment interactions This issue is illustrated most clearly in our understanding of heritability estimates, estimates derived from twin or adoption studies that parcel the variance in an outcome into genetic and environmental components. When we hear statements like ``schizophrenia is 60% genetic and 40% environmental,'' there is a tendency for us to assume that this means that genetic factors are more important in the etiology of schizophrenia than environmental factors. This is often taken as a fact given in the condition; the estimate (if established in a methodologically sound way) is accepted as having real meaning, that, is of providing limits on the impact of ``nurture.'' To take heritability estimates literally, however, and assume that a heritability estimate of 60% has any real meaning above and beyond stating that genetic factors play a role, masks the signicant judgements, indeed value judgements, that lie behind these estimates. When heritability estimates are derived from twin studies, for example, they include not only ``straight genetic eects,'' but also certain types of gene/environment interactions. These gene/environment interactions include all of the ways in which genotype inuences the environments that people seek out, the aspects of their environments that they value, and the ways in which

others in the individual's environment react to the genotype. These eects can be conceptualized as genetic eects only when the environment is seen as given and unchanging or only in a context where these environmental potentials are ubiquitous. The most frequently invoked concept of gene environment interaction in psychiatry is the stress-diathesis model genetic vulnerability to a disorder (for example depression) or a precursor to the disorder (for example mastery orientation) interacting with environmental factors (for example life events) to lead to disorder. In this conceptual model, the genetic eect is seen as related to the outcome itself, that is, genes ``for'' depression or precursors of depression. While those interactions may be important, there are also gene/environment interactions where the genes may have no meaningful relationship to the outcome. In some cases, this is obvious, in others not so obvious. For example, there may be a signicant heritability to criminal behavior that derives from a genetic component to IQ. However, IQ per se may have nothing to do with a propensity for crime. Rather, in an opportunity structure where monetary reward is tied to school achievement, children who don't do well in school may nd few opportunities for a lucrative straight career and make an alternative career choice. Is this genetic eect meaningful? Yes, if you think that it is right, fair, and just for monetary reward to be tied to school achievement. If you don't, however, the genetic eect is meaningless except as a trigger for social factors to be set into motion. Similarly, a proportion of the variance in socio-economic status mobility, and indeed of depression, may be genetic through physical appearance. To the extent that ``good looking'' people are seen as more competent, and more desirable to have around, they may have better socio-economic status and therefore lower depression. This genetic factor would be included in the heritability estimate for depression. Environmental specicity A related point is that estimates parceling nature/nurture inuences are environment specic (Plomin 1990). If you change the environment, the heritability estimate changes as well. Suppose you have two species of corn that are genetically distinct that you plant in the same eld. One produces 50% more corn on average than the other. Since the environment in this case was the same, it is clear that the variance in output is mainly due to genetic characteristics of the corn. However, when we look at the mechanism through which the gene aects the corn's output, it turns out that the corn with the superior production has a gene that suppresses a protein that is appealing to grasshoppers. The genes improve output by decreasing the proportion of the crop eaten by grasshoppers. Therefore, this heritability estimate was applicable only in an environment containing a signi-

359

cant grasshopper population (Jencks 1980). Furthermore, the size of the estimate would depend on how hungry and how mobile the grasshoppers were if they were very hungry and lazy, they may eat some of the less attractive crop due to these other factors. Thus, the estimate of the relative importance of nature/nurture factors would change based on the environment in which the study was done. In addition, this example points out the disjuncture between apparent causal importance and appropriate intervention. For despite the fact that the heritability estimate is huge, a successful intervention may be entirely environmental. A high heritability estimate tells you nothing about either the importance of environmental causal factors, nor about the potential ecacy of environmental interventions (Gould 1993). The value issue enters when you decide upon a reasonable environment to examine the nature eect. Explaining the variance does not explain the existence or size of the factor When we try to identify the causes of a condition, we usually begin by examining how that factor varies in a population. If you want to know the causes of depression, you look at how depression is distributed in the population which individuals are aected and which are not. However, as Lieberson (1985) points out, what we learn from this is the cause of the variation of this factor in a population, but not necessarily the cause of the factor itself or its magnitude. Similarly, what causes the dierence between individuals may not cause the dierences between groups of those individuals (Rose 1985). One example regarding heritability estimates would be height. Variance in height is largely genetic heritability estimates are in the range of 90%. Yet there are substantial intergenerational dierences in average height that are likely caused entirely by environmental factors (Gould 1993). Depending on the level of analysis chosen to examine the issue, the heritability estimate will be dierent. Indeed, depending on the level of analysis at which a question posed, the cause may be dierent. Even if, as a group, for example, a particular social status is related to risk for depression, when looking for the cause of depression in any individual person, that social status may be totally irrelevant. Under some circumstances, however, answers from either set of analyses can provide clues for cause and/or intervention. But what you think is an important cause and, indeed, what your parameter estimate will be depends on your level of analysis. Reication of our measures All parameter estimates are, of course, measurement dependent, that is, our choice of how to measure our variables inuences the size of the heritability estimate.

