Stability of genetic influence on morningness–eveningness: a
cross-sectional examination of South Korean twins from
preadolescence to young adulthood
Medical Research Center, Seoul National University, Seoul, South Korea
Accepted in revised form 3 October 2006; received 18 February 2006
A cross-sectional twin design was used to study the developmental nature of genetic and
environmental influences on morningness–eveningness (M–E). A total of 977 South
Korean twin pairs aged 9–23 years completed 13 items of a Korean version of the
Composite Scale through the telephone interview. The total sample was split into three
age groups: preadolescents, adolescents, and young adults. Twin correlations did not
vary significantly with age, suggesting that genetic influences on M–E are stable
throughout the developmental span. Results of model-fitting analyses indicated that
genetic and environmental factors explained, respectively, 45% and 55% of the
variance in all three age groups. Environmental factors were primarily those factors
that twins did not share as a consequence of their common rearing; family
environmental factors in M–E were consistently near zero in all three age groups.
The present study is the first to demonstrate genetic influences on M–E in preadolescent
children as young as 9 years old. In spite of differences in culture and frequencies of
genes between South Koreans and Caucasians, genetic and environmental influences on
M–E found in the present sample were remarkably similar to those reported by
previous studies on the basis of late adolescent and adult Caucasian twins.
circadian, development, environments, genetics, morningness-evening-
Circadian rhythms are biological rhythms that display 24 h
cyclic patterns of behavior. One of the most important aspects
of human individual differences in circadian rhythms has been
identified as a degree of ?morningness? and ?eveningness?
(Kerkhof, 1985; Merrow et al., 2005). Morningness–evening-
ness (M–E) is mostly reflected in the preference in sleep–wake
timing. Morning-type individuals get up easily and are more
alert in the morning than in the evening, have a hard time
sleeping late, fall asleep quickly in the evening, and prefer
daytime activities. Evening-type individuals are more alert at
night, able to sleep late in the morning, take a long time to fall
asleep at night, and prefer nighttime activities. It has been
reported that there is a significant relationship between M–E
and mental illness in particular, major depression, with
depressed patients having higher eveningness than normal
controls (Drennan et al., 1991). M–E also has been associated
with sleep disorders: morningness was related to difficulty in
maintaining sleep and the impossibility to return to sleep in the
early morning (sleep phase-advance syndrome); and evening-
ness was related to difficulty in initiating sleep and morning
sleepiness (Taillard et al., 2001).
Previous studies have shown that with advancing age, the
circadian rhythms of many variables undergo changes. Typ-
ically, preadolescent children tend to be morning-oriented. But
as children go through the transition from childhood to
adolescence, they become evening-oriented (Carskadon et al.,
1993; Gau and Soong, 2003; Kerkhof, 1985; Park et al., 2002;
Roenneberg et al., 2004). In the Shinkoda et al. (2000) study,
the authors administered an M–E questionnaire to over 500
Japanese students aged 6–18 years, and found that scores
significantly changed toward a preference for ?eveningness?
Correspondence: Yoon-Mi Hur, Medical Research Center #110, Seoul
National University, College of Medicine, 28 Yongon-dong, Chongno-
gu, Seoul 110-799, South Korea. Tel.: 82-2-741-6179; fax: 82-2-741-
6190; e-mail: firstname.lastname@example.org
J. Sleep Res. (2007) 16, 17–23
? 2007 European Sleep Research Society
over advancing grades; the change occurred most notably
around the seventh grade.
It has long been recognized that circadian rhythms are
largely determined by genetic factors. The twin method is
particularly useful for investigating genetic and environmental
influences on a trait variation within a population. The twin
method decomposes the total variance of a trait into variance
components attributable to genetic and environmental factors.
Genetic variance is further divided into additive and non-
additive genetic variance components, whereas environmental
variance is further decomposed into shared and non-shared
environmental variance components. Additive genetic variance
refers to genetic effects that simply add up across genes. Non-
additive genetic variance involves the effects of interaction
among alleles at a single locus as well as interactions among
alleles at different loci. Shared environmental factors represent
the environmental factors that are shared by two members of a
twin pair. Examples of shared environmental factors include
parental socioeconomic status, parental childrearing styles and
practices, and effects of schools that two members of a twin
pair attend together. Non-shared environmental factors rep-
resent those environmental factors that are not shared by two
members of a twin pair. Examples of non-shared environmen-
tal factors include accidents and peers that the two members of
a twin pair do not share (Plomin et al., 1990).
As the first twin study of M–E, Hur et al. (1998) analyzed
310 pairs of adult reared-together and reared-apart MZ and
DZ twins who completed an M–E questionnaire. Hur et al.
(1998) found that genetic factors explained approximately
54% of the total variance of M–E; age accounted for 3% of the
total variance; and the remaining variance was attributable to
non-shared environmental influences and measurement error.
Shared environmental effects on M–E were not significant in
the Hur et al. (1998) study.
