Twin Research in Mental Defect

    In attempting a summary of the results obtained by twin investigations in psychiatry, and taking at first the subject of mental defect, the large number of case reports that have been made on cases of mongolism occurring in twins provide an elegant example of the weakness of the methods used hitherto. In spite of the rarity of the condition, it has aroused so much clinical interest that Rosanoff in 1934 (1934a) was able to collect 59 cases from the literature, and his collection was by no means complete. To these 59 cases he was able to add 5 new pairs of his own. Of these 64 pairs, in five the sex of the twins is not stated. It is difficult to imagine why a clinician can have left out such an important point in his report. A further 23 pairs were of opposite sex. This is about the number one would expect in a random selection; so mongolism forms the only exception to the rule stated by Luxen burger to the effect that cases from the literature represent a selection for uniovularity. Rosanoff states that eight pairs were monozygotic and all concordant. All the same‑sexed pairs held to be dizygotic and all the opposite‑sexed pairs were discordant in respect of mongolism. These findings might by themselves be interpreted as in­dicating that a hereditary factor played a major role in the etiology. Rosanoff, however, agrees with the majority of other workers in thinking that the con­dition is not hereditary, and that the twin findings do not stand in contradiction to this belief. Family investigations do not disclose an increased frequency of mongolism among the sibs of mongols. The most widely held theory of the causation of mongolism is that the condition is due to faulty implantation of the ovum in an abnormal uterine mucosa. Monozygotic twins with closely neighbouring implantation sites and with a common placenta would naturally be likely to be concordant in regard to a character caused by deficient or abnormal nutrition consequent on the location of the placenta. Rosanoff's theory is that the condition is caused by injury to the ovum on ovulation, the injury being brought about by scars produced in the ovary by previous ovula­tions. The frequency of mongolism would on that theory naturally increase with increasing age of the mother, as has repeatedly been shown to be the case. One of the cases which Rosanoff has not collected from the literature is Siegert's case (1922, quoted by Siemens 1924) of discordant monozygotic twins. Were the uniovularity of Siegert's twins established (no evidence beyond a statement is provided), it would be a direct proof of the incorrectness of Rosanoff's theory, which obviously demands that all monozygotic twins without exception should be concordant.

   The fact is that no definite statement can be made about the incidence of mongolism in monozygotic and dizygotic twins from cases collected from the literature, as Rosanoff has collected them. The diagnosis of ovularity is not satisfactorily carried out in one of them. Nevertheless, the example of mon­golism provides a lesson in the necessity of caution in the interpretation of concordance‑discordance differences in monozygotic and dizygotic twins as a direct measure of the relative prepotence of heredity. In any case where a particular character may be the result of intra‑uterine processes or of the process of twinning, it is clear that the results of researches into twins must be interpreted with especial caution.

   Apart from mongolism, case reports on the occurrence of special forms of mental defect in twins are few and unimportant. Dorf (1934) has reported a case of cretinism occurring in one of binovular twins; in uniovular twins Wilson and Wolfsohn (1929) have reported cerebral deplegia with mental defect and congenital nuclear ophthalmoplegia with mental defect, von Engerth (1933) tower skull, Fabing (1934) epiloia, Jenkins and Glickman (1934) birth injury (one only affected). In most cases a selection of the appropriate literature is given, in Fabing's case, which appears to be unique, no data are given relevant to the diagnosis of ovularity. In both his and von Engerth's case there are interesting data regarding mirror‑imaging in pathological features. Von Engerth's case is beautifully worked up from the biometric aspect. A not uninteresting case is reported by Schmitz‑Lückger (1932), who observed mental defect of about the same degree and with many clinical similarities in a pair of opposite‑sexed twins, and suggests it was due to the contraceptive employed by the parents, which had not prevented the occurrence, but had damaged the result of conception.

