Twin Research in Psychiatry

A Critical Review

Journal of Neurology and Psychiatry, Vol. 1 n. 3, 1938


General Principles and Theoretical Basis

It is not attempted in the following review to cover the whole of the now enormous literature on twin research, even as far as it concerns psychiatry. Only works of statistical or theoretical importance will be dealt with in a critical spirit. Extensive reviews of twin literature from the general aspect are given by Newman (1923), Siemens (1924), Dahlberg (1926), von Verschuer (1927), Siemens (1932), von Verschuer (1932), Luxenburger (1932c), von Verschuer (1933), Newman, Freeman, and Holzinger (1937). Reviews of the literature as far as it specially affects psychiatry are given by Siemens (1924), Lange (1929), Luxenburger (1929, 1930a and b, 1932a, 1933, 1935), and Conrad (1937). No account will be given of the now very large number of papers dealing with work on twins of purely psychological nature. It is, however, time that these works, which are of very varying merit, were submitted to critical consideration by a competent reviewer.

   Twin research in psychiatry, as in other fields of human biology, is founded on a particular theory, that there are two kinds of twins, the monozygotic and the dizygotic. Both members of a pair of monozygotic twins are supposed to have exactly the same hereditary equipment. The members of a pair of dizygotic twins are supposed to be no more like each other than ordinary siblings. If this theory is correct, we are enabled to observe the results of a controlled experiment performed by nature. For if the hereditary endowment is the same, any real differences observed in monozygotic twins should be due to the environment. The differences in dizygotic twins will be due to differences both in the environment and in the hereditary equipment. This theory is probably correct (though it cannot be taken as proven), although its theoretical embryological basis has become more dubious and speculative than it seemed at first.

   The distinction between monozygotic and dizygotic twins was originally based on the observation that some twins were born with separate placent and membranes, some with these structures in common. The latter variety were found to be always of the same sex, and as they grew up they grew increasingly like one another, so like in fact that it was often difficult to distinguish them. From a statistical point of view it was found that there was an excess number of twins born of like sex, and were one to deduct double the number of opposite‑sexed twins from the total, the remaining proportion represented about the same figure as the proportion of monochorial or of similar twins. Further, it has been found that this group of similar twins, if selected on grounds of their similarity to one another in a number of characteristics commonly held to be hereditary (such as eye colour, stature, etc.), would be found to be similar in a vast number of other hereditary characteristics. A similar phenomenon would not be observed in a group of similarly chosen opposite‑sexed twins. It was clear, therefore, that this group of twins formed a special group in which the hereditary constitution could be taken to be very nearly if not absolutely the same in both members of the pair. For a very long period the diagnosis of these two groups was by means of the membranes only. As, however, the placental membranes of adult twins are not easily to be rescued from the incinerator to which they have many years before been consigned and as any reliable history of the formation of those placente is often quite unobtainable, the method of diagnosis by similarity has gradually replaced the diagnosis by membranes, and is now to be regarded as the method of choice. For the diagnosis by membranes, even when carried out with the most exact care, can no longer be regarded as reliable. Smith (1930 and 1931) has described almost certainly uniovular dichorial twins and Curtius (1930) and Lassen (1931) have made a systematic examination of the point. Between them they examined 86 twin placente by macroscopical, microscopical, and Roentgenological methods, and carried out a very careful anthropological investigation of the like‑sexed twins who could be followed up. Lassen gives their results in the following table

    They were unable to find any biological difference between the monochorial and the dichorial similar twins, from which it follows that the presence of one or two choria is more or less an accidental phenomenon and without much significance for the diagnosis of constitutional similarity or dissimilarity. As they found among 14 similar twins no less than five dichorial pairs, they conclude that dichorial uniovular twins are not at all uncommon ; and this opinion is borne out by the fact that the percentage of monochorial twins has been pessistently found to be smaller than the percentage of uniovular twins calculated from the sex distribution. Curtius (1930) considers that his results indicate that the method of diagnosis by membranes is unreliable and should be discarded. The occurrence of monochorial dizygotic twins appears to be much less common, if it occurs at all. Von Verschuer (1932a) quotes two cases of opposite‑sexed monochorial twins from the literature, and has himself described (1925) a case of monochorial dizygotic same‑sexed twins. The evidence for monochoria rests on the statement of an obstetrician. In view of the rarity of this phenomenon, even if it does occur, an exact diagnosis of monochoria, supported by microscopical examination, must be taken as very strong evidence of uniovularity. There is no particular difficulty from the embryological point of view in explaining the occurrence of dichorial uniovular twins, as long as the point of separation of the twin embryos is considered to be capable of occurring before the development of the chorionic anlage. Siemens (1924), who was to a large extent the author of the modern method of diagnosis by similarity, has stigmatized the diagnosis of ovularity by membranes as a petitio principii. It must, however, be pointed out that the same logical criticism applies to the diagnosis by similarity.

