In a pioneering study using this method, Robert Kendell and Jane Gourlay studied the symptoms of nearly 300 patients.8 They used a computer to assign negative scores to schizophrenia symptoms such as delusions and hallucinations and positive scores to mood symptoms such as abnormal gregariousness, manic speech and depressed behaviour. When the scores for each patient’s symptoms were added together, it was found that most patients fell in the middle range, close to zero, indicating that there is a continuum between schizophrenia and manic depression, rather than two separate illnesses. Subsequent studies using variations on the technique,9 although producing results that have varied in some respects from those obtained by Kendell and Gourlay, have all failed to find the neat dividing line between schizophrenic and bipolar symptoms that was the one of main assumptions of the Kraepelinian paradigm, and which, for this reason, was described by Kendell and Gourlay as one of the ‘cornerstones of modern psychiatry’.
Which symptoms go together?
Although psychotic symptoms do not fall into two main categories as supposed by Kraepelin, this does not mean that the relationships between them have no structure. Symptoms may cluster together in ways that were never envisaged by the father of modern psychiatry, perhaps in ways that cut across the major categories described in modern diagnostic systems. To discover whether this is the case, researchers have resorted to another statistical procedure, known as factor analysis. In the initial phase of this kind of analysis, information about symptoms is collected from a large number of patients and a matrix (table) of correlations is calculated showing the relationships between all possible pairs. (If two symptoms are highly correlated this means that they typically occur together.) Further analysis of the matrix then leads to the discovery of different clusters or groups of symptoms, known as factors. One of the main goals of factor analysis is to discover the minimum number of factors required to describe the data. If only one factor is discovered, there is only one cluster of symptoms, and patients can be placed on a single dimension running from well (no or very few symptoms) to very ill (many symptoms). If two or more factors are discovered, it is likely that there are two or more symptom clusters and that patients vary according to their severity on each.
Factor analysis was developed in the 1920s by psychologists who were interested in the relationships between different measures of human intelligence. They wanted to know whether there is just one type of general intelligence, or several different intellectual faculties. As the invention of the computer lay some way in the future, the necessary calculations were carried out by teams of clerks, who would often take several weeks to complete each analysis.10
The technique was first applied to the problem of psychiatric classification by an American psychologist, Thomas Moore, who published his findings in 1930.11 From data on the symptoms of a group of 367 seriously ill patients, Moore was able to extract no less than eight symptom clusters. Hallucinations and delusions seemed to occur together (a finding that has been consistently repeated in later research). However, Moore found two types of depression, two types of mania, a syndrome of uninhibited behaviour, a syndrome of catatonic behaviour, and a syndrome of cognitive deficits.
Following Moore’s work, interest in the use of factor analysis in psychiatry has waxed and waned as the technique has been progressively refined.12 However, the method has been used extensively in the past ten years, partly in response to an influential theory proposed by the British psychiatrist Tim Crow. In a paper published in 1980, Crow broke away from the Kraepelinian paradigm and suggested that there are two different types of schizophrenia.13 According to the theory, the main symptoms of Type I schizophrenia are hallucinations and delusions, which are caused by a biochemical imbalance in the brain. Borrowing from the literature on neurological illnesses, Crow referred to these as positive symptoms (because they consist of experiences and behaviours that we would prefer to be absent).* The symptoms of Crow’s Type II schizophrenia consist of apathy, emotional flatness and social withdrawal. Crow called these negative symptoms (because they involve the absence of desirable behaviour) and argued that they are caused by a progressive atrophy of certain areas in the brain. Although this theory made bold use of evidence from biological studies of brain structure and function, it was essentially speculative and was not based on analyses of symptom data. However, the theory leads to some fairly obvious predictions about what should emerge from a factor analysis of psychotic symptoms.