   Of the more usual types of mental defect a number of single cases occurring in twins have been reported, which it is not worth while to detail, as more valuable material is available. Both Smith (1929, 1930) and Rosanoff (1931, 1934) have made fairly systematic investigations in this field.

   Rosanoff (1931) has published his results with mentally defective twins with a brevity that robs his paper of much value. He gives no account of the methods of selection, no account of the cases from a clinical point of view, no account of the methods used for the diagnosis of ovularity, nor is there any statement of any quantitative measurement of the degree of defect. The body of the paper is given over to a discussion of a theory of the method of in­heritance which will be dealt with later. From the brief and incomplete analysis he gives of his material the following table can be constructed:

Of the 16 discordant opposite‑sexed pairs, the male was affected in 11 cases, the female in five. In all there were 91 mentally defective in 101 males, 69 mentally defective in 89 females. Rosanoff attempts to interpret his findings as recon­cilable with a dirneric inheritance of mental defect, the factors being in both cases recessive, one autosomal, the other located in the X‑chromosome. He finds support for this theory in the results of the family investigations of Goddard, which he has subjected to an analysis with this theory in view. In a later paper (1934) he reports the findings in a further 65 (making in all 81) pairs of opposite‑sexed twins. Among these he found 34 pairs where both twins were mentally defective, 31 where the male alone was affected, 16 where the female alone was affected. He further reports the findings in 233 pairs of opposite‑sexed twins, of normal or superior intelligence (1.Q. 76‑161), and found that the girls tended to have a higher intelligence than the boys. He further examined the intelligence scores of the parents of 100 of these twins and found that the fathers had higher J.Qs. than the mothers in 57 of the 100 cases. The sex‑distribution of higher I.Q. in the parents had no correlation with the sex­distribution of higher I.Q. among the twins‑a fact that speaks strongly against Rosanoff's theory, though he does not treat it as such. On this basis Rosanoff and his co‑workers build up a theory of the greater vulnerability of the male fetus and sex‑linked recessive inheritance as independent causes of the higher intelligence among the girls. A full analysis of Rosanoff's figures and his argument would not repay itself, but there is one important possibility to be pointed out. Rosanoff nowhere states what the ages of his twins were. It seems extremely probable that he was dealing with children, as he speaks everywhere of boys and girls. Now it is a well‑known phenomenon that in the curves of growth of intelligence with age of boys and girls, the curve is steeper in the case of girls than boys, and that in consequence the girls reach maturity in intelligence before the boys do. If this factor has not been taken into account by Rosanoff it would make just such a systematic difference between male and female twins as he observed.

   Luxenburger, stimulated by Rosanoff's theory, has published (1932) an analysis of the material of Brugger and [okay, which as it is not a twin‑material, is not the immediate concern of this review. Nevertheless, the figures he supplies are of interest and importance. In the families of 209 propositi he found the frequency of mental defect among children classified according to the constitution of their parents to be as follows

Luxenburger analyses the expectations of what these proportions would be on the theories (1) of simple autosomal recessive inheritance, (2) of simple recessive sex‑linked inheritance, (3) of both methods of inheritance occurring in a mixed population, (4) of dimeric recessive autosomal and sex‑linked inheritance (Rosanoff). He finds that the last theory provides the best explanation of the facts. Luxenburger's mathematics are not above reproach. He has, for instance, made no allowance for the comparative frequencies of the different types of crossings which would be productive of defectives. The mathematics of this problem are certainly more complicated than Luxenburger represents them, and would be worth the attention of a mathematical specialist. I think it would be found that Rosanoff's theory provides an inadequate explanation of the empirically found figures, and it seems to me likely that it would involve a greater relative frequency of mental defect in males as compared with females than is actually found. Rosanoff is himself aware of the inadequacy of his theory, and brings in the supplementary hypothesis of the greater vulnerability of the male fcetus. Luxenburger also does not suppose that this method of dirneric inheritance would be applicable to all cases of hereditary mental defect, but thinks that other methods of inheritance also play a part, if only a minor one. The chief criticism of the work of these two authors must be that it represents an over‑simplification of the problem. Human beings are not simply to be divided into normal and defective. Any theory of the inheritance of mental defect must provide an explanation of the curve of normal distribution of intelligence found in both males and females. It seems to me that the only two theories that will simply explain this phenomenon are: (1) that there are a large number of factors for mental defect, which are of different frequency in the population, and each produce mental defect of different degrees ; or (2) that there are a large number of factors for intelligence, any number of which may be present in a single individual, the effect of any one of which will be to enhance the action of any of the others‑that is, the more of these factors anyone has, the more intelligent he will be, the less the less. The two theories are not mutually exclusive and there is much to suggest that both have some truth. But even that represents an over‑simplification, as among other things the environment must play a role.

