Minnesota Twin Family Study

The Minnesota Twin Family Registry:

Some Initial Findings, page 2

D.T. Lykken, T.J. Bouchard, Jr., M. McGue, A. Tellegen

Department of Psychology, University of Minnesota, Minneapolis, USA

Please be patient. Images under construction. Thank you.

DEMOGRAPHIC CHARACTERISTICS OF THE REGISTRY TWINS

Because the total of Black, Asian, and native American minorities is less than 2% in the Minnesota population, we do not identify race nor make any racial comparisons.

Rates of Twinning

As suggested in Fig 1, there was a decline in the rate of DZ twinning in Minnesota from about 7.0 pairs per thousand confinements in 1936-55 to about 5.0 pairs per thousand in 1971-83; a similar decline has been reported for the US, Britain and northern European countries, and in Australia [1, 11, 12, 20, 23]. The rate of MZ twinning, in contrast, 3.5 in 1936-55 and 4.0 per thousand in 1971-81, has increased slightly. Doherty and Lancaster [12] report a similar increase in the MZ rate in Australia, from 3.6 in 1936-55 to 4.2 in 1971-81, while Bressers et al [5] report similar trends in several European countries. This increase in the rate of live-born MZ twins, at least in Minnesota, seems likely to have been due to to improvements in prenatal and obstetrical care. As shown in Table 3, neonatal mortality is higher in male than in female twins, higher also in MZ than in DZ twins, and decreased by more than 50% in Minnesota from 1936-55 to 1971-81. If the MZ mortality rate had been as high in the latter period as in the former, the viable MZ birth rate would not have increased at all. Additional evidence comes from a comparison of the distributions of gestational ages, shown in Fig. 5. In 1971-81, 15.7% of the total group of surviving male pairs had gestational ages of less than 32 weeks, as compared to 1.5% in 1936-55. Had the 14.2% surplus not survived in 1971-81, the MZ rate then would have been (0.858) x 4.0 or 3.4 per thousand, close to the 1936-55 MZ birth rate. By this same reasoning, the rate of DZ conceptions has decreased even more than is indicated above; were it not for this decrease in perinatal mortality, the viable DZ birth rate would have declined from about 7 to about 4.3 (instead of 5) per thousand.

Twin pairs lost by infant death (%) -a
			MM	FF	OS	MZ-c
1936-55	Mean		23.70	18.30	17.10	25.30
	SD		5.87	5.10	5.41	5.22
1971-81	Mean		11.90	8.4	7.3	12.7
	SD		2.33	3.44	2.69	2.66
Infant mortality risk-b
			MM-MZ	MM-DZ	FF-MZ	FF-DZ
1936-55			0.149	0.099	0.112	0.076
1971-81			0.073	0.044	0.050	0.030

a Percent of live-born twin pairs in which one or both twins died before age 6 months. There was a highly significant decrease in twin-infant mortalitiy from 1936-55 to 1971-81.

b Risk of mortality within frist 6 months after birth.

c MZ figures estimated from MZ infant mortality risks computed separately for male and female twins, using known opposie-sex, male and female same-sex twin mortality data and assuming (1) that male risk is the same in OS and SS DZ pairs, and (2) that the ratio of male to female risk is the same in MZ as in DZ twin pairs.

Fig. 5 Gestational age at birth as estimated from the time form the mother's last menses, for the male twins of the older cohort, born 1936-55, and the younger cohort, born 1971-81. In 1936-55, more than half of the live-born MZ and DZ pairs were carried to full term, ie, 40 weeks. In 1971-81, the mean nominal gestational age had decreased from 38.5 to 33.8 weeks, likely due to better methods of assessing fetal maturity accompanied by much greater use of cesarean delivery.

