![[Home/Contents]](../images/
home90.gif){kind=small}
![[email]](../images/
email90.gif){kind=small}
Edward M. Miller
Professor of Economics and Finance
University of New Orleans
New Orleans, LA 70148
emmef@uno.edu (E-Mail)
From Personality and Individual Differences, Vol. 17, October 1994, No. 4, 511-529.
In rodents, sex hormones transfer between fetuses. Females that are adjacent to males in utero later develop more masculine behavior and morphology. Human sex hormones can pass the placenta, as well as from the fetus to the amniotic fluid. It is plausible that hormones from one human fetus could masculinize or feminize an opposite-sex twin fetus.
Many results described in the opposite-sex twin literature could be explained by sex hormone transference. Such results involve dental asymmetry, twin death rates, resemblance to parents, sensation seeking, spatial performance, femininity or masculinity, dominance, myopia, mathematics performance, perceptual speed, responses to public opinion questionnaires, and high opposite-sex twin resemblance in intelligence and personality. Studies of hormonal transfer effects provide a new reason for studying opposite-sex twins, and a possible new funding source for twin studies.
New analysis of answers to public opinion questionnaires shows evidence that the opinions of females twins resemble the male answers more if the cotwin is male than if it is female, a result which is interpreted as providing evidence of testosterone transfer.
Key words: opposite-sex twins, sex hormones, testosterone, behavior, prenatal
The biological literature reports that females that are adjacent to males in the uterus develop more masculine behavior and morphological traits. This is usually attributed to hormones passing between fetuse. This in turn alters the developing brain.
It appears plausible that the same effect occurs in humans. This implies that twins whose womb mates were of the opposite sex (abbreviated OS) would be morphologically or behaviorally different from those whose womb mates were of the same sex (SS). This paper will review the animal literature, and the literature reporting effects in humans that may reflect hormone transfer. Since the hypothesis that hormones transfer in humans is new, it was not considered by the authors of most of the studies discussed. Thus, this paper will reconsider older results in the light of the hormonal transfer hypothesis.
A good survey of the uterine position literature has been provided by vom Saal (1989). Gandelman (1992, Chapter 3) has summarized the uterine position literature as well as the evidence for effects of fetal androgens on animal and human behavior.
Clemens (1974) was the first to report an uterine position effect. He found that female mice adjacent to male mice in the uterus had larger ano-genital distances and later exhibited more masculine behavior than females adjacent to females. This was attributed to testosterone from the male fetuses somehow affecting the female fetuses, since the effect on ano-genital distance did not appear when the mothers were treated with antiandrogens. A possible mechanism was provided when Fels and Bosch (1971) showed that testosterone could diffuse across amniotic membranes in rats. Gandelman et al. (1977) showed that the position of female mice in the uterus affected the potential for female aggressive behavior. After the mice were injected with testosterone, females that had been between two male mice showed more aggressive behavior than those that had been between females. These female mice actually more closely resembled male mice in aggressiveness than they did female mice that had been between two females. The weight of both male and female mice depends on intrauterine position (Kinsley, Miele, Wagner, Ghiraldi, Broida, & Svare, 1986). In mice, males are the heavier sex. Within each sex, those who were between two males (hence influenced by testosterone from them) were heavier than those who were between two females (significantly so for females, and almost so for males). The effect was large enough so that females between two males (high testosterone exposure) had body weights that were in the male range (except for the last days of the experiment). These masculinized females actually weighed more than the feminized males who had been between two females.
In the same set of experiments, female mice located between two males (higher testosterone levels) were less active than females between females (Kinsley, Miele, Konen, Ghiraldi, Broida, & Svare, 1986). Female mice are generally more active than male mice. The effect was large enough so that the females who had been affected by testosterone from males fetuses actually resembled male mice more in activity than they did the unexposed (due to being between females in utero) female mice. This shows that not only can hormones transfer, but that sufficiently large amounts can transfer to have effects comparable in magnitude to the normal male/female differences.
Aggression was twice as common for females between two males during pregnancy and over twice as great during lactation in the same sample of mice (Kinsley, Konen, Miele, Ghiraldi, & Svare, 1986). There was also a statistically significant uterine position effect for the size of litter females produced.
Clark, Galef and associates (Clark and Galef, 1988; Clark et al., 1988; Clark, et al. 1990; Clark and Galef, 1989; and Clark, Tucker & Galef, 1992; Anonymous, 1992) found that numerous aspects of gerbil morphology and behavior depended on the sex of adjacent fetuses. Males between females weighed more, had larger scent glands, were able to sire larger litters, and were more attractive to females. Conversely, females between two females matured faster and gave birth to more liters than those between two males. Male gerbils who were adjacent to two males had higher blood levels of testosterone than males who were between two females (Clark, vom Saal and Galef, 1992).
Vom Saal and Bronson (1980) showed that blood serum and ammonitic fluid concentrations of testosterone for female mice depended on uterine position, as did the attractiveness of female mice to males. Other traits found in mice to depend on uterine position include ano-genital distance (Lephart, Fleming, & Rhees, 1989) in males and ano-genital distance and time of vaginal opening in females (McDermott et al., 1978), and infanticide and parental behavior in males (vom Saal, 1983). In mice, these effects are attributed to hormones diffusing across amniotic membranes.
The conditioned taste aversion reaction (in which rats learn to avoid a substance which has made them ill) following testosterone treatment depends on uterine position in female rats (Babine and Smotherman 1984), a finding suggesting that testosterone transferring from male fetuses caused female rats to develop a more masculine brain. Also, in female rats, Glick and Shapiro (1988, p. 153) showed that the preferred body rotation direction depends on the number of males in the litter, a result they interpreted as being due to testosterone transfer from the male fetuses. Mankes et al. (1991) found that the alcohol consumption and hepatic alcohol dehydrogenase activity of male rats between two males was significantly less that that of males between two females. The effect was strong enough so that the males between two females were shifted to the female range with regard to alcohol consumption and hepatic alcohol dehydrogenase activity.
In rats and gerbils adult testosterone levels are influenced by fetal position. If humans experience similar effects, effects that occur only after puberty may still be influenced by prenatal hormones.
Although Fels and Bosch (1971) had shown that testosterone could diffuse across rat amniotic membranes, Meisel and Ward (1981) presented evidence that masculinizing effects on female rats of womb position were probably due to transfer of testosterone through the maternal blood circulation. These effects were observed in female fetuses downstream from males, but not in those upstream. Richmond and Sachs (1984) found supporting evidence by showing that the presence of a male on the caudal side of the uterine horn had a more critical influence than contiguity per se. Gandelman (1986) found that in the guinea pig, both contiguity to males and positioning of the male fetus were necessary conditions for females treated with testosterone to exhibit male type copulatory behavior.
Pigs are the only non-rodent for which a clear effect of uterine position has been observed. In pigs, Rohde Parfet et al. (1990) found no
difference with uterine position in ano-genital distance or body weight at birth or after 120 days of ad libitum access to food. However, there was a statistically significant tendency for males between two males in utero to gain more weight from 175 to 270 days of age. During this period, the pigs were under restricted feeding conditions, in which they had to compete with each other for food. The higher weight gain under these conditions was interpreted as being due to greater aggressiveness, and confirmed by observations that indicated such males tended to dominate in competitive encounters.
In cows, there is the well known freemartin effect. Freemartins are bovine OS females with ovotestes. They are frequently sterile. This is apparently due to the sharing of placental circulation and hormone transfer. However, the substance producing the effect does not appear to be testosterone, since injecting testosterone into pregnant cows does not produce the same effects. Thus, the mechanism causing the freemartin effect differs from that found in rodents, where the effects have been shown to be due to transfer of testosterone or estradiol (vom Saal 1989, p. 1827)
An obvious question is whether similar effects occur in humans. The human equivalent would be OS twins. Several authors have speculated that one twin fetus could affect the other. Vom Saal et al. (1983) have documented that not only does behavior of male mice depend on their uterine position, but that prenatal hormone concentrations of estradiol vary with the sex of the adjacent fetuses. They noted that in humans much of the estrogen in the maternal circulation is of fetal origin. The prominent neurologists, Geschwind and Galburda (1987, p. 141) speculated from the animal studies that, "Since male twins both produce testosterone, each will conceivably be exposed to higher levels than he would be if he were a singleton. By this hypothesis, the females of opposite-sex pairs should have a high rate of left-handedness, because of exposure to testosterone produced by the male co-twins." Apparently, no effort was made to search the literature for such an effect. Evidence supporting this hypothesis will be presented later in discussing Koch's work.
Phoenix (1974) and associates (Phoenix et al. 1968) showed that Rhesus monkeys could be masculinized both morphologically and behaviorally by injections of testosterone propionate into the mothers. This experiment was inspired by observing that testosterone propionate injections into female guinea pigs masculinized their female offspring (Phoenix et al., 1968, p. 34), as it did in rats (p. 38). Later, eight such artificially masculinized monkeys were shown to display more masculine play behavior than control monkeys at 12 months of age (Goy and Resko, 1972, p. 720). The importance of the latter experiment is not just that behavior is affected by prenatal testosterone, but that testosterone (injected into the mother in the experiment) can pass through the placenta from the mother to the fetus in primates, and plausibly in humans.