   The effect of the environment on intelligence and character has been the subject of an elegant and painstaking investigation of Newman, Freeman, and Holzinger (1937), an investigation which will probably remain unique. It must be briefly referred to, as it is not principally concerned with mental defect. The unique part of the work consists in the collection of 19 pairs of mono­zygotic twins who had been separated in early life and brought up in different environments. The authors were able to show that the intelligence, as estimated by standard tests, is subject to greater variation with varying environ­ment than is commonly supposed. The same was true of character‑traits, as estimated by other tests, but this part of their work is less satisfactory, owing to the greater difficulty in interpreting the meaning of such tests. The result of their work is to throw into relief the limitations of twin work and to illu­minate the fundamental difficulty of stating what the problem is, which is involved in the often formulated antithesis, heredity vs. environment. It is clear that if one were to take two individuals of like innate ability, and subject one from infancy to adult life to a stimulating and well‑balanced education, and to keep the other from education of any sort whatever, the final result would be very different in the two cases. This is what the above‑named authors have shown, in a less extreme case, to be true, tinder the widest range of environ­mental circumstance they found the extreme range of variation in intelligence quotient to be 24 per cent. In this case the variation was from 8 points below to 16 points above the normal value of 100 per cent. It is not admissible to argue from this that at lower or higher levels of intelligence a comparable change in the environment would produce an equal change in the test‑in­telligence. Both environment and heredity in the individual set limits to the range of variation in intelligence, of which the lower limit is nil. If the environ­ment is depressing, an improvement in it may be expected to produce a marked improvement in the intelligence. If, however, it is already favourable, a comparable variability cannot be expected. That quite large changes in the environment do not as a rule produce large changes in the I.Q. is shown by Holzinger's calculations that the mean difference in t.Q. in these 19 twins was little more than 8 points. That it was found to be as much as this will be a matter for surprise for some.

   The investigation of Newman, Freeman, and Holzinger was principally concerned with the effect of the post‑natal environment. To the prenatal environment and to the various environmental influences brought to bear at the time of birth, a probably much larger rôle is to be ascribed. The effect of birth‑trauma in producing mental defect has not yet been satisfactorily in­vestigated. Brander (1935) has drawn attention to this, and has criticized especially the work of Smith, still to be discussed, as taking insufficient account of birth‑trauma as a cause of defect, especially in concordant twins. In his own series of 10 mentally defective twin‑pairs birth anomalies had played a considerable role, and the twin with the lower I.Q. had in every case been subjected to their influence. He emphasized especially the importance of minor degrees of birth‑trauma in producing the less gross degrees of mental retardation. There seems to be much justice in his criticism ; and there is no reason to think that birth trauma, like the environmental factors responsible for mongolism, may not produce mental defect in both of a pair of uniovular twins.