In fact, the first study that attempted to test Crow’s theory obtained not two symptom clusters but three. Analysing data from a group of chronically ill schizophrenia patients, British psychiatrist Peter Liddle obtained one factor consisting of positive symptoms and one consisting of negative symptoms. The unexpected third factor consisted of symptoms of cognitive disorganization, for example disturbed speech and problems of attention.14 These findings have since been replicated by many other researchers,15 although, to complicate matters further, some have argued that the three clusters can be further subdivided.16
Of course, it is impossible to conclude much about psychosis in general from these findings, as they were obtained from groups of patients who had already been selected for a diagnosis of schizophrenia. What we would like to know, therefore, is whether the same three dimensions can be used to describe the symptoms of psychotic patients with other diagnoses. There is at least some evidence that they can. For example, in a recent study published by researchers at Harvard Medical School, factor analysis was used to investigate the symptoms of psychotic patients diagnosed as suffering from schizophrenia, major depression and bipolar disorder according to DSM criteria. Although there was some variation in the results when the analyses were restricted to the individual diagnostic groups, the overall picture that emerged was consistent with the three-factor model.17
In this account, I have simplified the steps involved in carrying out a factor analysis, and underplayed the difficulties involved. The number of factors revealed can be affected by decisions about which symptoms to include in the analysis (if important symptoms are not recorded the results are likely to be misleading)or about which patients to study (the broader the sample, the more likely that the results will be meaningful). Moreover, a number of different methods of carrying out the analysis employ different rules to determine the number of factors, so that researchers, finding that a particular analysis is uninterpretable, are sometimes tempted to experiment with other methods before selecting the one that yields the most credible results. Despite these opportunities for subjectivity, and similar limitations affecting discriminant function analysis, it is striking that analyses of patients’ symptoms have never provided clear support for Kraepelin’s theory. Indeed, at first sight, the results obtained from the two statistical approaches we have considered appear to deviate from Kraepelin’s system in opposite directions. The findings from discriminant function analysis imply that we should collapse all the psychotic disorders into one, whereas the findings from factor analysis seem to suggest that we need to subdivide the psychoses into many more components than those proposed by Kraepelin. This apparent paradox is resolved when we remember that the symptom dimensions revealed by factor analysis seem to describe efficiently the symptoms of both schizophrenia and bipolar patients. Therefore, although the distinction between schizophrenia and bipolar disorder does not seem to survive examination with these techniques, we can be fairly confident that all psychotic symptoms cannot be explained in terms of a single underlying disease process.
What Families Can Teach us about Psychiatric Diagnoses
A second function of a diagnosis is to group together people whose problems are likely to have a common aetiology. Indeed, as we have already seen, the most common strategy employed in psychiatric research – comparing those with a particular diagnosis with people who are psychologically healthy – assumes that those with the diagnosis have something in common that is of aetiological significance. Although discussion of many of the factors thought to in
fluence the development of psychosis must wait until later chapters, there is one aetiological factor that has been thought to have special significance for psychiatric classification, and which we will therefore consider here. This will require a short digression into the difficult world of genetics, an effort that will save us time later in the book.
On the assumption that psychiatric problems are, at least in part, inherited, some researchers have attempted to address the validity of diagnoses by observing whether they breed true in families. If this is so, people who suffer from a particular disorder are likely to have many relatives who suffer from the same disorder, but few relatives who suffer from any other kind of mental illness. Some researchers conducting this kind of work have argued that genetic studies provide an unassailable riposte to those who question the Kraepelinian paradigm. For example, the American psychiatrist Seymour Kety once commented that, ‘If schizophrenia is a myth, it is a myth with a strong genetic component.’18
As we have already seen, genetic research into psychiatric disorders has suffered from an unhappy history. Many of the German pioneers in the field held political views that were in sympathy with the Nazi movement. The first genetic studies of schizophrenia were carried out by Ernst Rüdin, who later served with Heinrich Himmler on a committee that, in 1933, drafted legislation enabling the compulsory sterilization of psychiatric patients.19 His student Franz Kallman advocated the sterilization, not only of mentally ill people, but also of their relatives.* Ironically, Kallman was forced to move from Germany to the United States in the 1930s because he was half-Jewish. Given these unpromising beginnings, it is understandable that some critics have viewed all genetic research into madness with suspicion.20
Before proceeding further, it will be useful to consider the three main research strategies that psychiatric geneticists have followed when attempting to determine the contribution of heredity to mental illness. All exploit the fact that we share alleles (particular versions of genes) with other individuals according to how closely we are related to them. For example, we obtain half of our alleles from each of our two parents who, in turn, obtained half of theirs from each of their two parents. Half of our alleles are therefore shared with each parent and a quarter are shared with each grandparent. As ordinary brothers and sisters are each a product of the random combination of alleles from the same two parents, it turns out that sibs, on average, also have half their alleles in common.