Meulenberg and Hofman (1990) have shown that maternal blood testosterone levels depend on the sex of the fetus being carried. They conclude (p. 53) that "as a consequence of a maternal-fetal gradient unbound testosterone crosses the placenta from the male fetus toward the maternal circulation, whereas the opposite direction applies to a female fetus." This observation makes very plausible an effect of a male twin on the female co-twin's testosterone level, since testosterone appears able to pass from fetus to mother (human), and from mother to fetus (at least in monkeys, guinea pigs, and rats).
There is strong evidence that pregnancy estrogen concentrations in blood and in urine are correlated with birth weight (Ekbom et al., 1992, p. 1017). While it could be that estrogen somehow promotes fetal growth (or is correlated with a factor that promotes it), the simplest explanation is that estrogen produced by the fetus is entering the maternal circulation. Since large fetuses presumably produce more estrogen, the observed estrogen/birth weight correlation is easily explained.
When hormones in the amniotic fluid of fetuses were measured (Carson et al, 1982), two male hormones, testosterone and androstenedione, were found to have significantly higher levels in the male fetuses than in female fetuses at mid-term (15-21 weeks). In a sample of 20, there was only one fetus from each sex whose testosterone fell in the range of the other sex. The average male value was .30 ng/mL in males versus .08 ng/mL in females. Thus transfer of a third of the testosterone would more than double the typical female concentration. For androstenedione, the concentrations were .96 ng/mL in the male and .56 ng/mL in the female, a much smaller difference. Near birth, the
males still had higher levels, but the differences were no longer statistically significant. Other measured hormones, including progesterone, did not display statistically significant differences between the sexes, with males actually showing the higher concentrations of progesterone in late gestation. Thus, in humans, a potentially significant hormone transfer would be for a male fetus to raise appreciably the level of testosterone, or of an another male hormone, in an adjacent female fetus. While female hormones probably also transfer, the absence of large sex differences suggests that a male's level would not vary with the co-twin's sex.
Individuals with inadequate levels of sex hormones might develop to be neither fully male nor fully female, and leave no descendants. Hence, natural selection would be expected to provide for a large safety margin in fetal sex hormones levels. Thus, even under adverse conditions, fetuses would develop morphologically and behaviorally into reproducing adults. Human twins are rare (and OS twins even rarer). Thus, if genes for high male hormone levels reduce the risk of inadequate hormones reaching male singleton fetuses, a small effect (even if adverse) on co-twins would not have appreciably retarded the evolution of such high fetal sex hormone levels. If even a small part of these hormones reach the other fetus, the quantity could be large relative to what is required for observable behavior changes.
Prenatal female hormones do not appear necessary for genital feminization in mammals, while male hormones are necessary for masculinization. Thus, an effect of male hormones diffusing into females may be more likely than an effect of female hormones diffusing into males. However, prenatal exposure to female hormones does affect the brain (Halpern, 1992, pp. 111-112). That neonatal ovariectomy defeminizes behavior, and alters the rat's corpus callosum in a male direction at a development stage corresponding to when the human is prenatal (Fitch, Cowell, Schrott, & Denenberg, 1991), suggests that in the human prenatal female hormones may affect development.
There is a large and complex literature on maternal and fetal hormones. This literature's underlying assumption is that maternal, fetal, and placental hormonal states affect each other, either by direct transfer of the hormones from one compartment to the other, or through transfer of precursors (see Tulchinsky & Ryan, 1980; or Schindler, 1982). Solomon (1988, p. 2085) states that "the steroids readily cross the placenta." Although the author has not found direct discussion of the possibility of one fetus directly affecting another, nothing has been found that would preclude such effects. Evidence that the fetus affects its mother's hormones, and that the mother affects its fetusesU hormones, makes an effect of one fetus on another quite plausible. Even more complex interaction effects may occur.
It should be noted that the sex related hormones are chemically quite similar to each other, all consisting of three six-membered carbon rings and one five-membered ring joined to each other by common sides (Schulster et al., 1976, pp. 4-5). This makes it plausible that if one steroid hormone can be shown to diffuse, the others can also. In particular, cortisol (which is structurally similar to testosterone, as shown in Schulster, Burstein, & Cooke's p. 5 chart) is believed to transfer from the mother to the fetus. The maternal contribution to fetal cortisol is calculated at 25 to 50%, and to fetal cortisone at nearly 100% (Gibson & Tulchinsky, 1980, p. 134). By injecting radioactive forms into the mother and observing radioactivity in the fetus (Schindler, 1982, p. 91), cortisol and dehydroepiandrosterone have been shown to transfer from the maternal circulation to the fetal circulation. If steroids can transfer between maternal and fetal circulation, it is very likely that they can also transfer between two fetuses.
There is some evidence that handedness (Ellis and Peckham, 1991) and sexual orientation in humans are affected by maternal stress (Ellis et al., 1988). The effects are argued to occur through the stress affecting the levels of stress hormones (adrenaline, cortisol, and corticosterone) in the mother. These then reach the fetus through the placenta in humans, as in experimental animals. If these effects are real, it would provide human evidence that hormones structurally similar to sex hormones can pass from the mother to the fetus. As noted above, cortisol has a structure very similar to testosterone.
TambyRaja and Ratnam (1981) have reported higher levels of plasma estrogen in mothers of twins than were found in mothers of singletons. They speculate that (p. 193), "The high oestrogen levels in twin pregnancy may have been as a result of the four fetal adrenal glands producing an excess of precursor DHEAS and two fetal livers producing enough 16-hydoxylase for formation of these steroids." Trapp et al. (1986) report 40% higher estriol levels in mothers of twins than in mothers of singletons. If these female hormones transfer between the maternal and the fetal circulations, it makes it more plausible that the structurally similar male hormones could transfer from one fetus to the other.
Having shown plausibility of hormone transferal between fetuses, let us turn to reports in human twins that OS and SS twins differ in a way that could be explained by hormonal transfer.
Boklage (1985) examined the asymmetry in dental diameters in
human twins. He found that males generally had larger teeth in the right
jaw (which is consistent with the larger sizes for their right
hemisphere of their brains), and that it was rather easy to separate
males from females. He went on to say (p. 601, SS refers to same-sex, OS
to opposite-sex, and DZ to dizygotic):
"A linear discriminant function sex-classification rule
correctly classified 42 of the 44 SS-DZ in our sample, but misclassified
12 of our 20 OS-DZ twins (X2l = 21.6, exact P= 3 /X 10-7), eight of the
10 females, and four of the 10 males.
A quadratic discriminant function sex-classification rule that
correctly classified 123 out of 128 SS twins of both zygosities
misclassified 16 of the 20 OS-DZ twins (X2l = 72.1), seven of the 10
males, and nine of the 10 females."
The most plausible interpretation of these results is that something is transferring from the co-twin (presumably testosterone, estradiol, or another hormone) that causes OS females to become more masculine, and/or male fetuses to become more feminine.
Boklage (1987, Table 3, on p. 282) has also reported, from an analysis of another database that fetal death rates (i.e. deaths before birth) in OS males are lower than for other male twins. A plausible explanation would be a hormonal effect, with transfer of a female hormone being most plausible, although elevation of testosterone in SS twin males could play a role. No similar effect is found in females, whose fetal death rates do not depend on their twin's sex.
For neonatal deaths, there is a tendency for death rates in both sexes to be greater in SS DZ twins than in OS twins. Combined fetal and neonatal deaths show lower rates for OS twins (Table 2 on p. 281). Again, a possible explanation is hormonal transfer.
Zazzo (1960, p. 698) has found an interesting effect regarding whether twins were reported to more closely resembled their mothers or fathers. Not surprisingly, twins normally resembled more closely the parent of their sex. However, this effect was weaker (to a statistically significant degree) in OS twins. In particular, the OS females more closely resembled their fathers (56.4%) than their mothers (43.6%), while the SS DZ females more often resembled their mothers (57.1%). While this effect was statistically significant, OS males had only a slightly greater tendency to resemble their mothers (47.2%) than did DZ SS males (44.4%). However, they greatly (significantly at the .001 level) exceeded the maternal resemblances shown by the monozygotic (MZ) males (31.9%). While OS males differed from SS males, OS females differed little from SS females.
The most obvious interpretation is that the OS twins were less sex typical due to hormone transfer. They were thus perceived as having less resemblance to their same sex parents. Presumably, since sex differences are very obvious, the resemblance to a particular parent is reported after mentally adjusting for the expected sex difference. When the expected sex differences are lacking, an adjustment for the expected sex difference is then made, and the twin is then seen as less likely to resemble the parent. Thus, a relatively non-masculine looking male twin is less likely to be reported as resembling his father. Likewise, a relatively masculine female would be reported as more closely resembling her father rather than her mother. The greater paternal resemblance of OS female twins could be explained if transfer of male hormones contributed to a more masculine appearance. In a corresponding table for psychological resemblance, no significant differences were reported among twin types.
A puzzling result (p. 642) is that for a number of characteristics, the OS twins were less alike than MZ twins, but more alike than SS DZ twins. While the first is expected, the latter is unexpected. One would normally expect OS twins to be even more different than SS DZ twins, rather than more similar, since they have sex differences in addition to genetic differences. A possible hormonal explanation is that the sex differences are less than normally expected,
making it easier for OS twins to resemble each other as much as SS twins.