While this applies equally well to scales and diagnoses, independent and dependent variables, here it will be discussed only in terms of diagnostic outcomes. In the course of our struggles to develop reliable and valid measures, we often forget that what underlies our measures are constructs that we have developed for particular purposes to examine particular issues. There is a considerable literature discussing the underlying meaning of psychiatric diagnoses, with some general consensus that the basis of a psychiatric disorder is that there is a dysfunction in the person regardless of the source of this dysfunction, there is now something awry in the person. For example, depression is seen in essence as mood disregulation that is, the person's mood responses are abnormal. This ``mood abnormality'' becomes codied in diagnostic criteria. However, after attempts to ensure the reliability and validity of these categories, somehow the basic underlying construct is lost and the criteria themselves become indications of the disorder. The criteria become decontextualized; that is, having these symptoms, regardless of the context, becomes an indication of disregulated mood they become, by denition, abnormal. We may have general agreement about this regarding some symptoms (e.g., suicidal ideation) but not about others (e.g., guilt). To call symptoms the same thing regardless of their context is a decision about standards of behavior we expect from people. That this constellation of symptoms is called ``depression'' and not ``grief'' or ``sadness'' or ``penance'' has signicant implications about heritability estimates, or our understanding of what is inherited. When this issue of the dierences between normal responses to life and disease is discussed in the literature, the language of discovery is often used. That is, the question is asked, ``How can we tell what is a normal response and what is an abnormal response?'' The question might, however, be more appropriately phrased as ``What do we want to call a normal response and what do we want to call an abnormal response?'' We often think that biology will help us resolve this issue; that is, if we nd a genetic defect or some neurotransmitter decit we will distinguish normal responses from disease on that basis. However, the argument is circular. We will only call biological variation a biological decit if the outcome with which it is associated is considered negative and abnormal. If we consider responses normal, variations in serotonin level become normal biological variation related to ``sensitivity, caring, communality'' etc. If we consider responses abnormal this will be called ``decits in serotonin levels'' etc. This becomes clear when we consider genetic and/or biological correlates of sexual orientation. If you think ``homosexual'' orientation is abnormal, the genetic or biological correlates become decits, if you don't, they are normal biological variation. The crucial question is, to what range of environments do we expect individuals to adjust and to what range of human variation do we expect our environments to be hospitable. In addition, we can choose to divide diagnostic categories based on

360

their genetic components leading to disorders that are largely genetic. If, however, we divide the disorders based on levels of neurotransmitters, we can make the genetic component in any disorder smaller. Thus, our choice of principles for categorizing disorders determines the importance of various causal components. The point is not that any particular estimates parceling out nature and nurture components are ``wrong'', but rather that they have no concrete correspondence to something in the world out there they are malleable. Based on what, in any interaction, we consider a meaningful genetic eect, what we consider an appropriate environment in which to ask a research question, whether we're interested in individuals or groups, if we're interested in the cause of a factor's magnitude or its distribution, what measures we decide upon, the causal importance of nature and nurture indicated by our estimates will change. These issues are all based on judgements inuenced by our values. By way of concrete examples, the eect of this problem on research into life events and depression and one example of research on violence are discussed below.