   Smith gives an account of his investigations in two papers. An account at length is given in the Danish language (1929), and is not available to the reviewer. A shortened account is also given in German (1930), and appears to contain the more essential facts. In the course of the registration of all the mental defectives in Denmark, Smith ascertained all the multiple births among the first 6,700 mental defectives registered. In 67 pairs of twins full details were available. Of these 35 pairs were of opposite sex. A careful attempt was made to establish the ovularity of all the same‑sexed pairs, and though this was not possible with certainty in every case (for instance where the twin partner was already dead), there is no reason to think that errors of ovularity diagnosis are a large source of error in his results. He found in all 50 pairs of binovular twins, of whom four were concordant in respect of mental deficiency. In one case no decision could be reached as to ovularity. In three cases uniovularity was regarded as probable : all of these were concordant. Thirteen pairs were regarded as certainly uniovular, of whom 11 pairs were concordant, two discordant. The proportion of concordance in uniovular pairs (11/13 ‑=846 per cent.) compared to that in binovular pairs (8 per cent.) speaks, as Smith remarks, strongly for the hereditary determination of the majority of cases of mental defect. Smith's cases were all quite ordinary cases of mental defect, with or without history of fits, with or without neurological abnor­malities. There were no cases of the rarer and special forms of mental defect. He was unable to find a clinical picture that was characteristic for a hereditary or for an acquired form of the condition. His most interesting finding is that a number of the concordant uniovular pairs showed various neurological abnormalities. In pair 109 both twins had a left‑sided spastic hemiplegia, both had convulsions as children ; one of the pair got progressively worse and died at the age of 22 ; the other improved, though neurological signs remained. Both twins of pair 115 showed disturbance of speech, with defective articulation, both were somewhat deaf, one showed a positive Babinski. In another three pairs (perhaps four) both twins had convulsions, and one of one pair died in status epilepticus. This conformity of neurological picture Smith is inclined to interpret as suggesting that even those cases of mental defect which show an abnormal neurological picture may be due to hereditary and not to any exogenous cause. The possibility must be conceded, but the evidence is not conclusive. Other explanations cannot be excluded, as in the case of mongolism.

   Smith's two discordant uniovular pairs both show an obviously exogenous disorder. In pair 120 the affected twin had fits from the third day after birth, recurring at intervals since. The left side of the body became with further growth markedly underdeveloped, and he died in a state of idiocy at eight; his brother was normal. This would appear to be a case of birth trauma. In case 121 (triplets), the affected twin and his sister had a severe exanthematous illness at the age of three months, by which the healthy brother remained unaffected.

   As regards the degree of defect, Smith does not provide very detailed information in his German paper ; but from what facts he does give it is clear that there were perceptible though not very large differences in this respect Only in one pair (116) is the difference at all a marked one. Characterological differences were also noted in the concordant twins. One gets the impression from the work that an inborn tendency to mental defect manifests itself in a high proportion of the cases where it is present, and generally to its full effect. In other words, the gene or genes responsible show, in the language of genetics, high values for manifestation rate, expressivity, and specificity. This is not to say that environmental factors are not important. There may be cases of mental defect largely determined by heredity where the environment is largely without power of modification, and others caused solely by the environment (e.g birth trauma) where there has been no hereditary predisposition.

   Working on Smith's material, Luxenburger has attempted to obtain an actual value for the manifestation rate of the hereditary factors responsible for mental defect. By manifestation rate is meant the average percentage of cases where the genotypic factors are present in which they show themselves pheno­typically.

   Luxenburger provides a formula

in which M represents the manifestation rate, mg the number of concordant twins among the like‑sexed pairs, v the empirically found expectation of sibs to be affected similarly to the propositus (in this case with mental defect), y’ the expectation of a twin reaching the average age of the material investigated and z the total number of twin proposity. This formula is for use on a total twin material, including all cases that died in infancy, and can only be applied to material gathered by a method of selection where such factors as age, sex concordance, etc., have no selective effect. Its great convenience is that it avoids the determination of ovularity altogether. Its weakness lies in the fact that the factors v and y' are themselves subject to large errors. From Smith's figures Luxenburger calculates the " expected figure" for the manifestation rate of the genes of mental defect as 0•835, and compares this with the " observed figure of 0•846 (the number of concordant uniovular twins divided by the total number of uniovular twins, here 11/13). Luxenburger points out that the agreement is good. It is, however, without significance.