Before leaving morphology, one negative result should be reported. It was hypothesized that if hormones transfer, some of the sexually dimorphic body dimensions might differ depending on the co-twin's sex. This was thought to be especially likely for the ratio of the shoulder to hip. Tanner (1989, p. 68) states that "cartilage cells in the hip joint are specialized to respond to female sex hormone (oestrogen) and cartilage cells in the shoulder region are specialized to respond to male sex hormones (androgens, primarily testosterone)." He also states that, "The shoulder-hip dimorphism has long been used as a measure of bodily androgyny, i.e. the degree to which a male resembles a female, or vice versa." Dahlberg (1926) in a classic book provided an appendix giving anthropomorphic dimensions for 486 Swedish twins, many of which were OS. This data was entered and analyzed. No statistically significant effects for co-twin sex were found for the ratio of shoulder to hip widths.
McFadden (1993b) has reported that the number of spontaneous otoacoustic emissions in female OS twins is in the male range. SS females show appreciably higher rates than OS females and higher rates than males, as do female non-twins. Spontaneous otoacoustic emissions are continuous sounds that are produced in the cochlea and propagate into the external ear canal, where they can be recorded. It is not known exactly how these are produced or why they are more common in females, but McFadden (1993a) believes they may be due to differences in the strength of efferent inhibition delivered to the cochleas. If he is right, the reported opposite sex twin effect in female may be evidence that hormonal effects on these neurons differ between SS and OS females twins.
Whitfield & Martin (1992) have examined in human twins the reaction to a variety of variables considered relevant to alcohol consumption. They found no difference in female behavior depending on whether the co-twin was male or female. For males, they found a significant effect for only one of fourteen variables. However, for six out of seven tests, the OS males had reactions that were more feminine. Of course, these results could have occurred by chance. The one statistically significant result in males related to the number of incorrect responses to the motor co-ordination task. Here the score dropped from 17.7 (Standard error =1.2) for SS DZ males to 12.5 (Standard error =1.6) for OS DZ males. The SS female values were 12.8 (Standard error =1.2), versus 9.9 (standard error 1.2) for OS females. Thus, the OS males have values in the female range. While this result could be due to chance, further research is called for since the possible effect is of large size and in the direction predicted.
Resnick et al. (1993) have examined sensation seeking in a sample of 422 British adult twin pairs including 51 OS pairs. Age adjusted measures of sensation seeking behavior showed a statistically significant increase in OS females as compared with SS females. The increase was statistically significant for an overall measure, as well as for subscales apparently measuring disinhibition (interpreted as tapping interest in socially and sexually disinhibited activities), and experience seeking (interpreted as seeking of new experiences in a non-conformist way through travel, new aesthetic interests or consciousness altering drugs). OS females had elevated scores for thrill and adventure seeking and for boredom susceptibility. For all scales, there was a statistically significant tendency for males to score higher. Thus, the OS femalesU attitudes were consistently more masculine than those of the SS females. No consistent pattern was found in comparing the SS male to the OS males.
Cole-Harding et al. (1988) have reported that OS DZ females had significantly higher baseline spatial scores on the Vandenberg modification of the Shepard-Metzler Mental Rotations test than SS females. Also, the improvement shown by the OS females over three trials was greater than for the SS females. "By the third trial, the scores of these OS/DZ females were not significantly different from those of their twin brothers." although sex differences were otherwise found. Notice that the effect was large enough to eliminate the sex differences. This was interpreted as "These results suggest the possibility that exposure to testosterone in utero improves spatial ability in females, thus supporting the theory that differences in prenatal exposure to testosterone are at least partially responsible for the gender differences in spatial ability."
Several studies of abilities and school performance have been examined to see if the performance of OS twins shifted in the direction of their co-twins. Naturally, such an effect would be expected only where performance showed an appreciable sex difference. Since it is intended to report these reanalyzes elsewhere, just a summary will be given here.
Record et al. (1970) report verbal ability scores derived from the British 11 plus examination (which they interpreted as a test of verbal intelligence) for 2164 twins born 1950 to 1957 in Birmingham, England. The females had the higher scores. The OS males did not differ appreciably in scores from the SS males. However, females were different. The 478 SS females had an average score of 97.6. The scores of 342 OS females averaged 96.15, which was below the 97.6 average of 478 SS females. Having a twin brother appeared to lower the female score 43% of the way towards the male average. While the published data does not give standard deviations for the different groups, given the large sample size, this effect is probably not due to chance. The direction of the effect is consistent with testosterone transfer from the male fetus to the female fetus, with a resultant brain masculinization lowering verbal abilities. Of course, an alternative is that having a brother makes studying harder, and this lowers the female scores.
Fischbein has reported (1978) a twin study of Swedish studentsU aptitude and school achievement that included 53-70 (number depending on the exact data reported) OS pairs. The results were reanalyzed (with the aid of some additional data kindly provided by Professor Fishbein) with the sex hormone transfer hypothesis in mind. It is intended to report the details elsewhere. For Swedish third grade children, there were no significant sex differences in performance, nor any evidence that performance depended on the co-twin's sex.
However, for sixth grade mathematics, when puberty is beginning to occur, there is evidence consistent with a possible hormonal effect. In this sample, as in other studies, males do better at mathematics than females. The 70 OS twin girls average 38.5, versus the average of 34.65 (a weighted average of Fischbein's MZ and DZ data) for SS twin girls. A male co-twin appears to have a significant positive effect on female mathematics performance. This is consistent with the joint hypothesis that hormones transfer, and that prenatal testosterone exposure contributes to the ability to learn mathematics.
Fischbein kindly provided data from the intelligence tests taken by these children. The only large sex difference was in perceptual speed, for which females were superior. A comparison of the SS females with the OS females showed that having a brother (rather than a sister) made a statistically significant difference. Having a twin brother appeared to shift the female score over halfway towards the male value.
Although a social explanation is possible, it is not obvious what type of socialization difference would cause females to have higher perceptual speed differences. Even less obvious is a social mechanism that would cause perceptual speed to vary with the co-twin's sex. However, the joint hypothesis that testosterone from a male fetus affects the female fetus, and that prenatal testosterone exposure reduces perceptual speed, can explain these results.
One other peculiarity emerged from studying this set of twins. Fischbein et al. (1991) reported that OS males were most popular, and OS females unusually unpopular. While sampling variability or social interpretations are possible (and offered in the paper), hormonal interpretations are also possible. Testosterone is believed to lead to higher levels of aggression (see Kemper, 1992, for citations), and to a personality high in psychoticism, characteristics that would lower popularity. The OS females, having been exposed to extra testosterone, would be less popular, while OS males, having had their testosterone partially off-set by female hormones, would be more popular than other males.
Husen (1959) reported a study of Swedish draftees (virtually all of the male population aged about 20 years) for 1949-1952 which includes data on SS male twins and OS ones. In most cases the SS male twins and the OS ones did not appear to differ in intelligence (statistical significance tests are not supplied). This is not surprizing since sex hormones probably do not affect intelligence.
However, an interesting peculiarity is that on all four subjects for which primary school marks were available, the OS males did better than either the MZ or the DZ SS twins. This is highly unlikely if the co-twin's sex is unimportant.
In comparing SS male twins with OS ones, it is found that OS ones have a statistically significant (5% probability) advantage in reading and history, and an almost statistically significant advantage in writing (t=1.78). Husen did not give standard deviations for each figure, but reported that standard deviations had an average of about .45 (p. 62), which is the figure used in these calculations. These are subjects in which girls generally do better. The better performance of OS males is consistent with female hormones passing into male fetuses, and this raising performance.
Husen also reported the percentages who were regarded as having had unsatisfactory school marks. OS draftees consistently had lower such percentages than SS ones, although the absence of standard deviations makes tests of statistical significance impossible. However, the same pattern for all four subjects would be unlikely if there were truly no difference.
Unfortunately, most studies that have utilized OS twins have been interested in estimating heritability or other genetic parameters, and have reported only the correlations between the members of twin pairs, not the absolute values tabulated by zygosity and sex. A finding of a number of these studies has been that the OS twins correlated about as closely, or even more closely, as SS twins. This is surprizing because one would have expected that the sex differences would lead to an appreciably lower correlation between the twins. For instance, the average over 9 studies of the correlation coefficients for the intelligence of OS twins is the same .53 average obtained for 11 studies of like sex twins (Erlenmeyer-Kimling and Jarvik, 1963). If hormones do transfer, any sex related differences within twin pairs would be attenuated, helping to explain these otherwise puzzling results.
Carter (1932) examined the extent to which twins, predominantly junior and senior high school students, were similar on the Strong Vocational Interest Inventory. His major finding was that MZ twins were more similar than SS DZ ones, indicating a substantial heritability. An unexpected finding was that OS twins correlated .30, which greatly exceeded the .20 correlation for SS males, and was approximately equal to the .32 correlation for SS females. This was puzzling since male and female interests are normally quite different (probably more so in the thirties), and the sex differences would be expected to cause the OS twins to correlate much less than the SS twins. It is possible that hormonal transfer reduces the effect of sex differences.
However, Carter and Strong (1933), in a later paper on sex differences, provided scores on the Strong Vocational Interest Blank for 34 OS pairs, and a second sample of 100 boys and 100 girls. To test the hormone transfer hypothesis, I compared the sex differences in the two samples. This did not support the hypothesis of OS twins affecting each other in the hypothesized direction. The differences in scores between the twins were actually a little greater than between the singletons (rather than less as predicted by the joint hypothesis that hormones transfer and that hormones affect vocational interest scores).