to what extent, life events are important in the etiology of depression. Violence The issues are even more dicult regarding current research on violence, particularly in the area of interindividual dierences in a propensity for violence among high-risk individuals. For example, there has been a study examining the propensity for violent behavior among the siblings of juvenile oenders found guilty of violent crimes and remanded to the criminal justice system. The sample was almost entirely African-American and Latino youth, reecting the incarcerated juvenile population who met the study criteria (Pine et al. 1996). What should be noted about this type of design is that it limits the variability of virtually all social factors related to juvenile delinquency, because all neighborhood, poverty, justice-system, racism, and family factors are held relatively constant. The main variance left to explain resides in factors dierentiating individuals. In such a study it may be that children who have certain biological characteristics (e.g., high blood pressure) or score higher on hyperactivity scales are more likely to commit crimes than their less hyperactive counterparts. One could indeed nd that such factors explain a signicant proportion of the variance in determining who went on to commit a crime. This is not scientically wrong individuals within this sample may indeed be more likely to commit crimes if they are ``hyperactive.'' However, by decreasing most environmental variance, factors of the individual, divorced from social position, are the only kinds of factors that could explain variation. Interpreting ``large proportion of variance explained'' and ``signicant result'' as meaning this is an important cause of violence, is misplaced concreteness. Deciding that examining individual variance in this highly constrained environment is a viable and meaningful place to look for the cause of violence would seem to be a decision guided by values and with considerable value implications. It highlights certain aspects of a process and puts other aspects in the shadows. The construct issue is also crucial in this type of work. For example, after the riots in Los Angeles (LA) in response to an incident of police brutality, as after the inner-city riots in the United States in the 1960s, there was a urry of initiatives to look at individual characteristics in violent behaviors (Hupert 1994). The logic was that since not everyone in the community participated in the looting etc., only some people did, there must be variation in the propensity to commit violent acts. This assumes, of course, that you understand the LA riots as ``LA riots'' and not as ``LA rebellion'' or ``LA uprising''. It assumes that being violent in these conditions is something that is unreasonable, something that represents a problematic response on the part of the individual. Surely this is true from the victims' per-

Examples of the value component in psychiatric research

Life events The issue of whether or not life events cause depression poses very complicated conceptual issues. If we consider depression a thing, out there in the world, be it a disease or misery, then the issues are methodological. If however, we conceptualize depression contextually, then the very factor that we are looking to as a causal factor is part of the context. Take, for example, a women who has a well paid, fullling job, a loving egalitarian spouse, two children in Harvard Medical School who devote their free time to Doctors without Borders, and who is down and blue, with six symptoms of depression for 8 weeks. It would be hard to imagine that she has the same thing or that the same scale score means the same thing for her as for her counterpart whose husband died in a job-related accident, whose child was injured when she was driving and didn't put a seat belt on the child, who got red from her job due to sexism, and who was recently diagnosed with cancer. To use a medical analogy, high blood pressure has a dierent meaning depending on the context Has the person just been running? Is the person nervous about seeing a doctor? etc. The same symptoms mean something dierent. It may be that we decide that this symptom level is problematic in and of itself, regardless of the context; that is a judgement about what responses we expect from people. Do we want to consider genetic and/or biological components to these reactions vulnerabilities to depression? What does this say about what we, as a society, think about appropriate emotional response? Are these symptoms always bad? Can they have positive functions? These value judgements critically inuence whether or not, or

361

spective and from the point of view of general law and order, but a parallel can be drawn with drapetomania. Not all slaves ran away. There was variation about who would runaway. Maybe it was related to risk taking, closeness to parents, tolerance for uncertainty and biological correlates of those factors. Therefore, when comparing runaway slaves with those who stayed, one could theoretically nd biological vulnerability for that behavior. From the perspective of the society that thought slaves should stay, they belonged to their owners, this could theoretically be a scientically validated position. From the point of view of the mother of slaves who might run away, whose child then had a good chance of being tortured or killed, biological intervention to keep her child from running away might be a desirable outcome. From our perspective, the whole question behind the diagnosis and research is morally repugnant. Our decisions about whether we consider aggressive behavior as violence or rebellion, an expression of individual problems or a defective environment, will determine what we consider a cause.