   However, the use of the similarity method though faulty in logic is justified in fact. For though the diagnosis cannot be made a matter of certainty, it may be made with such a high degree of probability that there is no longer room for reasonable doubt. Some investigators have examined their twins with a plurality of refined methods which would be beyond the capacity of any single individual and only conveniently carried out by an anthropological institute. Such methods of reducing the chance of binovularity from one in millions to one in billions can only be necessary in exceptional cases. A case of mongolism in one only of monozygotic twins might well be examined by such elaborate methods, for instance : for so far anything approaching proof of such an event is lacking. For practical purposes, however, the diagnosis of uniovularity does not offer serious difficulties in the great majority of cases. A history of very striking similarity in appearance, supported by a short and convenient range of anthropological tests and measurements, including the very useful finger‑print method, is adequate in the general run of cases. Nevertheless, it must be emphasized that the satisfactory diagnosis of ovularity is the necessary foundation of any reliable twin research, and its lack in the great majority of the earlier published cases very seriously detracts from their value. The method of diagnosis by membranes should not now be used unsupported by data from the similarity method.

   At present there is little certain knowledge of the mechanism by which twins are produced. The commonly held opinion is that dissimilar twins are binovular and have arisen from the accidental production of two ova at the same time while similar twins are uniovular and have arisen from a single ovum by fission of the blastoderm, plural gastrulation, or by fission of the embryonic axis. In an interesting paper Curtius and von Verschuer (1932) investigated the fre­quency of twin‑parenthood in the families of 931 twins. They came to the conclusion that there was a single recessive factor for twinning, which might express itself in the homozygotic state as the parenthood of twins, was trans­mitted and exhibited equally by the male and the female, and showed no predilection for either the monozygotic or the dizygotic type of twin. The chance of this tendency to the production of twins manifesting itself at any particular birth they calculate (after Dahlberg) as 0•06. Their paper, as well as that of Meyer (1932), which goes in a similar direction, has been subjected to a destructive criticism by Lenz (1933), who considers that the capacity of twinning must be held to be a universal human characteristic and its manifestation governed by purely environmental factors. It would seem to the reviewer that the existence of a non‑universal hereditary tendency to twinning has, however, not yet been disposed of, and that the theory of a recessive inheritance which may manifest itself in the homozygotic state in the actual twin pair, a possibility which is especially suggested by Meyer's paper, has not yet received adequate consideration. As a result of Lenz's criticism von Verschuer has subsequently (1933) retreated from his former position, but he still holds to the possibility of a uniovular origin of dizygotic twins, a rather far‑fetched theory put forward by himself and Curtius to explain their finding that the father could be re­sponsible for the production of dizygotic twins. The theory briefly is that the hereditary disposition to twinning manifests itself in a splitting tendency (Spaltungstendenz), which may be exhibited by the ovum or given to it by the sperm either after or before the final maturation division. They suppose that in the last of these cases, the ovum, before it has given up the second polar body, as a result of the Spaltungstendenz given it by the spermatozoon splits off a polar body which is as large and fertilizable as the remaining fertilized ovum, and that this polar body is itself fertilized by another sperm. The result is twins of differing constitution (e.g. perhaps differing sex) but uniovular origin. The possibility of such an event occurring had already been suggested by Oluf Thompson (1929). As Thompson stated, the result of this process would be two twins of identical maternal but differing paternal heredity, and the theory might be used to explain the occurrence of very similar opposite‑sexed twins. Curtius' and von Verschuer's generalization would seem to be unjustified, however, as there is no evidence that such twins exist in any number, if at all. I cannot see that in their paper the authors recognize that this peculiar variety of twin would be the necessary result of the process they suppose to take place, or that they take note of the general opinion of embryologists that in man, as in animals, the division of the primary oocyte is the reduction division, that of the secondary oocyte is an equational division. The theory is fundamentally that put forward by Dahlberg, the " Spaltungstendenz" corresponding to Dahlberg's "doubling tendency." Neither theory provides any explanation for the increasing frequency of twinning with increasing age of the mother, and all the authors are constrained to suppose that polyovulation also occurs and is independent of any hereditary factor. The theory that there is a common hereditary disposition for the production of monozygotic and dizygotic twins, a theory to which von Verschuer appears to adhere, runs counter to the finding that in Japan there is a normal uniovular‑twin birth‑rate but a very low dizygotic­twin birth‑rate.