Maternal Age

The rate of DZ twinning increases with maternal age more sharply than the rate of MZ twinning [6] which suggests that mothers of DZ twins should average somewhat older than the mothers of MZs. Table 4 confirms this expectation for mothers of both male and female twins. Surprisingly, the mothers of the younger cohort of male twins were significantly younger than the mothers of the male twins born in 1936-55. Our mean maternal age for the 1971-81 mothers, 26.8 years, can be compared with the mean of 26.1 years for mothers of all twins born in Minnesota and surrounding states in 1982, as reported by Allen [1].

	Female twins		Male twins
	1936-55		1936-55		1971-81
	MZ	DZ	MZ	DZ	MZ	DZ
Mean	28.10	29.30	28.20	29.70	26.60	27.10
SD	6.42	5.90	6.36	5.64	4.90	4.59
N	1,356	1,299	1,007	1,017	513	338
MZ vs DZ
CR-b	5.0, P < 0.0001		5.6, P < 0.0001		1.5 (ns)
Older vs younger males
CR-b	MZ = 5.4, P < 0.0001		DZ = 8.5, P < 0.0001

a -Only for male twins has so far been collected for the younger cohort.
b -CR is the ratio of the indicated difference to its standard error.

Gestational Age

Until 1981, the nominal gestational age, computed as the time from the mother's last menses, was listed on Minnesota birth records. More than half of the Minnesota twins born in 1936-55 were carried to 40 weeks, the normal term for singletons. If we define prematurity as a gestational age less than 37 weeks, following Watson and Campbell [45], then about 21% of the twins were born prematurely. It is this higher risk of prematurity, of course, that is responsible for the higher rate of perinatal injury in twins than in singletons which, in turn, increases the within-pair variance on traits susceptible to the effects of such injury, thus leading to underestimates of the heritability of such traits in single-born persons [33].

The 1,775 MZ twins in our older cohort had a mean gestational age 0.3 weeks shorter than that for the 2,265 DZs; this small difference is statistically significant and identical to the difference reported in a study of 1,855 twin pairs born in Belgium from 1964 to 1987 [41]. Our older cohort, born from 1936-55, had a mean gestational age of 38.5 weeks, longer than the Belgian mean, 37.0 weeks, which is the value reported by Bulmer [6] as the average length of a twin pregnancy, and much longer than the mean. 33.8 weeks, for our younger cohort, born 1971-81. In Fig. 5 are plotted gestational ages for the male twins of the older and younger cohorts separately. Only about 3% of the younger twins were carried to 40 weeks, compared to 57% of the older group; 82% of the younger cohort, versus 20% of the older, were more than 4 weeks "premature". This marked change, over 20 years, in the distribution of gestational ages suggests both an increased ability to sustain life in very premature infants with gestational ages suggests both an increased ability to sustain life in very premature infants with gestational ages less than 32 weeks, and also a great increase in the proportion of twins for whom birth is induced early or who are delivered by cesarean section (CS) at nominal gestational ages from 32 to 40 weeks.

Fig. 6 Birth weight as a function of nominal gestational age for male twins born 1936-55 and for male twins born 1971-81. The latter curve suggests that better-developed infants were being taken by cesarean section at earlier nominal gestational ages.

Birth Weights

The distributions of birth weights of the Registry twins were compared by sex and by twin type. Contrary to the recent Belgian findings [41], male twins were slightly heavier than females (94.5 vs 89.3 oz, P < 0.0001) and DZs were significantly heavier (P < 0.0001) than MZ twins. Birth weight increases, of course, with gestational age, as is shown in Fig. 6. The linear increase for the older cohort can be compared with Bulmer's Fig. 3.1 [6, p.49]; the two curves are quite similar. The birth weights of the 1936-55 DZ twins average 6.3 oz heavier (P < 0.001) than in the Belgian study, about the difference to be expected given the 1.5 week longer mean gestational age.