question, they often claimed greater expertise in determining the outcome than non-scientists. 5. It helps us to see that our choice of research design and methods inuence the answers we can nd for our questions. The more restrictive or ubiquitous the environment in which we look for a cause, the larger genetic and other individual-level variations will loom. Conversely, the more genetically homogeneous the population, the more important the environment will appear. 6. Acknowledging the value premises that underlie our work helps us to take responsibility for the ethical implications. We can't hide under the cloak of scientic objectivity. Gould (1993, p 23) sums up the benets of letting values enter the debate as a legitimate factor. Science cannot escape it[s] curious dialectic. Embedded in surrounding culture, it can, nonetheless, be a powerful agent for questioning and even overturning the assumptions that nurture it Scientists can struggle to identify the cultural assumptions of their trade and ask how answers might be formulated under dierent assertions. Scientists can propose creative theories that force startled colleagues to confront unquestioned procedures. But science's potential as an instrument for identifying the cultural constraints upon it cannot be fully realized until scientists give up the twin myths of objectivity and inexorable march toward truth. Particularly in the area of nature/nurture issues, where rigid positions have caused so much damage in the past, and in the area of mental health research, which is so intimately tied to the deep and powerful questions about how we conceptualize human nature and the kind of world in which we want to live, it is important for us to spend some time considering and arguing about the basic value issues that underlie our work.

Conclusion
The basic point is that the causal chain is long and complex. Where in the causal chain we see a link as being important, explaining a large proportion of the variance or a good place for intervention, depends on our values about what we think should and can be changed and what behavior we expect from people under what conditions. Examining our values in the nature/nurture debate could have the following benets: 1. We can focus our debates more sharply by recognizing basic value assumptions that may underlie our positions. 2. It allows for greater exibility. Causes become tied to context, purpose, and usefulness. Once we see our estimates in terms of usefulness rather than truth, we naturally ask, ``Useful for whom under what conditions?'' It allows for highlighting dierent causes for dierent purposes. It also allows for multiple understandings of the same outcomes. We can do research within a context that sees schizophrenia as a disease, for those who nd that conceptualization useful, without ignoring the interests of those who see a disease framework as constricting for understanding this set of symptoms. 3. Change can happen more quickly. When you are not dealing with ``the truth'' but ``a truth'' there is less resistance to change. 4. It forces us to give up our claim to special expertise when it is not appropriate. For example, when reading historical accounts of the debate on the removal of homosexuality from the Diagnostic and Statistical Manual (Bayer 1981), it is clear that when researchers did not recognize what was a value

References
Ausubel N (1948) A treasury of Jewish folklore. Crown, New York Bayer R (1981) Homosexuality and American psychiatry: the politics of diagnosis. Basic Books, New York Breggin PR (1991) Toxic psychiatry. St. Martin's Press, New York Cartwright S (1981) Report on the diseases and physical peculiarities of the negro race. In: Caplan AL, Engelhardt HT, McCartney JJ (eds) Concepts of health and disease: interdisciplinary perspectives. Addison-Wesley, New York, pp 305 325 Degler CN (1992) In search of human nature. Oxford University Press, New York Ehrenreich B, English D (1978) For her own good: 150 years of the experts' advice to women. Anchor Press, Garden City Gottesman II (1991) Schizophrenia genesis: the origins of madness, 3rd edn. Freeman, New York Gould SJ (1993) The mismeasure of man. Norton, New York Hupert N (1994) Reducing urban violence: a critique of the new public health approach. Pharos 57: 610

362 Isaac RJ, Armat VC (1990) Madness in the streets: how psychiatry and the law abandoned the mentally ill. Free Press, New York Jencks C (1980) Heredity, environment, and public policy reconsidered. Am Sociol Rev 45: 723736 Klein DF, Wender PH (1993) Understanding depression: a complete guide to its diagnosis and treatment. Oxford University Press, New York Kuhn TS (1970) The structure of scientic revolutions. University of Chicago Press, Chicago Lewontin RC, Rose S, Kamin LJ (1984) Not in our genes: biology, ideology and human nature. Pantheon, New York Lieberson S (1985) Making it count: the improvement of social research and theory. University of California Press, Berkeley Mirowsky J, Ross CE (1989) Social causes of psychological distress. de Gruyter, New York Pine DS, Wasserman G, Coplan J, Staghezza-Jaramillo B, Davies M, Fried JE, Greenhill L, Shaer D (1996) Psychosom Med 58: 342353 Plomin R (1990) Nature and nurture: an introduction to human behavioral genetics. Brooks/Cole, Pacic Grove Rose G (1985) Sick individuals and sick populations. Int J Epidemiol 14: 3238 Thomas A, Sillen S (1991) Racism and psychiatry. Carol, New York