Incidentally, the Carter and Strong rationale for studying the OS twins was that they "were exactly paired for age, family environment, social and economic status, etc.", leaving only the sex related differences (whether due to biology or socialization). If hormones transfer, this is a incorrect strategy. Of course, they could not have been aware of this potential problem.
Koch (1966) has reported a very large number of measures for a sample of twins and matched controls, young children cross-classified by both zygosity and sex. Unfortunately, she had only 19 OS twin pairs. Thus, relatively few individual items show statistically significant differences. However, the effects were often in the direction hypothesized here. Her most revealing statement is (p. 163) "The striking divergence noted between the sexes when two groups, uniform in sex, are compared is not apparent in the two sex groups derived from opposite-sex pairs." This follows on the sentence that "The DZOFfs'did not seem to be 'masculinized' by their brother's influence as much as the brothers were 'feminized' by theirs." In the quote, DZOFf's refers to the OS female twins. Koch (writing in the environmentally oriented sixties) interprets the results as being due to differential socialization, but the result is also consistent with this paper's hormonal hypothesis.
The section on OS male twins states the following (p. 159). "These DZO" males were described as rather sober and subdued when compared with nontwins: they were rated lower in activeness, loudness, confidence, intensity, selfishness, and inclination to project blame. There was an intimation of a tendency on their part to play more with
girls. At first it was thought these DZOSm's showed their sisters' influence in being relatively feminine in behavior patternsQless selfish, more responsible, more obedient, less exhibitionist, less moody, and less active than parallel singletons.S Thus, in comparisons with the SS matched controls, the results are consistent with the males displaying more feminine behavior patterns. However, Koch follows the above immediately with "The DZOSm's did not, however, differ significantly from the DZSSm's in these traits. Hence it looks as if the former were, at least in part, conforming to generally approved behavior patterns rather than merely copying the sisters' behavior pattern." The interpretation is a socialization one, which Koch consistently prefers. However, she does go on to note that, "Although the DZOSm's differed insignificantly from the DZSSm's, the small differences between them were in the direction of the feminine." Her underlying tables (Table 52) supports her conclusion.
Inspection of the tables shows how large a difference can be and still not be statistically significant in a small sample. She constructed a "masculinity in attitude" by averaging normalized ratings by teachers for six traits (resistance, moodiness, revengefulness, tendency to project blame, tendency to tease, and social apprehensiveness) (pp. 92-93) with a mean of 4.0 and a standard deviation of 1.0. There were, after adjustment for social class, (Table 72) appreciable sex differences on these for MZ twins (4.11 for males versus 3.74 for females). Almost identical averages were computed for DZ twins (4.11 versus 3.72), suggesting a difference of .37 to .39 standard deviations between the sexes. However, for OS twins the difference shrunk greatly. The averages were 3.87 for males (i.e. the males moved almost two thirds of the way to the female value) versus 3.83 for females (the females moved over a quarter of the way towards the male values), leaving a male/female difference of only .04 standard deviations. This is rather striking. Essentially, the sex difference has been eliminated. However, this effect does not achieve statistical significance. Unfortunately, with her sample size, given the small size of the male/female differences. (With a population whose average age is six, these differences are not as large as they become later) a complete shift of the males to the female pattern would not be statistically significant.
Often, an effect that falls short of statistical significance is concluded not to exist. Here, where the effect as measured was large (an almost total disappearance of the sexual difference), a better conclusion is that there is a potentially important effect that should be studied further, but with a larger sample. Such a study is especially desirable now that there is a theory that predicts such a reduction in sex differences.
A similar result was found for her "femininity in attitude," which was based on the six traits of affectionateness, tenacity, obedience, cheerfulness, responsibleness, and friendliness to children. For MZ twins, the score were 3.73 for males versus 4.20 for females. Similar values were found for DZ SS twins, 3.77 versus 4.23. The sex effects were .47 and .46 standard deviations respectively. However, the male OS value was 4.05 and the female value was 4.19, shrinking the sex effect to .14 standard deviations. About 70% of the sex effect has disappeared through the males moving almost two thirds of the way towards the female values, while female values remain essentially unchanged. Again the results were reported as not statistically significant.
Some interesting results show up in the comparisons of twins with matched singletons (Table 73). There are statistically significant (P<.01) tendencies for the MZ females to be less masculine (.41 standard deviations) and more feminine (.34 standard deviations). For DZ females, the trends are in the same direction but much weaker (.13 and .15). As pointed out earlier, TambyRaja and Ratnam (1981) and Trapp et al. (1986) report higher levels of female hormones in mothers of twins. Thus, a possible hormonal explanation is that extra hormones from one female twin made the other twin even more feminine. A social explanation might involve modeling on the other twin. The size of the female twin feminization and demasculinization effect, when compared with the controls, were about the same (.13 for masculinity and .10 for femininity), which is not statistically significant.
Male twins with male co-twins were more masculine in all four comparisons with singletons (MZ and DZ each for masculinity and femininity), but the effects lacked statistical significance. SS males were less masculine (by .23 standard deviation units) and more feminine (by .12 standard deviation units), but neither effect was statistically significant.
One variable reported on by Koch and believed to be influenced by hormones (testosterone) is hand preference. Earlier, the Geschwind and Galaburda (1987, p. 141) hypothesis that OS twins should show more left-handedness was mentioned as an early speculation about sex hormone transfers in humans. Koch's data (Table 29) shows that the percentage of OS females that were left-handed (24% at some time before seven, and 18% at the time of the study) exceeded the corresponding percentages for OS males (18% and 12%), while in SS twins there was the usual tendency for the males to be more often left-handed. However, due to the small sample size (19 OS twins), the difference could be due to chance.
Left-handedness is often considered to be caused by testosterone. Koch's data (Table 30) provides some indirect evidence for a hormonal effect. She finds a significant tendency for the more dominant twin to be left handed (P reported as between .01 and .001), and also a tendency for the dominant twin to be more competitive (probability between .1 and .05). Both competitiveness and dominance are today believed to be testosterone related (Kemper, 1992).
Koch reports a tendency for physical defects of various types to be more common in OS females than in SS females. In SS twins, males had more defects than females, while in OS twins this was reversed (Table 12). In non-twin studies, males have more birth defects, an effect that is often attributed to testosterone. Thus, extra testosterone could create defects in female co-twins. The reversal of the usual effects in OS twins appears due to much higher defect rates in the OS females than in SS female twins.
Koch reports one other interesting finding. In 80% of the OS twins, the dominant twin was female. This is surprizing because in most social dyads (i.e. marriage) the male is usually dominant. She attributes this striking finding to the greater verbal skills (probably due to earlier maturation) of females in early childhood. (Her subjects were about age six.) This is plausible, but the females may have been assisted in achieving dominance by a reduction in the usual prenatal hormonal differences.
Another report of female dominance in OS twins exists. Zazzo (1960, p. 645) also reports that OS females are more commonly dominant in a sample aged from one year on up. High female dominance is reported for all ages, although it is not statistically significant for those aged 11 to 15, or for those over 20. It is especially striking that in the 16-20 age range, where males would normally be dominant, there were 14 female dominant pairs versus only 5 male dominant pairs, a statistically significant effect. Greater female prenatal testosterone exposure could be part of the explanation here.
The Koch study is one study in which the data is tabulated in such a way as to permit comparisons by sex of OS twins on a number of variables. Unfortunately, the small size of her OS sample and the fact that the research concerned young children at an age (about six years) when sex differences are small, prevent saying more than that her results are consistent with hormones produced by one twin somehow reaching and affecting the OS twin. Of course, if there really is an effect by which OS twins become more like each other with regard to sex typed behavior, it could be due to learning behavior from the OS twin, rather than to hormonal effects. If typical feminine or masculine attitudes were learned, one would expect that similar learning would occur in OS siblings.
Fortunately, Koch (1955) had done a very similar sibling study prior to her twin study. Children of the age used in the twin study, drawn from a similar population, and living with only one other sibling (the twins studied were selected to be without other siblings, which implied each twin had only a single sibling), were compared using the same psychological measures. Her description of this study makes no mention of a general tendency for siblings to become more like their siblings of the opposite sex. The strongest sex specific effect found is that "children with brothers were rated as more competitive, ambitious, enthusiastic, and less wavering in decisions than children with sisters. .... our results suggest that children with brothers are also less likely to build alibis and to be more tenacious of purpose than are those with sisters. In all of the traits mentioned, except ambition and tenacity, the effect of a male sibling is significantly different from that of a female only in the case of girls." (p. 47).
The author's impression, after reading both of Koch's reports, is that the sibling effects and the twin effects are quite different (the reports' formats differ, making exact comparison difficult). If she had thought her twin study results were similar to her earlier sibling
study, she would have commented on the resemblance. However, she makes no such comparisons. If learning from one's sibling was the primary mechanism by which one sibling affected another in both studies, one would expect the nature and direction of the effects to be similar in both studies. Both studies used children of the same age, from the same population, and controlled to have only a single sibling (the co-twin for the twin study). Such similarity is not apparent. The most parsimonious way to account for the absence of similar effects from siblings in the two studies is that the effects of a twin on its co-twin are caused by a different mechanism than the effects of one sibling on another sibling. Since hormonal effects are not found among non-twin siblings, but could be among twins, hormonal effects are an obvious candidate for this mechanism.