Fig. 7 Distributions of birth weights for the male twins born in 1936-55 and 1971-81. The similarity in the curves, except at the extremes, supports the view that the younger cohort, although their nominal gestational age averaged nearly 5 weeks less than for the older cohort, were selectively delivered on the basis of fetal maturity. (Only male twin data are available on the younger cohort.)

the birth weights for the male twins born in 1971-81, also shown in Fig. 6, increase sharply between 31 and 32 weeks of gestational age. This supports the conjecture that, after a nominal gestational age of 32 weeks, many of these younger twins were taken by CS, but only after determining that they were sufficiently well developed. It should be remembered that "gestational age" as recorded on the birth record is inexact and that modern sonography provides a more accurate indes of true fetal maturity. Thus, although the younger cohort were being delivered at nominal gestational ages that seem alarmingly short (Fig. 5), their actual maturation at delivery was as far along as for the older cohort. The distributions of birth weights for the older and younger male twins, shown in Fig. 7, are virtually identical; there were more very large twins in 1936-55, that would have been taken by CS in 1971-81, and more very small twins in 1971-81 that were live born due, presumably, to improved obstetrical techniques.

Years of education		0-9	10-11	HS	13-15	Coll.	Grad.	N
Males		2	4	41	22	16	15	2,449
Females		1	3	49	24	15	9	3,133

Occupational code-a			1		2		3		4		5		6		7		8		N
Twins	Males		10		13		21		11		29		12		4		0		2,462
	Females		2		16		19		28		5		11		5		14		3,147
Spouses	Males		11		14		21		12		27		12		3		0		2,469
	Females		2		15		16		23		4		9		3		29		1,857

Family income (in $1,000s)		0-10	10-20	20-30	30-40	40-50	50-75	75-up	N
Males		5	16	26	22	13	11	5	2,367
Females		8	20	26	20	13	9	5	2,972

a - The Occupational Code is from Hollingshead and Redlich [21]; #1 is professional, #2 is managerial, #7 is unskilled labor, #8 is homemaker.

Social Class and Education

Table 5 shows that 5% of the adult Registry twins did not complete high school, 45% stopped with high school graduation, and about 27% are college graduates. For the male twins, and the spouses of females twins, the most populous Hollingshead socioeconomic status (SES) categories are #3 (decorator, insurance agent, manager), #4 (bank clerk, farmer, Army non-com.), and #5 (hair stylist, locksmith, machinist.). Ten percent of the male twins (11% of the husbands) are professional people (SES #1) while some 16% of both groups are in category #6 (bartender, machine operator, meter reader, etc.) or #7 (laborer, car washer, welfare recipient). The modal combined family income was $20 to $30 thousand, about 25% earned less than $20 thousand, and about 15% had a combined income of more than $50 thousand. Thus, the Minnesota Twin Registry includes a broad cross-section of this north central state with greater representation of less well-educated, less affluent, working class and rural citizenry than is common among volunteer samples participating in psychological research.

These 29 to 52 year old twins averaged about 13.5 years of education, just over 2 years more than their parents received. The correlation between mother's and father's years of education was 0.53 for 269 pairs whose parents both completed BQs; the twins' estimates of these parents' education were accurated enough to yield a spousal correlation of 0.52. The correlation between mothers' and fathers' education as reported by some 2,800 twin pairs was 0.55.

Thus, it appears that there was fairly strong assortative mating in the parental generation, considerably stronger than in the twins' generation where the correlation between twins and their spouses is only 0.37. A similar decrease in spousal correlation, from 0.86 to 0.52 over about one generation, has been reported for a large sample of Norwegian male twins [36]. Years of education completed by both male and female Registry twins correlated 0.48 with their mothers' and 0.52 with their fathers' educational attainment.

Fig. 8. Scores on Positive Emotionality (extraversion) and Negative Emotionality (neuroticism) as a function of years of education for 2,242 female twins born from 1936 to 1955. The ordinate is in T-score units (mean = 50, SD = 10). The correlation for males is slightly higher than for the females shown here.