   The "manifestation rate" of an inherited disposition is a term derived from animal genetics. It signifies that where the inherited tendency towards the development of a certain character is present it will manifest itself in a certain proportion of cases. In the remainder it will remain inhibited. The inhibition may be either by the action of other inherited factors or by the presence of an unsuitable environment. If the inhibition is solely by reason of other in­herited factors, it will take place, if at all, in both of a pair of twins, which will then entirely escape observation. The calculation of the manifestation rate from concordance‑discordance figures from twin studies gives only an expression for the effect of the environment. The environment may, however, be of such a quality that the character is produced (or imitated) without the inherited tendency being present at all. This is probably the case with Smith's two discordant pairs, where there are certainly no grounds for the assumption that there was a hereditary disposition to mental defect present in both twins which showed itself in only one. These two cases would more properly have been excluded from the calculation, and the " observed " manifestation rate is in Smith's cases apparently 1•0.

   Luxenburger, however, in his analysis of Smith's results is concerned with a particular point. He quotes the results of Brugger and Lokay, who investigated the frequency of mental defect among the sibs of defectives, and found it to be 0•13 to 0•18. Both figures are well behind the expected 0•25 on the assumption of simple recessive inheritance. Luxenburger is concerned to show that this discrepancy is not due to deficient manifestation of the inherited factor. It is clear that a manifestation rate of 0•835 or 0•846 or 1•0 will not explain the total discrepancy, but at most a part of it. Luxenburger seems not to have been aware at this time of the possible additional disturbing effect of other inherited factors than the specific ones (factors conveniently classed as the "genotypic milieu"), but only to have reckoned with possible environmental interference. As Luxenburger has himself subsequently pointed out, the effect of the genotypic milieu is most conveniently to be measured together with the effect of the environment (when certain assumptions are made as to the mode of inheritance) by discovering the frequency of a character among the children of parents, both of which show it. From Luxenburger's figures, previously quoted, it appears that the incidence of mental defect among the children of two defective parents is about 0•9.

   Luxenburger next considers the possible influence of lethal effects of the genes concerned. Obviously if a particular gene were to have a lethal effect the number of its carriers in a given population would be reduced and in families known to carry the gene the number of those giving evidence of its presence would be below the expected value. Even if the lethal effect were only exhibited in a percentage of cases, this would still be the case. Obviously the genes for mental defect are not absolutely lethal, because then there would be no subjects of mental defect that would survive. Luxenburger attempts to test the possi­bility of a lethal effect showing itself in a proportion of cases by calculating the relative and absolute frequency of uniovular and binovular pairs in Smith's material.

   The method is based on a statement of Weinberg's to the effect that among the bearers of a character which involves an increased prenatal mortality, a relatively reduced percentage of binovular twins is to be expected. Luxenburger calculates that the percentage of twins in Smith's material (122 in 6,689) corre­sponds to average expectation and that the opposite‑sexed twins (i.e. certainly dizygotic twins) are neither more nor less frequent than could be expected. He deduces from this that the fact of twinning as such does not predispose to mental defect, and that there is no evidence of any lethal effect exercised by the genes of mental defect.

   These conclusions may be true, but the reasoning does not seem to be altogether sound, and there are reasons to be sceptical. Twin conception is known to predispose to increased prenatal mortality, twin birth to birth injury, and birth injury to mental defect and to reduced viability. Luxenburger seems to have made no direct comparison of the duration of life of the twin partners in Smith's material with the life tables of the general Danish population, a population of defectives, or with a population of normal twins. Facts gained from such comparisons would be more valuable than the somewhat indirect evidence supplied.