Mitchel et al. (1989) examined masculinity and femininity in twins, but since their sample included only 9 OS twin pairs, it has little power for hypothesis testing. However, on the two femininity scores reported (Table 1), there is a shift of the males towards greater femininity and of the females towards greater masculinity, such that the OS males actually have higher femininity scores than the females. No such shifts appear for the masculinity scores. Given the small sample size, it is not surprizing that none of these effects even approach statistical significance. Still, it is interesting that the effect is similar to that reported by Koch, and to that predicted by theory.
Although many of the effects reported by Koch and by Mitchel et al. are not statistically significant, the absence of statistical significance is consistent with a practically important effect. Given a theoretical reason for suspecting an effect, further research with larger samples would appear to be justified.
In Koch's study, a very striking difference between the OS group and the SS groups was a much higher rate of ocular defects, as evidenced by wearing glasses, in the OS twins, which was significant at the .001 level. The rate in twins was much higher than in the singleton population. The comparison of an OS twin rate of 38% with the 4.39% rate among the 3,000 controls is striking. She notes that retrolental fibrosplasia played a role. Severely premature children born in the period she studied were routinely given oxygen, which can lead to this condition. OS twins had greater rates of prematurity than did SS ones. There exists one other study that (Stocks and Karns, 1933) also reports a relatively high fraction of eye defects in OS twins.
The author has read much literature on myopia (see Miller, 1992) and found no discussion of hormonal effects. However, a possible hormonal mechanism can be imagined. The growth of certain brain tissues is affected by sex hormones, including estrogen. Possibly sex hormones affect differentially those tissues that become the lens and those that determine the size and shape of the rest of the eye. An eye whose axial diameter is too long or too short for the power of the lens produces an ocular defect, which is usually corrected by glasses. Natural selection should have produced a situation in which the different parts of the eye develop in the proper ratio in the typical hormonal environment for their sex. However, if the hormonal environment is unusual, and the eye's different parts are affected differentially, the result could be increased myopia.
Tanner (1989, p. 17) states that "the eye probably has a slight adolescent growth spurt." and that "although myopia increases continuously from at least age 6 to maturity, a particularly rapid rate of change occurs at about 11 to 12 in girls and 13 to 14 in boys, as would be expected if there was a rather greater spurt in the axial dimension of the eye than in its vertical dimension." Since the puberty growth spurt is caused by sex hormones, this would suggest some sensitivity to sex hormones then, and possibly earlier, making the above story more plausible. He also (p. 67) states that, "There is a clear adolescent growth spurt, coincident with the height spurt, in most facial dimensions, though not in the width of the eyes or the distances between them." Since the adolescent growth spurt is known to be caused by sex hormones, it is possible that some parts of the eye respond to sex hormones more than do other parts.
A large scale twin study conducted in London, summarized in Eaves, Eysenck, & Martin (1989) provided suitable data for another test. As part of a larger study of 1650 twins in the early 1970's, a sixty question Public Opinion Inventory (in Appendix D) had been asked dealing with such topics as sexual behavior, religion, and politics.
For the majority of questions there was not a statistically significant difference between the answers of those twins that shared a womb with one of the same sex, and those that shared a womb with one of a different sex. For females, in only 16 out of 60 questions was the difference statistically significant at the 5% level. However, only three such significant results would be expected from chance alone, thus suggesting that who one shared the womb with has important effects. However, for 14 of these 16 statistically significant comparisons, the difference took the form of the female twins answering the question more as males did. In only two cases (question 16, dealing with the importance of family and question 55 dealing with homosexuality) were the answers more in the feminine direction. For both of these questions the magnitude of the sex effect was small.
For males there were only five questions (3, 15, 30, 36, and 52) in which there were statistically significant effects for the sex of the womb mate. However, in all five of these cases, the male twins that shared a womb with a female had more feminine opinions than did those with twin brothers.
A preliminary question was whether the standard of comparison for a twin that shared the womb should be all twins that did not share a womb with an opposite sexed twins, or just the other dizygotic twin. The sex differences were quite similar, whether estimated from monozygotic or dizygotic twins (r=.836). Since there were appreciably more monozygotic twins (650 female twins and 240 male) than there were dizygotic twins (388 females and 118 males), more precise estimates of sexual differences could be obtained using all SS twins. Thus the difference between the answers of all the male and all the female same sex twins were taken as measuring the sex difference. The different sex twins were excluded because of the possibility that they might display somewhat different sex-typical behavior because of a different hormonal environments within the womb (testing this hypothesis was the study's goal).
As a measure of the sex effect the differences between the sexes in answers to the questions were computed (the average male score minus the average female score). The term feminine or masculine here refers merely to how females or males typically answered, rather than to any a prior designation of certain answers as typical of males or females. For each item, the difference in average scores was computed between the female twins with a twin brother and the female twins with a twin sister. This series measures the effect on a female twin of the sex of her cotwin. The two series (the sex series and the sex of wombmate series) had a correlation of .54. This is statistically significant, and far too high to have occurred by chance. Figure 1 plots the magnitude of the sex of wombmate effect against the magnitude of the sex effect. The sex of the wombmate effect is in the same direction as the sex effect. Adult females with twin brothers answered in a more masculine way than did those with sisters.
[Insert Figure 1 about Here]
A similar computation for the males gave a correlation of -.22. The negative sign means that the male answers were shifted towards the female answers. This is to say that their opinions became more feminine. The effect, while in the predicted direction, is not quite statistically significant (p=.096).
One statistical problem arises because the data on dizygotic females is used in calculating both the difference attributed to sex of wombmate and for calculating the difference between the sexes. The errors in the two variables are not independent. An error in this variable would contribute to both differences, possibly creating a spurious effect. For instance, suppose that by luck the female same sex dizygotic twins happened to answer a question so as to raise their score by .1 over what the score would be if we sampled an infinite number of twins on a question on which males had the higher score. This would reduce the male female difference. If female twins with male cotwins normally answered more in the male manner, this would also reduce the size of the wombmate effect. Now suppose luck (sampling variability) had the opposite effect of reducing these twins scores by .1. This would raise the sex difference and the wombmate effect. Thus, the effect of sampling variability is to increase the correlation.
While this effect is probably minor, it could produce a non-existent correlation. To eliminate this statistical effect, the wombmate effect was also calculated in a way that would not include the same twins used to calculate the sex effect. Thus the sex differences were determined using only monozygotic twins, and the wombmate effect using only dizygotic twins. Thus, the magnitude of the sex effect was calculated by subtracting the female monozygotic twins' average score
from the male monozygotic twins' average score. This increased the error due to sampling variability in the sex effect. The wombmate effect was then calculated using only dizygotic twins (again by subtracting the scores for those females who had a female wombmate from those who had a male wombmate). This had the additional benefit of making the wombmate effect depend only on dizygotic twins, eliminating any possible error from some uncontrolled for difference between monozygotic and dizygotic twins.
The effect was to reduce the correlation coefficient for female twins from .54 to .47. The percentage adjustment in the female values towards the male values was reduced from 31% to 24%. Since the estimate of the coefficient was still approximately four times its standard error, the effect still existed.
The cost of eliminating the statistical problem described is to reduce the size of the sample used to calculate the sex differences from 1038 to 650 and, more seriously, to reduce the size of the sample used to calculate the differences by twin type from the same 1038 to the 388 dizygotic female twins. This reduction in power was judged to be a greater loss than avoiding a small degree of common error variance, and emphasis is consequently placed on the above equations.
It was hypothesized that if the effect were related to what was being inherited (perhaps because the inheritance was related to testosterone), the correlation coefficient would be larger for the ten items found by the authors of the original study (see their p. 322) to have the largest additive component of variability. These are items believed to be heavily genetically influenced.
For females, the correlation coefficient was .77 (p=.01) for these items. The high correlation suggests that the genetic influence could be primarily through the level of testosterone, or another genetically influenced hormone.
For the ten items identified as having the largest between families component of environmental variance, the correlation coefficient was also high, .70 (p=.02). This high level could have happened if these items were also influenced by testosterone levels, but mothers differ in the testosterone levels their children are exposed to for environmental reasons (possibly stress or parity related). The high correlations found here could also be merely because these items are those heavily influenced by personal contact with other family members, and the most influential family member for a twin would be the cotwin.
The ten items with the largest within families component of variance had a correlation coefficient of -.10 (which is non-significant). Testosterone appears to have very little to do with these items.
The process was repeated for males. The correlation coefficients were +.10, +.25, and -.42. None were large enough to be statistically significant, probably because each group had only ten questions.
There is one previous study that can help cast some light on the hypothesized plausibility of the behavioral variables in question responding to hormones that could diffuse between fetuses. Saki, Baker, Jacklin, and Shulman (1992), working with identical and fraternal same sex twins (unfortunately, no OS twins were included), found evidence of significant genetic influences on levels of progesterone and estradiol, but not testosterone. However, evidence was found for a strong effect of shared environments on testosterone levels. If the public opinion questions that exhibit additive genetic effects and shared environmental effects are being influenced by hormones that move between fetuses, the results found would seem to be consistent with what is known about the inheritance of fetal hormones.