Some 2,200 female and 1,100 male twins completed the Multidimensional Personality Questionnaire (MPQ) [38, 40] and it was found that the super-factors, Positive and Negative Emotionality, vary systematically with educational attainment. As shown in Fig. 8, Positive Emotionality (PE: a kind of "talent for happiness", often called Extroversion) is about one SD higher in women with doctorates than in those who did not complete high school; the correlation between educational attainment and PE is 0.20 in both sexes. Negative Emotionality ("talent for misery" or Neuroticism) also varies systematically with educational attainment but in the opposite direction; the correlation is -0.20 in women and -0.25 in men.

Fig. 9. Scores on Alienation and Self Esteem as a function of Socioeconomic Status (inverted so that 7 = professional). Ordinate is in T-score units. Male correlations are slightly higher.

The twins averaged a higher SES than their fathers, 4.14 vs 3.83 for 2,369 male twins, and 4.21 vs 3.85 for 2,643 females. These differences, while small, were highly significant with these large samples. The correlation with the fathers' SES was 0.35 for males, and the 0.27 for females. SES, which is correlated about 0.60 with years of education, also showed similar personality correlates. As illustrated in Fig. 9, for example, low SES female twins were about one SD lower than high SES twins on a questionnaire measure of Self Esteem and about one SD lower than high SES twins on a questionnaire measure of Self Esteem and about one SD higher on the Alienation scale of the MPQ: the correlations were 0.20 and -0.18, respectively, for women; 0.28 and -0.31, respectively, for men. (Twins who identified themselves as homemakers were excluded from these computations since the socioeconomic status of this category is ambiguous.)

	MZ twin pairs			DZ twin pairs			OS twin pairs
	MM	FF		MM	FF
	(N=433)	(N=392)		(N=632)	(N=571)		(N=380)
Years of education	0.64	0.66		0.44	0.50		0.40
Occupational Status	0.59	0.53		0.38	0.32		0.24
Spouse's occ. status	0.28	0.38		0.06	0.23		-0.01
Gross family income	0.53	0.41		0.37	0.25		0.29

a -Intraclass correlations used for same-sex twins, product-moment correlations for opposite-sex pairs.

The intrapair twin correlations for SES and education are shown in Table 6 and indicate significant effects of both genetic and shared environmental influences. The Falconer heritabilities range from 0.30 to 0.44.

	Ever married	Ever divorced	Never married	Total
Male twins
MZ	730	133 (18%)	146 (17%)	876
DZ	665	129 (19%)	128 (16%)	793
OS	340	73 (21%)	42 (11%)	382
Female twins
MZ	1,128	207 (18%)	156 (12%)	1,284
DZ	1,024	223 (22%)	117 (10%)	1,141
OS	349	93 (27%)	30 (8%)	379

Number of offspring-a
	Parents	Offspring	Ratio
Male twins	2,036	3,701	1.82
Female twins	2,647	5,311	2.01
Total: 4,683 twins produced 9,012 offspring, ratio = 1.92

a -Ever-married twins only.

Marriage, Offspring, and Divorce

Some 17% of MZ twins, and 16% of DZs, had never married (Table 7). Once married, however, there is a slightly lower divorce rate among MZ twins (18%) than among DZs (19%). Nearly 1/5 of these twins, aged 29 to 52 at the time of reporting, had produced no offspring. The total number of living children produced by all 4,683 twins is 9,012, giving a reproductive ratio of 1.92. Separately by birth-year cohort, we computed the correlation between number of offspring and the number of siblings in the family in which the twins were reared; the root-mean-squared weighted average of these correlations was 0.13 for both males and females, small but statistically significant for these large samples.

Fig. 10. Number of offspring as a function of the age of the (twin) parent. Although these twins are from 10 birth-year cohorts, their ages at time of data collection yielded only 4 data points. The younger twins started their families later and will probably have smaller families, on average, than the older twins.

Is Fecundity Heritable?