   Luxenburger's argument, that the normal distribution of uniovular and binovular twins in Smith's material is against the assumption of any lethal effect, rests entirely on the validity of Weinberg's statement that a lethal effect would show itself in a reduced number of binovular twins. Luxenburger does not give a reference to the particular work of Weinberg's in which this state­ment occurs and I am unable to trace it. The statement itself, though pro­ceeding from a statistical authority, seems to me of questionable truth.

   One might approach the problem as follows

   Taking a group of crossings which produce the character under discussion, let

   a=the number of uniovular twin pairs born,

   b=the number of binovular twin pairs born,

   v=the frequency of the character in the children of these crossings.

Then there will be

   av                uniovular pairs with the character,

   a(1‑v)           uniovular pairs without the character,

   bv2               binovular concordant pairs,

   2bv(1‑v)       binovular discordant pairs,

   b(l‑v)2          binovular pairs without the character.

Starting from twin propositi one will discover (by Weinberg's "Probanden­methode," used in all the German calculations)

   2av              concordant uniovular pairs,

   2bv2            concordant binovular pairs,

   2bv(l‑v)        discordant binovular pairs.

The relation of the uniovular twin pairs found to the binovular pairs found will be:


Now let the character have a lethal effect e which is exerted at random. Among the uniovular pairs who bear the character there now will be

   av(1‑e)2        pairs where both twins survive,

   2ave(1‑e)     pairs where one twin survives,

   ave2             pairs where neither twin survives.

Among the binovular twins there will be

   bv2(1‑e)2         concordant pairs where both twins survive,

   2bv2e(l‑e)        concordant pairs where one twin survives,

   bv2e2               concordant pairs where neither twin survives,

   2bv(1‑v)(1‑e)   discordant pairs where both twins survive,

   2bv(1‑v)e         discordant pairs where one twin survives.

Starting again from twin propositi one now will find

   2av(l‑e)2          uniovular twin pairs, all concordant,

   2bv2(l‑e)2         binovular twin pairs, concordant,

   2bv(1‑v)(l‑c)     binovular pairs, discordant.

and the relation of uniovular to binovular twin pairs found will be


 But as v and e are by hypothesis both positive fractions,   must be less than 1. This resolves itself to a statement that under the conditions laid down above, the lethal effect of a hereditary factor would show itself in a larger and not a smaller proportion of dizygotic twins in a total twin material.

   The fundamental assumption in the above argument is that the lethal effect is exhibited in a proportion of cases distributed at random. If the lethal effect were dependent for its manifestation on favourable environmental circum­stances, this might well be the case. If, however, the lethal effect were dependent on a favourable genotypic milieu, the result would be different.

   Let us suppose that the character exhibits its lethal effect only where are a considerable number of certain other common genes also present. These other genes are only found all together in a proportion e of cases. Now the circumstances among the binovular twins will remain as in the example worked out above ; but among the uniovular twins the lethal affect will show itself in a proportion e of pairs, in all of which both members of the pair will be affected. Starting again from twin propositi, the number of monozygotic twin pairs found will now be no longer 2av(1‑e)2, but 2av(l‑e); and the relation of uniovular to binovular twins found will be no longer   but   In this case the proportion of dizygotic twins found will be unduly low, as Weinberg is said to have stated.

   But we do not know whether this purely hypothetical lethal effect is exhibited in accordance with one or other or both of the possibilities outlined above. It is therefore not admissible to argue from a normal distribution of uniovular and binovular twins that no lethal effect of the genes concerned is present. The matter is not of purely theoretical interest, as it might appear, for in the twin work on epilepsy we shall meet with it again and find far‑reaching conclusions based on arguments similar to Luxenburger's.

   If we attempt a summary of what twin work has added to our knowledge of the inheritance of mental defect we may say that its findings are in conformity with those supplied by family investigations, and indicate that hereditary factors are of great importance, though certainly not the only ones concerned in the production of the syndrome. Much further investigation is required, and not least in the rarer and special forms of mental defect. The systematic ascertain­ment of all twins in institutions for mental defectives would be likely to be productive of a very valuable material for research.