It should be noted that their work was done with umbilical cord blood, which reflects hormonal levels at birth. The maximum sex related differences in fetal testosterone levels occur well before birth. At birth the sexes actually overlap in hormone levels. If levels of sex hormones produce sexual differentiation in the brain, the effect most likely occurs well before birth, when the hormonal differences are largest. Thus, this study's failure to find evidence for inheritance of testosterone levels at birth does not show that earlier levels are not genetically determined. Indeed, the strong sexual differences found earlier make it very likely that genes play a major role (since sex is a genetically determined variable).
An obvious alternative to a pre-natal hormonal effect is a post-natal socialization effect. Since twins spend much time with each other when young and develop a special bond, perhaps they come to think alike, and this similarity persists into adulthood even though the twins
no longer live together. This could cause female twins with brothers to be more masculine in their opinions, and brothers with sisters to be more feminine. If females (at least as adults) are more affected by socialization by their twins than males are, a greater female than male shift could be explained. (The hormonal explanation would be that testosterone affects the females greatly, but that either little female hormones are produced by the ovary prenatally, or the male is little affected by any female hormones that may reach him.)
The closest analogy to an opposite sex twin relationship for intensity of feeling is marriage. Spousal opinions on public opinion questionnaires are known to be highly correlated. If spousal resemblances result primarily from living together, the opinions of married couples resemble each other more the longer the couples had been married. However, the resemblance of spouses does not appear to increase with age (a good surrogate for length of marriage), suggesting that living with and having extensive contact with someone of the opposite sex does not cause ones opinions to shift in the direction of that person's opinions (Eaves, Eysenck, and Martin 1989, p. 376).
Ernst and Angst (1983, pp. 173) report that both imitation (siblings resemble each other) and contrast (siblings contrast with each other) hypothesizes have been proposed for how siblings affect one another, and provide examples of studies reporting both types of effects. They conclude (p. 175) that,"The hypothesis that sex of sib has no general and lasting influence on personality has not been refuted." Leventhal (1970) reported in a study of college students that evidence (depending on the item) could be found for both contrast and imitation hypothesizes among second born males. Using the Gough Scale of Psychological Femininity, Rosenberg and Sutton-Smith (1968) reported significant imitative effects among college females from families with two siblings. While the score of females with sisters moved from 23.61 (for females with sisters) to 22.41 (females with brothers), a statistically significant shift, this move was still small in relation to the male-female gap (the femalesU brothers averaged 15.9). While the sex of the sibling had a statistically significant impact, the femininity score of a brother had virtually no impact (correlation of .03) on the femininity score of his sister, which goes against the imitative hypothesis. Landers (1970) found evidence for imitation effects in sports participation and femininity among college females.
In considering the above studies with college students, it should be remembered they are probably still living with, or very recently lived with, their siblings. However, the adult twins discussed here had typically lived apart from their cotwins for many years. Typically, among the adult twins, the member of the opposite sex they saw the most was a spouse rather than their cotwin. Given the amount of time spent with a spouse, having an adult opposite sexed twin probably added little to the exposure to the opinions of the opposite sex.
Butcher and Case (in press) present evidence from three large data sets that women who had brothers have more years of education (about .5 years more) than similarly situated women who had sisters. This goes against the belief that families would give preference to males in financing education, causing the females with brothers to obtain less education than those with sisters. The simplest explanation for this result in that brothers influence their sisters in the direction of more education, perhaps by causing them to value education more highly. This would be evidence of an imitation effect on the attitudes of sisters. Another possible explanation is that parents urge sons to obtain more education, and then to maintain consistency among the siblings, also urge the female siblings to obtain more education.
There is obviously considerable further research to be done. It would be very useful to directly confirm that the hormonal environment experienced by a twin does depend on the sex of its womb-mate. Even if there were no evidence for the joint hypothesis of hormonal transfer and hormonal effects on behavior and morphology, a showing that hormones transferred would be useful. If OS twins were known to develop in an unusual hormonal environment, but a certain aspect of their behavior, morphology, or disease risk did not depend on the co-twin's sex, this would be evidence that that aspect of behavior, morphology, or disease risk was not sensitive to prenatal hormone levels. This would be very important evidence for many questions concerning the role of prenatal hormones in the origin of sex differences
There are several possible ways to study hormone transfer. One would be to examine the hormonal content of umbilical cords and placentas at birth. Such measurements have been made in singletons (Maccoby et al., 1979) and have shown that umbilical cord hormone levels correlate with childhood behavior (Jacklin et al., 1983). Another approach might use blood samples from stillborn fetuses. The latter might provide evidence on hormonal environments experienced during earlier stages in pregnancy.
It is possible to measure amniotic fluid hormones. For instance, tables in both Schindler (1982, pp. 55-57) and Belisle and Tulchinsky (1980, p. 174) show that testosterone levels are much higher (especially in mid-pregnancy) in fluid surrounding males than in a females' fluid. The various amniotic fluid measurements are usually interpreted as indicating fetal levels. While the risk of taking amniotic fluid samples precludes doing so only for research purposes, amniotic fluid sampling is common enough so that studies of OS twins should be possible. The mothers of DZ twins are typically older than the mothers of singletons. There are often valid medical reasons for sampling amniotic fluid of older mothers. Thus, it may be easier to find twin samples than the number of twin pregnancies would suggest.
Since fetal testes produce the highest testosterone levels in mid-pregnancy, amniotic measurements might tell more about the relative levels during the critical periods for sex differentiation than umbilical cord measurements would.
It would also be useful to have information on physical variables, since these are unlikely to be influenced by merely interacting with an OS twin after birth. A particularly interesting variable to study might be pelvic shape, since this is reported to vary considerably between males and females. Individuals with pelvises shaped like those of the opposite sex have been reported to resemble that sex more in behavior (see Eysenck and Wilson, 1979, pp. 37-39).
Assuming that hormones do transfer between twins, this provides a whole new reason for studying twins. Not only can twins provide unique information about heritability, but they can provide evidence about the effects of prenatal sex hormones. There is evidence for prenatal hormonal effects on human behavior (see Hines, 1982; Levin, 1987, pp. 79-88; Halpern, 1992, pp. 110-120; Kemper, 1992, or Moir & Jessel, 1992 for reviews), but the topic remains highly controversial, with some believing all gender differences in behavior are due to differential socialization.
Inability to directly manipulate prenatal exposure to sex hormones has prevented the question from being definitely resolved for humans. Effects in experimental animals are well established. Some information has been obtained by studying children whose parents were given the nonsteroid estrogen diethylstilbestrol, which has been shown to have a masculinizing effect on behavior. However, for good reasons, this drug is no longer commonly given to pregnant mothers, and new subjects are no longer becoming available. Additional information has been obtained from observing children who have unusual hormonal environments due to various disease states (such as congenital adrenal hyperplasia), unusual karotypes (XO, XYY, etc.), or prenatal exposure to adrenal androgens (for a review see Reinisch & Sanders, 1992). However, suitable subjects are rare, and many of these conditions are diseases with multiple effects (not all of which are sex hormone related). Thus such evidence has not been decisive.
Fortunately, OS twins are common, and easily identified from birth records. Being a twin is not considered a disease state, and twins are broadly representative of the general population. Thus, the probability that OS twins are exposed to an unusual prenatal hormonal environment could provide a valuable new research tool.
Most twin studies have collected data only on SS twins. Even most of the large Scandinavian population based twin registers have included only SS twins. This is unfortunate, since it would have been very economical to include OS twins. The first stage in constructing such registers is to search the birth record for children with the same surnames born on the same date and at the same place. This gives a first approximation to a list of all twins, including the OS ones. Even if the ultimate goal is a register of SS twins, the list of OS twins excluded should be retained for future use. A major expense in constructing a SS twin register is determining whether the twins are DZ or MZ. This expense is avoided for OS twins. They are always DZ.
Fortunately, one population based twin register does include OS twins. This is the Finnish twin register, which was recently expanded to include 23,000 sets of twins born between 1958 and 1986 (Kaprio et al.,
1990). This register includes 7,922 OS twin pairs. This large sample could provide the basis for many useful studies of OS twins.
In the standard genetically oriented twin study, no useful data is obtained if only one twin responds. Fortunately, even if only one twin participates in a hormonally motivated study, useful data is still obtained. Since such studies require only comparing twins with male co-twins to twins with female co-twins, the co-twin's refusal to participate doesnUt make the the responding twin's answers useless. In questionnaire studies, data from only one member of a pair should be retained. It may be useful for studies of prenatal hormones.
If a questionnaire is returned without any identifying marks, as may be done in studies of such sensitive topics as sexual behavior, it is important to ask the co-twin's sex. This permits examining possible hormonal effects. For instance, this would be a very useful question if OS twins were studied to determine if differing exposures to prenatal sex hormones affected the incidence of homosexuality, as is widely suspected (see Blyne & Parsons, 1993 for a review). Although study of OS twins might cast light on the widely discussed hypothesis that lower than normal exposure to prenatal sex hormones caused or contributed to homosexuality, the major published studies to date have used only SS twins (see Bailey & Pillard, 1991; and Bailey et al., 1993, for references). Fortunately, a study is underway using Australian data by Martin and Bailey that does include OS twins.
The possibility that twins provide an opportunity to study individuals raised in an unusual hormonal environment could assist in obtaining financing for twin studies. In many studies, prenatal hormonal hypothesizes could easily be included. This might make it possible to obtain sufficient additional funding to cover at least the added cost of including OS twins (especially since there would be no expenses for determining zygosity).