The average number of offspring increases steadily over the full age range of these Registry twins (see Fig. 10) even though the 4 oldest cohorts were aged 45 to 50 at time of testing. This suggests that there are two effects working additively; the younger twins have not yet completed their families while the older twins had larger families on average than the younger ones will ever have. At the time these older twins were beginning their families, the frequency of newborn infants in families of size N in Minnesota was a linear-decreasing function of N (see Fig. 11). Over the ensuing 20 years, as the younger Registry twins were beginning their families, this function became increasingly sigmoid; family size decreased rapidly during this period while the age of Minnesota mothers at primiparity increased.

Fig. 11. Secular trend in parity of current birth, 1963 to 1983, all live births, from Minnesota Health Department records. In 1978-83, 42% of all births were first-born children, as compared with 26% in 1963-67. These curves show a regular trend toward smaller families over the 20-year period.

To compute within-pair similarity in number of offspring, independent of this confounding secular trend, requires that we somehow remove the trend from the raw data. We did this in two ways. For each of the 10 birth-year cohorts, and for the sexes separately, we computed the mean and variance of number of offspring and also the within- and between-mean squares. We first excluded all pairs in which either twin had never been married, ensuring that both members of the included pairs had been equally "at risk" for the production of offspring.

We then divided each within- and between-mean square by the total variance for that cohort, obtained the overall within- and between-mean squares as the Ni-weighted average of the 10 values (Ni = number of pairs in the ith cohort), and computed the grand average intraclass correlation. Our second method of removing the secular trend was to convert each twin's number of offspring and dividing by that cohort's standard deviation. We then computed the overall intraclass correlation from these deviation scores. Both sets of correlations are given in Table 8.

		MZ twin pairs			DZ twin pairs			OS twin pairs-b
		MM	FF		MM	FF
		(N=317)	(N=482)		(N=281)	(N=446)		(N=314)
Method 1		0.29	0.32		0.12	0.18		0.09
Method 2		0.30	0.36		0.14	0.26		0.11

a -Only pairs in which both twins had been married at least once were included in the analysis. The secular trend of offspring vs parental age was removed in two ways (see text) before computing the correlations.
b -OS twin values are product-moment correlations.

The average correlation for number of offspring, weighted by sample size, is 0.322 for MZ twins and 0.160 for DZs, yielding a heritability of 0.32. Since both parents play a role in determining the number of offspring, except when one spouse is biologically infertile, the upper limit of heritability in fecundity would seem to be about 0.5. Thus, an estimated heritability as high as 0.32 must be regarded as substantial and, perhaps, surprising.

Is the Risk of Divorce Heritable?

Again limiting consideration only to ever-married twins, the divorce rate for these 30- to 50-year old twins was more than double the rate for their parents and the risk of divorce in twin offspring of divorced parents was about 46% higher than for offspring of parents who had not divorced (Table 9). Cotwins of divorced MZ twins were nearly three times as likely to be divorced themselves than were cotwins of still-married twins. The increase in risk for DZ cotwins, 46%, was about 1/4 that for MZ pairs but also statistically reliable.

When divorce rate (proportion ever divorced) is plotted against age as in Fig. 12, the inverted U-shaped curve appears to be the sum of two opposing secular trends; the younger twins are moving through the period of greatest divorce risk while the older twins, nearly through the risk period, display the higher threshold for divorce characteristic of their generation. The lower curve in Fig. 12 represents the proportion of the parents of the twins in each cohort who had ever been divorced from each other. Having remained married long enough to produce the twins, these parents are somewhat selected for lower than average risk but their curve illustrates the lower divorce rate (higher threshold) of the parental generation. It seems likely, had we recruited another 10 birth-year cohorts of older twins, the oldest of which would be about the age of the parents of our youngest twins, that the extended right side of the upper curve in Fig. 12 would have merged into the parental divorce-rate curve, ie, the left side of the lower curve in the figure.