   It is commonly assumed that similar twins are identical in respect of their hereditary equipment. This cannot by any means be taken as proven. It is an admitted theoretical possibility that in the division of the fertilized ovum there might be an unequal division of the chromosomes. Such an event has been shown to occur in the reduction division in Drosophila, resulting in in­dividuals with more or less than their proper quota of chromosomes, or in other cases with a fragment of a chromosome being additional to the normal quota in one individual, deficient in another. There is no cogent reason why this should not also occur in the somatic division, and in man. Furthermore somatic mutations are known to occur in a variety of different organisms, and if such a mutation were to occur in one of the two cells produced by the primary fission of the fertilized ovum it would produce a true hereditary difference in the two resulting monozygotic twins. These theoretical possibilities are, however, of little practical importance : for even were they shown to occur in man, they would be events of such rarity that statistical results obtained by the examination of large numbers of twins would be little affected by them.

   A more important objection to the assumption of the identical hereditary constitution of monozygotic twins has been raised, among others, by Dahlberg (1926) and Newman (1923, 1937). These authors have emphasized the importance of asymmetrical phenomena occurring in monozygotic twins, and drawn attention to the frequency of "mirror‑imaging," or of discordance in respect of such characters as handedness, unilateral mcvi, asymmetrical characters of finger and palm‑prints, etc. Both authors suppose there is an "asymmetry mechanism," specially affecting uniovular twins, which can be productive of real differences and cannot be called environmental in nature. Newman supposes that uniovular twins are derived from the partially developed right and left halves of a single embryo, and that the degree of asymmetry re­versal depends on the lateness of the time of separation of the two halves to make new individuals. It seems, however, that this reversal of asymmetry occurs, especially as regards right‑ and left‑handedness, in fraternal as well as identical twins. The figures provided by different workers are very conflicting. From Newman's compilation of his own figures and those of Dahlberg, von Verschuer, and Hirsch, given in his book (1937), and from the quotations from the literature and the original data supplied by Wilson and Jones (1932) the following table has been compiled

   Von Verschuer (1932) provides a mass of data on the occurrence of asym­metries in human twins, and comes to the conclusion that the distribution of the asymmetrical character between the two twins is a chance phenomenon. If the reversal of the asymmetrical feature is due to the process that Newman hypothecates, then one would expect the presence of one asymmetrical feature to show a high correlation with the presence of others. Von Verschuer has worked out a series of these correlation coefficients, all of which proved not to be statistically significant. The two following tables are quoted from him:

   It is to be regretted that von Verschuer does not state the number of twins on which these figures are based. The absence of any correlation between any two asymmetries, however treated, does, however, provide a strong argument against Newman's hypothesis. Newman himself states that he has found correlation between various asymmetrical features, such as palm‑prints and dental irregularities with handedness or crown‑whorl, but his data are hardly given in sufficient detail to permit of any judgement.

   The matter must therefore be regarded as still open. A judgement is made more difficult by the paucity of information on the frequency of these asym­metrical characters among the children of single births. So far it seems that left‑handedness alone shows a higher incidence in twins than in singletons, and that in twins it is not yet certain that there is a higher incidence in the nionozygotic than in the dizygotic. The increased frequency of left‑handedness in fraternal twins as compared to the general population, pace Newman, does suggest that left‑handedness in both varieties of twins may have a common explanation. This common explanation would besought in the twin‑pregnancy itself, a theory which would be quite in accordance with von Versehuer's failure to find any correlation between the presence of one asymmetry and the presence of another. The matter, as Newman rightly points out, is not thereby disposed of. An individual in whom a partial reversal of asymmetry has occurred, e.g. an ambidextrous person, is not the same as one in whom it has not occurred, even where other qualities, such as psychological ones, are considered. If partial reversals of the functional asymmetry of the brain, from whatever cause, are of common occurrence in twins, they would be of fundamental importance for psychiatry. What their importance would be, however, would be a most difficult matter to assess. Whether this phenomenon of asymmetry reversal, if it occurs, be regarded as of environmental origin or not, there seems to be little sound evidence for Dahlberg's view that it can be accompanied by differ­ences in the distribution of the germ‑plasm. There is no more reason to suppose that two monozygotic twins, even showing discordance in asymmetrical features, are of different genotypic constitution than to suppose that the two sides of the body are different in this respect. As long as we are dealing with symmetrical characters or qualities expressed by the individual as a whole, as is mostly the case in psychiatry, we are justified in neglecting the theoretical possibility of inequalities in twin pairs from this cause.