Breast cancer is a medical area receiving increased funding where the study of OS twins might be useful. A recent hypothesis proposes that prenatal estrogen exposure affects breast cancer risk (Ekbom et al., 1992). Reversed cerebral asymmetry in breast cancer patients was recently reported (Sandson et al., 1992). The most plausible explanation for this surprizing finding (who would have guessed a brain-breast linkage?) is that prenatal hormones affect both cerebral asymmetry and vulnerability to breast cancer. A small study (Hsieh et al., 1992) found data consistent with the hypothesis that breast cancer risk was higher in twins, as Holm (1988) had found earlier.
If hormones do transfer between twins, this may influence the disease risks of OS twins. Having an OS twin could prove to be a risk factor for many diseases. So far the only disease for which it has been shown to be a risk factor is myopia (see above), but there appears to have been little investigation of this possibility for other diseases. Since there are large numbers of OS twins, and membership in a such a pair is readily determined, the possibility that it is a risk factor should be studied. Doing so could provide an additional source of financial support for twin studies.
There exists reasonably strong evidence that hormones can transfer from one fetus to another. This is made plausible both on the basis of evidence from the medical literature of hormones crossing the placenta, and from various observations that OS twins differ from SS twins, as this paper's hypothesis predicts. While some of the behavioral observations can be explained by social learning effects or other theories, some cannot (especially the dental asymmetry and ocular defect results). The best explanation of all of the results in the literature is that hormones do transfer. Alternatives involve a series of ad hoc explanations for each study, and are much less parsimonious.
Examination of how twin behavior is affected by the sex of the co-twin provides a methodology for studying the effects of prenatal hormone exposure, especially exposure to male hormones. Since ethical considerations forbid directly manipulating prenatal hormonal exposure, this may be one of the few ethical ways to study the possible hormonal basis for behavioral sex differences.
Babine, A. M. and Smotherman, W. P. (1984). Uterine position and conditioned taste aversion. Behavioral Neuroscience, 98, 461-466.
Bailey, J. M., & Pillard, R. C. (1991). A genetic study of male sexual orientation. Archives of General Psychiatry, 48, 1089-1096.
Bailey, J. M., Pillard, R. C., Neale, M. C., & Agyei, Y. (1993). Heritable factors influence sexual orientation in women. Archives of General Psychiatry, 50, 217-223.
Blyne, W. & Parsons, B. (1993). Human sexual orientation: The biologic theories reappraised. Archives of General Psychiatry, 50, 228-39.
Boklage, C. E. (1985). Interactions between same-sex- dizygotic fetuses and the assumption of Weinberg difference method epidemiology. American Journal of Human Genetics, 37, 591-605.
Boklage, C. E. (1987). Race, zygosity, and mortality among twins: Interactions of myth and method. Acta Geneticae Medicae et Gemellologiae, 32, 275-288.
Belisle, S., & Tulchinsky, D. (1980). Amniotic fluid hormones. In Tulchinsky, D. and Ryan, K. J. (Eds.). Maternal-fetal endocrinology. Philadelphia: W. B. Saunders, 169-195.
Butcher, K. F. & Case, A. (in press). The effect of sibling sex composition on women's education and earnings. Quarterly Journal of Economics.
Carson, D. J., Okuno, A., Lee, P. A., Stetten, G., Didolkar, S. M., & Migeon, C. J. (1982). Amniotic fluid steroid levels. American Journal of Diseases of Children, 136, 218-222.
Carter, H. D. (1932). Twin similarities in occupational interests. The Journal of Education Psychology, 23, 641-655.
Carter, H. D. & Strong, E. K. (1933). Sex differences in occupational interests of high school students. The Personnel Journal, 12, 166-175.
Clark, M. M. & Galef, B. G. (1988). Effects of uterine position on rate of sexual development in female Mongolian gerbils. Physiology & Behavior, 42, 15-18.
Clark, M. M., Bone, S. & Galef, B. G.(1988). Uterine positions and schedules of urination: correlates of differential maternal anogenital stimulation. Developmental Psychobiology, 22, 389-400.
Clark, M. M., Malenfant, S. A., Winter, D. A. & Galef, Jr., B. G. (1990). Fetal uterine position affects copulation and scent markings by adult male gerbils. Physiology & Behavior, 47, 301-305.
Clark, M. M., & Galef, Jr., B. G. (1989). Sexual segregation in the left and right horns of the gerbil uterus: the male embryo is usually on the right, the female on the left (Hippocrates). Developmental Psychobiology, 23, 29-37.
Clark, M. M., vom Saal, F. S., & Galef, Jr., B. G. (1992). Intrauterine positions and testosterone levels of adult male gerbils are correlated. Physiology & Behavior, 51, 957-960.
Clark, M. M., Tucker, L., & Galef, Jr., B.G. (1992). Stud males and dud males: intra-uterine position effects on the reproductive success of male gerbils. Animal Behavior, 43, 215-221.
Clemens, L. G. (1974). Neurohormonal control of male sexual behavior. In W. Montagna & W. A. Sadler. Reproductive behavior. New York: Plenum Press.
Cole-Harding, S., Morstad, A. L., & Wilson, J. R. (1988). Spatial ability in members of opposite-sex twin pairs. Behavior Genetics, 18, 710.
Dahlberg, G., (1926). Twin births and twins from a heredity point of view. Stockholm: Bokforlags-A.-B. Tidens Tryckeri.
Eaves, L. J., Eysenck, H. J. & Martin, N. G. (1989). Genes, culture, and personality. London: Academic Press.
Ekbom, A., Trichopoulos, D., Adami, H., Hsieh, C., & Lan, S. (1992). Evidence of prenatal influences on breast cancer risk. The Lancet, 340, 1015-1018.
Ellis, L. & Peckham, W. (1991,Winter). Prenatal stress and handedness. Pre- and Peri-Natal Psychology Journal, 6, 135-144.
Ellis, L., Ames, M. A., Peckham, W., & Burke, D. (1988). Sexual orientation of human offspring may be altered by severe maternal stress during pregnancy. The Journal of Sex Research, 25, 152-157.
Erlenmeyer-Kimling, L. and Jarvik, L. R. (1963). Genetics and intelligence: A review. Science, 142 1477-1479.
Ernst, C. & Angst, J. (1983). Birth order: its influence on personality. New York: Springer-Verlag.
Eysenck, H. J. & Wilson, G. (1979). The psychology of sex. London: J. M. Dent & Sons.
Fels, E. & Bosch, L. R. (1971). Effect of prenatal administration of testosterone on ovarian function in rats. American Journal of Obstetrics and Gynecology, 111, 964-969.
Fischbein, S. (1978). School achievement and test results for twins and singletons in relation to social background. In Nance, W. E. (Ed.). Twin research: Part A, psychology and methodology. New York: Liss.
Fischbein, S., Frank, Ove, & Cenner, S. (1991). Popularity ratings of twins and non-twins at age 11 and 13. Scandinavian Journal of Educational Research, 35, 227-238.
Fitch, R. H., Cowell, P. E., Schrott, L. M. & Denenberg, V. H. (1991). Corpus callosum: ovarian hormones and feminization. Brain Research, 542, 313-317.
Gandelman, R. (1986). Uterine position and the activation of male sexual activity in testosterone propionate-treated female guinea pigs, Hormones and Behavior, 20, 287-293.
Gandelman, R. (1992). Psychobiology of behavioral development. Oxford: Oxford University Press.
Gandelman, R., vom Saal, F. S., Reinisch, J. M. (1977). Contiguity to male fetuses affects morphology and behavior of female mice. Nature, 266, 722-724.
Geschwind, N. & Galburda, A. M. (1987). Cerebral lateralization. Cambridge, Mass.: MIT Press.
Glick, S. D. & Shapiro, R. M. (1988). Functional and neurochemical asymmetries. In Geshwind, N. & Galaburda, A. M. (Eds.). Cerebral dominance: The biological foundations. Cambridge, Mass: Harvard University Press, 147-167.
Gibson, M. & Tulchinsky, D. (1980). The maternal adrenal. In Tulchinsky, D. & Ryan, K. J. (Eds.). Maternal-fetal endocrinology. Philadelphia: W. B. Saunders.
Goy, R. W. and Resko, J. A. (1972) Gonadal Hormones and Behavior of Normal and Pseudohermaphroditic Non human Female Primates, in Eastwood, E. B. (Ed.), Recent progress in hormone research. New York: Academic Press, 707-735.
Halpern, D. F. (1992). Sex differences in cognitive abilities. Hillsdale: Lawrence Erlbaum.
Hendricks, S. E. (1992). Role of estrogens and progestins in the development of female sexual behavior potential. In Gerall, A. A., Moltz, H., and Ward I. L., Handbook of behavioral Neurobiology, Vol. 11, Sexual differentiation. New York: Plenum, 129-149.
Hines, M. (1982). Prenatal gonadal hormones and sex differences in human behavior. Psychological Bulletin, 92, 56-80.
Holm N. V., (1988). Studies of cancer etiology in the Danish twin population. I Breast cancer. In Gedda, L., Parisi, P. & Nance, W. (Eds.). Twin research 1: Part A. Twin biology and multiple pregnancy. 211-216.