			MZ twins			DZ twins
			MZ twin offspring			DZ twin offspring
			Not divorced	Divorced		Not divorced	Divorced
Parents
	Not divorced		941	229		1,292	368
	Divorced		68	26		1,292	368
			x-squared = 3.5 (P~0.05)			x-squared = 11.0 (P<0.001)
			P(D|parents D) = 0.277			P(D|parents D) = 0.338
			P(D|parents not D) = 0.196			P(D|parents not D) = 0.222

			MZ twins			DZ twins
			Twin B			Twin B
			Not divorced	Divorced		Not divorced	Divorced
Twin A
	Not divorced		433	68		554	158
	Divorced		77	55		136	65
			x-squared = 52.7 (P < 0.001)			x-squared = 8.7 (P < 0.001)
			P(D|cotwin D) = 0.413			P(D|cotwin D) = 0.306
			P(D|cotwin not D) = 0.143			P(D|cotwin not D) = 0.210
Increased risk if cotwin is divorced
MZs = (0.413 - 0.143)/0.143 = 189%				DZs = (0.306 - 0.210)/0.210 = 46%

In any case, just as for assessing twin similarity in number of offspring, above, these secular changes in divorce rate pose a problem since they might tend to inflate twin concordance. For a dichotomous variable such as divorce, there is now wholly satisfactory way to partial out the secular variation. The main age-related change in twin divorce rate is the increase from 14 to 22% for twins aged 29 to 33. Therefore, all twins younger than 33 years were omitted from the calculations in Table 9 although, as it turned out, the MZ correlation was unchanged by this omission.

Fig. 12. Regression of divorce rate on age for ever-married twins and their parents. For the Parents' curve, the abscissa represents the age of their twin offspring; the parents of the younger twins were some ten years older than the age of the oldest twins. The right half of the twins' curve, together with the curve for the parents, illustrates the increase in cumulative divorce risk as the twins move through the risk period.

Conventional wisdom would suggest that the increase in risk for divorce among offspring of divorced parents is environmentally transmitted. Children from broken homes are expected to be less well-adjusted, more prey to later marital problems. Moreover, they have been raised by role-models whose marriages are characterized by conflict and intransigence and who "solved" these problems by recourse to divorce. Such reasoning would predict an increased risk among children of divorce. Only a genetic model, however, comports with all the data, eg, the fact that the increase in divorce risk for DZ cotwins of divorced probands is about equal to the increased risk for offspring of divorced parents, and less than half the increased risk for cotwins of divorced MZ twins. Therefore, a reasonable interpretation of the parent-offspring correlation is that the children of divorce are more at risk for divorce themselves largely because of the genetic proclivities that they inherited from their affected parents.

					Height (in)			Weight (lb)			Ponderal index
			N		Mean	SD		Mean	SD		Mean	SD
DZ twins
	males		361		70.3	2.7		176.1	26.7		12.60	0.53
	females		559		64.5	2.6		135.7	27.2		12.64	0.70
MZ twins
	males		384		70.3	2.9		176.1	24.8		12.59	0.57
	females		569		64.4	2.6		136.1	26.8		12.61	0.70
Spouses
	males		482		70.8	2.6		184.4	28.4		12.49	0.56
	females		414		64.5	2.4		137.0	23.8		12.59	0.68
Siblings
	males		67		69.8	3.6		181.3	30.3		12.38	0.52
	females		104		64.9	3.3		147.1	32.4		12.39	0.79
Parents
	males		112		69.9	2.9		176.1	24.8		12.59	0.57
	females		569		64.4	2.6		136.1	26.8		12.61	0.70
Offspring
	males		116		70.5	3.1		169.4	24.0		12.97	0.61
	females		175		64.9	2.6		130.0	19.9		12.86	0.64

Anthropometric Data

Self-reported heights and weights (Table 10) indicate that the participating MZ and DZ twins do not differ in stature not are the twins shorter on average than their singleton siblings or parents. The twins weigh slightly but significantly less than the contempoeraneous spouse sample but, as expected, the young-adult offspring average thinner than their parents.

Intra-class correlations of these variables, for all twins who returned the BQ, are given in Table 11. The correlations for self-reported height are slightly lower than those found when stature is measured in the laboratory [28].