Matters become slightly more complicated when it comes to twins. Dizygotic (sometimes called non-identical or fraternal) twins are born after two eggs, appearing at the same time in the uterus, are fertilized by two sperm. They are therefore sibs who happen to be born at the same time and, on average, have half their alleles in common. About 50 per cent of dizygotic twin pairs consist of a boy and a girl, about a quarter consist of two boys and about a quarter consist of two girls. Even when the twins are of the same sex, they often look quite different from each other, although this is not invariably the case. Monozygotic or identical twins, on the other hand, are born after a single zygote (fertilized egg) splits in two and both halves develop separately in the uterus. These twins are genetically identical and are always the same sex, although they may still be differentially affected by factors in the inter-uterine environment.*
The simplest method of exploiting these differences in genetic similarity is the family study. In this kind of research, cases of an illness are traced within large families. If the illness is mostly inherited, close relatives of an affected member should be more likely to suffer from the illness than distant relatives. The second widely used strategy is the twin study. If genes make a major contribution to an illness, the probability that an identical twin of an affected individual will also suffer from the illness should be very high and the concordance rate for fraternal twins should be lower. Of course, twins belonging to the same family may be raised similarly and this may also affect concordance rates. To control for this possibility, geneticists sometimes carry out adoption studies. For example, children of affected parents who have been adopted away at birth and raised by normal families can be compared with the adopted-away children of unaffected parents. If genetic factors play a role in the aetiology of the illness, the children of the affected parents should show higher rates of illness than the children of the unaffected parents, even though both groups have been raised apart from their biological families. An alternative adoption study strategy starts with the adoptees but follows the same logic. In this case, the researcher attempts to trace the biological parents of adoptees who have become ill, in order to compare them with the biological parents of adoptees who show no evidence of psychiatric illness.
It is fair to say that each of these strategies has proved to be fraught with difficulties. For example, the poor reliability of psychiatric diagnoses means that there must be some uncertainty about the diagnoses assigned by genetic researchers. (This problem was most evident in a famous series of adoption studies carried out in Denmark by Seymour Kety and his colleagues, who used such a broad definition of ‘schizophrenia’ that almost anyone who was slightly eccentric was regarded as mentally ill.21 Inevitably, some commentators have dismissed the results from these studies as almost worthless, although they are still cited in textbooks of psychiatry.)22 Surprisingly, it can also be difficult to decide whether twins are identical or fraternal without genetic testing, which has only become available in recent years.
When the individuals who are examined in a genetic study are relatively young, geneticists make mathematical corrections to their data to allow for the possibility that some unaffected individuals will become ill later in life (in effect, they guess how many of the well individuals will become ill as they get older). Confusingly, there are also several ways of estimating concordance in twin studies, some of which give higher values than others. The commonsense approach is the pairwise method. For a group of twins in which at least one member of each pair is affected, the pairwise concordance rate is the percentage of pairs in which both are affected. For example, if in three out of ten twin pairs both are affected, but in the remaining seven only one is affected, the pairwise concordance rate is 3/10 or 30 per cent. The less commonsensical probandwise method gives higher values and is therefore preferred by geneticists. This is calculated as the proportion of twins who have the illness who have an affected twin. In the above example, six twins have an affected twin (i.e. both members of each concordant pair are counted as having a twin who is also ill) whereas only seven do not, leading to a concordance rate of 6/13 or 46 per cent (an increase in concordance of more than a half).*
A further difficulty with twin studies is that twin status itself is associated with slightly abnormal development. Twins (whether fraternal or identical) have a higher risk of congenital difficulties and, on average, reach developmental milestones slightly later than singletons.23 It is also possible that they are treated differently from singletons by their parents, who will certainly face more than the usual difficulties when struggling to provide them with adequate attention. Whether these subtle influences might, in extreme circumstances, culminate in a higher risk of psychosis remains unclear. In an analysis of twin data from schizophrenia studies, Steven Rose, Leon Kamin and Richard Lewontin (distinguished neuroscientist, psychologist and geneticist, respectively) calculated that concordance rates for fraternal twins were significantly greater than those for non-twin sibs, an observation that clearly points to the adverse effect of being a twin.24 However, attempts to find out whether twins have an increased liability to psychosis compared to the general population have reported inconsistent results.25
Given these problems, it is remarkable that genetic researchers have been almost blind to obvious environmental influences on the people they have studied. The case of the Genain quadruplets, four genetically identical women believed to be suffering from schizophrenia, provides an astonishing example of this kind of myopia. American psychologist David Rosenthal and his colleagues estimated that t
he probability that identical quadruplets would be affected by the illness was one per one and a half billion births.26 The pseudonyms given to the quadruplets to maintain their anonymity give some indication of where the researchers were coming from. Genain was derived from the Greek for ‘dreadful gene’ and the first names selected for the unfortunate women, Nora, Iris, Myra and Hester, when written in their birth order, spelt out the initials of the US National Institute for Mental Health.
Despite his obvious bias, Rosenthal’s publications provide compelling reasons for questioning his exclusively genetic account of their difficulties. First, there were considerable differences in the psychiatric problems experienced by the women. Myra, who had married, suffered from much milder symptoms than the others, and preferred minimal contact with the investigators. Follow-up studies carried out in 1981 when the quadruplets were 51 years old27 and in 1996 when they were 6628 showed that Myra and Iris performed much better than Nora and Hester on a range of psychological tests, with Myra’s performance falling mostly in the normal range. Moreover, Myra and Iris showed much less deterioration than Nora and Hester when their medication was stopped.
Madness Explained Page 10