Hsieh, C., Lan, S., Ekbom, A., Adami, H., Petridou, E., & Trichopoulos, D., (1992). Twin membership and breast cancer risk. American Journal of Epidemiology, 136, 1321-1326.
Husen, T., (1959). Psychological twin research. Stockholm: Almqvist & Wiksell.
Jacklin, C. N., Maccoby, E. & Doering, C. H. (1983). Neonatal sex-steroid hormones and timidity in 6-18-month-old boys and girls, Developmental Psychobiology, 16(3), 163-168.
Kaprio, J., Koskenvuo, M. & Rose, R. J. (1990). Population-based twin registries: illustrative applications from genetic epidemiology and behavioral genetics from the Finnish twin cohort study. Acta Geneticae Medicae et Gemellologiae, 39, 427-439.
Kemper, T. D. (1992). Social structure and testosterone. New Brunswick: Rutgers University Press.
Kinsley, C., Konen, C., Miele, J., Ghiraldi, L., & Svare, B. (1986). Intrauterine position modulates maternal behavior in female mice. Physiology & Behavior, 36, 793-799.
Kinsley, C., Miele, J., Konen, C., Ghiraldi, L., Broida, J., & Svare, B. (1986). Intrauterine contiguity influences regulatory activity in adult male and female mice. Hormones and Behavior, 20, 7-12.
Kinsley, C., Miele, J., Wagner, C. K., Ghiraldi, L., Broida, J., & Svare, B. (1986). Prior intrauterine position influences body weight in male and female mice. Hormones and Behavior, 20, 201-211.
Koch, H. L. (1966). Twins and twin relations. Chicago: University of Chicago Press.
Koch, H. L. (1955). Some personality correlates of sex, sibling position, and sex of sibling among five and six-year-old children. Genetic Psychology Monographs, 52, 3-50.
Landers, D. M. (1970). Sibling-sex-status and ordinal position effects on female sport participation and interests. The Journal of Social Psychology, 80, 247-248.
Lephart, E. D., Fleming, D. E., & Rhees, W., (1989). Fetal male masculinization in control and prenatally stressed rats. Developmental Psychobiology, 22, 707-716.
Levin, M. (1987). Feminism & freedom. New Brunswick: Transaction Books.
Leventhal, G. S. (1970). Influence of brothers and sisters on sex-role behavior. Journal of Personality and Social Psychology, 16, 452-465.
Maccoby, E. E., Doering, C. H., Jacklin, C. N., & Kraemer, H. (1979). Concentrations of sex hormones in umbilical cord blood: Their relation to sex and birth order in infants. Child Development, 50, 632-642.
Mankes, R. F., Glick, S. D., Van der Hoeven, F. & LeFevre, R. (1991). Alcohol preference and hepatic alcohol dehydrogenase activity in adult long-evans rats is affected by intrauterine sibling contiguity. Alcohol: Clinical and Experimental Research, 15, 80-85.
McDermott, N. J., Gandleman, R. & Reinisch, J. M. (1978). Contiguity to male fetuses influences ano-genital distance and time of vaginal opening in mice. Psychology & Behavior, 20, 661-663.
McFadden, D. (1993a). A speculation about the paralled ear asymetries and sex differences in hearing sensitivity and otoacoustic emissions. Hearing Research, 68, 143-151.
McFadden, D. (1993b). A masculinizing effect on the auditory systems of human females having male co-twins. Proceedings of the national Academy of Science, U. S. A., 90, 11900-11904.
Meisel, R.L., Ward, I.L. (1981). Fetal female rats are masculinized by male litter mates located causally in the uterus. Science, 213, 239-241.
Meulenberg, P. M. M. & Hofman, J. A. (1990). Maternal testosterone and fetal sex. Journal of Steroid Biochemistry and
Molecular Biology, 39, 51-54.
Miller, E. M., (1992). On the Correlation of Myopia and Intelligence. Genetic, Social, and General Psychology Monographs, 118, 363-383.
Mitchel, J. E., Baker, L. A.,& Jacklin, C. N. (1989). Masculinity and femininity in twin children: Genetic and environmental Factors. Child Development, 60, 1475-1485.
Moir, A. & Jessel, D. (1992). Brain sex. New York: Dell Publishing.
Phoenix, C. H. (1974). Prenatal testosterone in the non human primate and its consequences for behavior. In Friedman, R. C., Richart, R. M. & Vande Wiele, R. L. (Eds.). Sex differences in behavior. New York: Wiley, 19-32.
Phoenix, C. H., Goy, R. W. & Resko, J. A. (1968). Psycho sexual differentiation as a function of androgenic stimulation. In Diamond, M. (Ed.). Perspectives in reproduction and sexual behavior. Bloomington: Indiana University Press, 33-49.
Record, R. G, McKeown, T. & Edwards, J. H., (1970). An investigation of the difference in measured intelligence between twins and single births. Annals of Human Genetics, 34, 11-20.
Reinisch, J. M. & Sanders, S. A. (1992). Prenatal hormonal contributions to sex differences in human cognitive and personality development. In Gerall, A. A., Moltz, H., and Ward, I. L. (Eds.), Handbook of behavioral neurobiology, Vol. 11, Sexual differentiation. New York: Plenum, 221-239.
Resnick, S. M., Gottesman, I.I., & McGue, M. (1993). Sensation seeking in opposite-sex twins: An effect of prenatal hormones? Behavior Genetics, 23, 323-329.
Richmond, G. and Sachs, B. J. (1984). Further evidence or masculinization of female rats by males located caudally in utero. Hormones and Behavior, 18, 484-490.
Rhode Parfet, K. A., Lamberson, W. R., Ricke, A. R., Cantley, T. C., Ganjam, V. K., van Saal, F. S., & Day, B. N. (1990). Intrauterine position effects in male and female swine: subsequent survivability, growth rate, morphology and semen characteristics. Journal of Animal Science, 68, 179-185.
Rosenberg, B. G. and Sutton-Smith, B. (1968). Family interaction effects on masculinity-femininity. Journal of Personality and Social Psychology, 8,117-120.
Saki, L. M., Baker, L. A., Jacklin, C. N., & Shulman, I. (1992). Sex steroids at birth: Genetic and environmental variation and covariation. Developmental Psychobiology, 24, 559-570.
Sandson, T., Wen, P., & LeMay, M. (1992). Reversed cerebral lateralization in women with breast cancer. Lancet, 339, 523-524.
Schindler, A. E. (1982). Hormones in human amniotic fluid. Berlin, Springer-Verlag.
Schulster, D., Burstein, S. & Cooke, B. A. (1976). Molecular endocrinology of the steroid hormones. London: Wiley & Sons.
Solomon S. (1988). The placenta as an endocrine organ. In Knobil, E. & Neil, J. D. (Eds). The physiology of reproduction. New York: Raven Press, 2085-2091.
Stocks, P. & Karns M. N. (1933). A biometric investigation of twins and their brothers and sisters, Pt. II. Annals of Eugenics, 5, 1-55.
TambyRaja, R. L. & Ratnam, S. S. (1981). Plasma steroid changes in twin pregnancies. In L. Gedda, P. Parisi, & W. Nance (Eds). Twin research 1: Part A. Twin biology and multiple pregnancy. 190-195.
Tanner, J. M. (1989). Foetus into man. Cambridge: Harvard University Press.
Trapp, M., Kato, K., Bohnet, H. G., Gerhard, I., Weise, H. C., & Leidenberger, F. (1986). Human placental lactogen and unconjugated estriol concentration in twin pregnancy: Monitoring of fetal development in intrauterine growth retardation and single intrauterine fetal death. American Journal of Obstetrics and Gynecology, 55, 1027-31.
Tulchinsky, D. & Ryan, K. J. (1980). (Eds). Maternal-fetal endocrinology. Philadelphia: W. B. Saunders.
vom Saal, F. (1983). Variation in infanticide and parental behavior in male mice due to prior intrauterine proximity to female fetuses: Elimination by prenatal stress. Psychology & Behavior, 30, 675-681.
vom Saal, F. (1989). Sexual differentiation in litter-bearing mammals: influence of sex of adjacent fetuses in utero. Journal of Animal Science, 67, 1824-1840.
vom Saal, F., Bronson, F. H. (1980). Sexual characteristics of adult female mice are correlated with their blood testosterone levels during prenatal development. Science, 208, 597-599.
vom Saal, F., Grant, W. M., McMullen, C., Kurt W., & Laves, E. (1983). High fetal estrogen concentrations: correlation with increased adult sexual activity and decreased aggression in male mice. Science, 220, 1306-1308.
Whitfield, J. B. & Martin, N. G. (1992). Sex differences in alcohol, use, peak concentration, and post-alcohol test performance in humans; A test of the effects of prenatal environment, Manuscript.
Zazzo, R.(1960). Les jumeaux: Le couple and la personne. Paris: Press Universitaire de France.
Zondeck, L. H., & Zondeck, T. (1979). Observations on the determination of fetal sex in early pregnancy. Contributions to Gynecology and Obstetrics, 5, 91-108.
![[The Latest!]](../images/
latest90.gif){kind=small}
![[Taboos]](../images/
taboo90.gif){kind=small}
![[Stalkers]](../images/
stal90.gif){kind=small}
![[Thoughts]](../images/
though90.gif){kind=small}
![[Library]](../images/
librar90.gif){kind=small}
![[Web Links]](../images/
links90.gif){kind=small}