   So much for the theoretical basis of twin research. In a comprehensive review of its application in psychiatry in 1930 Luxenburger points out that work on twins may be classified into three kinds. There is the single‑case report, the collection of a small or large number of cases in an unsystematic way, and the serial investigation in which all types of twins fulfilling certain conditions are examined, regardless of their ovularity, concordance, or dis­cordance, etc. The great wealth of the literature is made up of work of the first kind, and, as Luxenburger has shown, it is for statistical purposes very largely useless. For it has been convincingly shown that when such reports of single cases are collected together and analysed they give no true picture of the twin material as provided by nature. They represent a selection of a particular type of case, the selection being in the first place for concordance. It is where both twins have suffered from schizophrenia, for instance, that the interest of the clinician has been aroused. These collections of single cases from the literature contain a quite undue proportion of concordant and of monozygotic twins, and even in the concordant cases there seems to be an undue selection for a far‑reaching concordance down to the finest clinical details.

   Luxenburger (1930b) provides a table compiled from the literature up to 1930, from which the following is extracted

   It will be noted that, when mongols are omitted, over two‑thirds of the reported twins are uniovular. The normal expectation would be to find about one‑third uniovular. As exemplifying the differing value of the three types of work on twins, he provides the following striking table, showing the compara­tive results from (a) collection of cases from the literature ; (b) cases collected by circulars sent to doctors, institutions, etc. ; and (c) the systematic collection of an uninterrupted series:

As Luxenburger says, the necessity for systematic unbiassed collections of twins may, on these data, be regarded as proven. Statistical analysis of the frequency of twinning in a clinical group, the extent of concordance in mono­zygotic twins as to clinical picture, onset, course, and outcome, is therefore impossible on the basis of such investigations. For the same reason uncon­trolled collections, without guarantees against special and undesirable selection, are also unreliable. The last class of statistically unobjectionable serial in­vestigations, securing a representative collection of twins, concordant and discordant, monozygotic and dizygotic, were in 1930 of great rarity ; and even to‑day there are not many of them. It is largely due to Luxenburger that the merits of this type of investigation are now recognized and that psychiatrists have made as much use of the method as they have. The method, as he demands it should be carried out, is to start from a particular group of patients, selected on grounds of their being in particular hospitals at a particular time or otherwise having fixed and definite limitations of time and space, and by inquiries made at registry offices to discover all the twin births, irrespective of whether the twin survived or died, became subsequently ill or remained well, and to subject the whole of the material to an exhaustive examination. In this way a series is obtained which is without gaps and is representative (1) in respect of the frequency of twins among the bearers of the character under investigation ; (2) in respect of the comparative frequency of like and opposite­sexed pairs, of monozygous and dizygous pairs, of concordance and discordance. Information on all these points may be of the greatest importance for any judgement to be reached on the presence, importance, and nature of hereditary factors. Even this type of investigation, as Luxenburger has said, is not without statistical source of error ; for by its means those twins are never discovered where, in spite of the presence of the hereditary factor, neither twin has developed the illness, abnormality, or character investigated. This source of error is unavoidable, and must limit the value of all twin investigations. It is, however, often not feasible to approach Luxenburger's moderate ideal. It is in England, for instance, impossible to ascertain the fact of twinning from birth registers. Furthermore, for the final figures to be sound it would be necessary to make sure that the whole of the original material out of which the twins were selected is uniform‑that, for instance, all possibilities of incorrect diagnosis have been eliminated. When a large twin material is being collected this is obviously not humanly possible. Most twin investigators will, therefore, have to content themselves with something less than what would be theoretically desirable. If the method of ascertainment of the twins is such that there is no likelihood of either uniovular or binovular, concordant or discordant twins being more largely represented in the final material than their proportional representation in nature, this is as much as can be demanded. The necessity of undiscriminating attention to all the material extends itself throughout the whole of the work, however, and it is necessary that the twins should not only be ascertained but also worked up with equal care, regardless of whether they are monozygotic or dizygotic, concordant or discordant. Furthermore, the reliability of any such investigation demands the satisfactory diagnosis of ovularity.

   This is not to assert that all case reports of single cases are without value. A particular character may be of such rarity that its occurrence in a pair of twins is an event of importance and worthy of record. In such cases clinicians should be aware of the fact that all twins, monozygotic or dizygotic, concordant or discordant, are of importance. It is, however, obviously better that such twins should be ascertained by an automatic method which does not depend for its certainty on the interest of the clinician. All hospitals where any research work is done at all should make inquiries as to twinship about all their patients as a matter of ordinary routine, so that when an interested investigator appears he may have a chance of following up all cases showing a particular syndrome.