Brander, T. (1935). Mschr. Kinderheilk., 63, 276.

Conrad, K. (1937). Fortschr. Neural. Psychiat., 9, 197.

Curtius, F. (1930). Arch. Gynäk., 140, 361.

Curtius, F., and Verschuer, 0. v. (1932). Arch. Rass. u. Ges Biol., 26, 361.

Dahlberg, G. (1926). "Twin Births and Twins from a Hereditary Point of View." Stock­holm.

Dorf, G. B. (1934). Amer.J. Dis. Child., 48, 1,316.

Engerth, G. v. (1933). Z. ges. Neurol. Psychiat., 148, 670.

Fabing, H. (1934). Brain, 57, 227.

Jenkins, R. L., and Glickman, F. (1934). Amer. J. Div. Child., 48, 130.

Lange, J. (1929). Allg. Z. Psychiat., 90, 122.

Lassen, M. T. (1931). Arch. Gynäk., 147, 48.

Lenz, F. (1933). Arch. Russ. u. GesBiol., 27, 294 and 311.

Luxenburger, H. (1929). Fortschr. Neurol. Psychiat., 1, 82.

--- (1930a). Ibid., 2, 373.

--- (1930b). Zbl. ges. Neurol. Psychiat., 56, 145.

--- (1931). Z. ges. Neurol. Psychiat., 135, 767.

--- (1932a). Fortschr. Neurol. Psychiat., 4, 1 and 49.

--- (1932b). Z. ges. Neurol. Psychiat., 140, 320.

--- (1932c). Z. induki. Abstumm. u. VererbLehre, 61, 223.

--- (1933). Fortschr. Neurol. Psychiat., 5, 1.

--- (1935). Ibid., 7, 173.

Meyer, H. C. (1932). Arch. Rass. u. GesBiol., 26, 387.

Newman, H. H. (1923). "The Physiology of Twinning." Univ. Press, Chicago.

Newman, H. H., Freeman, F. N., and Hoffinger, K. J. (1937). "Twins : a Study of Heredity and Environment." Univ. Press, Chicago.

Rosanoff, A. J. (1931). Amer. J. Psychiat., 11, 289.

Rosanoff, A. I., and Handy, L. M. (1934). Amer. J. Dis. Child., 48, 764.

Rosanoff, A. J., Handy, L. M., Rosanoff, I. A., and Inman‑Kane, C. V. (1934). J. Nerv. Ment. Dis., 80, 125.

Schmitz‑Lückger, J. (1932). Psychiat. neurol. Wschr., 34, 437.

Siegert, . . . (1922). Contribution to discussion in Klin. Wschr., 1, 2,259.

Siemens, 1‑1. W. (1924). "Die Zwillingspathologie." Springer, Berlin.

--- (1932). Z. indukt. Abstamm. u. VererbLehre, 61, 208.

Smith, J. C. (1929). Medd. Danm. Antrop., 3, I (quoted Smith 1930, Luxenburger 1931, q.v. - not verified).

---(1930). Z. ges. Neurol. Psychiat., 125, 678,

---(1931). Acta Psychiat., Kbh., 6, 79.

Thompson, O. (1929). Quoted Smith (1931), q.v.

Verschuer, O. v. (1925). Münch. med. Wschr., 72, 184.

--- (1927). Ergebn. inn. Med. Kinderheilk., 31, 35.

--- (1932). Z. indukt. Abstamm. a. VerebLehre, 61, 147.

--- (1933). Arch. Gynäk., 156, 362.

Wilson, S. A. K., and Wolfsohn, J. M. (1929). Arch. Neurol. Psychiat., 21, 477.

Wilson, P. T., and Jones, H. E. (1932). Genetics, 17, 560.