With respect to birth weight (obtained directly from birth records), the DZ correlations are as high as for MZ twins. Birth weight is, of course, largely determined by gestational age at parturition, ie, by an environmental influence common to both twins. We attempted to remove the effect of gestational age by converting each weight to a standard score relative to all the twins of that sex and twin type who were born at the same gestational age as the index case. Within-pair correlations based on these corrected scores are somewhat smaller than before but still as large for DZ as for MZ pairs. Since about half the pairs were born during week 40 of gestational age, we also computed the correlations of (uncorrected) birth weight for this group separately. Once again, there is no suggestion of an MZ:DZ difference, ie, no evidence of heritability of birth weight. Vlietinck et al [41], modeling with LISREL the birth weight of 1,855 Belgian twin pairs born since 1964, attributed 22.5% of the variance to additive genetic factors; perhaps their prospective study achieved more accurate estimates of gestational age than were available to us. In any case, it is notable that birth weight is the only dependent variable we have so far encountered on which DZ twins are as similar as MZ twins for both sexes, and the only variable for which shared environment is the major source of phenotypic variance.

				MZ twin pairs			DZ twin pairs			OS twin pairs-b
				MM	FF		MM	FF
N				318	471		288	407		221
Birth weight
	Uncorrected			0.73	0.71		0.56	0.70		0.64
	Corr. for gest. age			0.57	0.60		0.58	0.60		0.47
	Gest. age = 40 wk			0.65	0.65		0.56	0.64		0.42
	(N)			(163)	(234)		(166)	(234)		(115)
N				435	638		392	570		380
Height (self-reported)				0.86	0.90		0.44	0.44		0.40
Weight (self-reported)				0.76	0.79		0.32	0.31		0.29
Ponderal index				0.73	0.78		0.30	0.36		0.38

a - The precision of the estimates is high due to large N, eg, the 95% confidence interval for the MZ female correlation for height runs from 0.884 to 0.914.
b - The values for the OS pairs are product-moment correlations, male vs. female.

Frequency of Contact

As can be seen in Table 12, female twins stay in closer touch than male twins do and MZ twins speak together, either in person or by telephone, significantly more frequently on the average than do DZ twins. Thus, for example, 29% of MZ females, and 23% of MZ males, speak together daily, as compared with 15% and 8% of female and male DZ pairs, respectively.

More detailed assessment of perceived closeness was conducted on twin cohorts born in 1946 and 1952. Approximately 100 pairs each of MZ, SS-DZ and OS-DZ twins responded to a 14-item questionnaire containing three clusters of inter-correlating items concerned with amount of contact, degree of closeness and intimacy, and perceived similarity. MZ pairs were in closer contact and saw themselves as more similar than either group of DZ twins. MZs also described themselves as closer to one another than did the DZ pairs; this difference was small but consistent, appearing for both sexes on all five questions in the Closeness cluster. Over the entire group, Closeness correlated about 0.40 with Contact and about 0.55 with perceived Similarity.

Rose et al [34] have argued that the greater similarity in personality of adult MZ than DZ twins may be a consequence of this greater frequency and intimacy of contact. The present findings corroborate once again the greater closeness among MZ twins and the strong correlation between closeness and similarity. However, we continue to believe that similarity leads to intimacy, rather than the other way about, in part because the obverse hypothesis leaves intimacy unexplained while similarity can be explained genetically.

					Males				Females
Index category			Index		MZ	DZ	OS		MZ	DZ	OS
Twins live together			6		5%	2%	1%		3%	1%	1%
Speak together
	Daily		5		18	6	3		26	14	2
	Weekly		4		34	26	19		40	30	17
	Monthly		3		26	30	32		22	32	34
	On holidays		2		7	17	25		3	9	25
	Seldom		1		10	20	21		4	14	21
Mean index value					3.6	2.9	2.6		3.9	3.2	2.6