Vink et al. (2001) analyzed 1650 pairs of late adolescent
twins (mean age ¼ 17.8 years) and their parents (mean
ages ¼ 48.0 years for fathers and 46.0 years for mothers)
and 124 pairs of adult twins (mean age ¼ 46.5 years) who
responded to an M–E question with five answer categories. In
the Vink et al. (2001) study, genetic influences on M–E were
44% for the younger generation and 48% for the older
generation. They also demonstrated that non-additive genetic
influences were significant (15% for the younger and 16% for
the older generation). In the Vink et al. (2001) study the
magnitudes of genetic and environmental variances in M–E
were not different between males and females. The Hur et al.
(1998) and Vink et al. (2001) studies agree with each other in
that shared environmental influences were negligible in indi-
vidual differences in M–E among late adolescents and adults,
while genetic influences on M–E were substantial and com-
parable in magnitude between late adolescents and adults.
Whereas the Hur et al. and Vink et al. studies assessed
genetic and environmental contributions to M–E among
Caucasian twins who live in free societies, Klei et al. (2005)
investigated genetic and environmental factors in M–E among
the Hutterites, an endogamous, religious group whose mem-
bers live under relatively fixed schedule in agrarian, self-
supportive communities. In spite of the social restriction
impinged upon the Hutterites, additive genetic influence on
M–E among the Hutterites was found to be significant (23%)
and only slightly lower than those estimated in the Dutch twin
study by Vink et al. (2001). The results of the Klei et al. study
suggest that environmental constraints may not exert substan-
tial influences on individual differences in M–E.
The present study examined genetic and environmental
influences on M–E among South Korean children and
adolescents who also live in relatively restrictive environments.
Compared with most Western countries, South Korea main-
tains a much stronger relationship between early academic
achievements and later educational opportunities and social
standing. Also, for the strong tradition of Confucianism in the
society, excellent early academic achievements and subsequent
performance on the college entrance examination taken at the
end of high school are considered family honor as well as
personal accomplishments in South Korea. For these reasons,
students in South Korea are expected to devote their time
primarily to their studies. It is very common for most South
Korean junior high school, high school, and even some of the
elementary school students to stay at private educational
institutions or have tutoring until very late at night for extra
study. This extremely competitive atmosphere and social
pressure significantly limit time available for sleep among
students and considerably affect the sleep/wake patterns of
students in South Korea. Investigators have reported that
South Korean students suffer from severe daytime sleepiness,
sleep/wake problem behavior, and depressed-mood (Yang
et al., 2005).
The present study aimed to investigate whether genetic
influence on M–E among South Korean twins is similar to
those found in Western samples. Especially, by studying
preadolescent, adolescent, and young adult twins cross-sec-
tionally, the present study attempted to determine whether
heritability estimates of M–E among South Koreans increase,
decrease, or remain stable from preadolescence to young
adulthood. To my knowledge, the present study is the first to
include the preadolescent period of the lifespan to identify
genetic influences on M–E.
MATERIALS AND METHODS
The sample consisted of 977 twin pairs who responded to a
telephone survey conducted by ongoing South Korean Twin
Registry (SKTR) in the years 2004 and 2005. The SKTR is a
nationwide volunteer registry of South Korean twins and their
families. A detailed description of the recruitment procedure of
the SKTR was reported by Hur et al. (in press). The telephone
surveys in the years 2004 and 2005 were conducted mainly
(>98%) for the twins who resided in Seoul, South Korea.
Twins? zygosity in the SKTR was determined from the twins?
parents? responses to a zygosity questionnaire that includes
18 Y.-M. Hur
? 2007 European Sleep Research Society, J. Sleep Res., 16, 17–23
questions regarding physical similarities and frequency of
confusion of the twins by family members and others. Twenty-
two pairs were excluded from data analyses because their
zygosity was ambiguous.
Table 1 provides a description of the sample. The total
sample ranged from 9 to 23 years of age and was split into
three age groups: preadolescents (9–12 years; 132 pairs of MZ
and 121 pairs of DZ twins), adolescents (13–18 years; 340 pairs
of MZ and 177 pairs of DZ twins), and young adults (19–
23 years; 194 pairs of MZ and 58 pairs of DZ twins). The
preadolescent group in the present study consisted of element-
ary school students (grades 4–6), the adolescent group, junior
high and high school students (grades 7–12), and the young
adult group, graduates of high schools.
The number of MZ twins was not very different from that of
DZ twins in the preadolescent sample; however, the numbers
of MZ twins exceeded the numbers of DZ twins in the
adolescent and young adult samples. These cohort differences
in the rates of MZ and DZ twins largely reflect a change of the
birth rate of DZ twin pairs in South Korean population over
the past 20 years (Hur and Kwon, 2005) and do not necessarily
represent a serious ascertainment bias. In the preadolescent
and adolescent samples, less than 55% of the sample was
female, suggesting that the samples consist of roughly equal
numbers of males and females. In the young adult sample,
however, 63% of the sample was female. It appears that an
overrepresentation of females in the young adult sample is
partly because some of the male twins were in the military
service at the time of telephone interview because young adult
males in South Korea have the obligation for the army service.
As part of the regular SKTR telephone interview, twins were
asked to respond to a Korean version of the Composite Scale
(CS) developed by Smith et al. (1989). The CS consists of 13
items regarding preferred rising and bed times, preferred times
of physical and mental performance, and subjective alertness
after rising and before going to bed, and subjective evaluation
of morningness and eveningness. The items of the CS were
adapted from M–E Questionnaire (Horne and Ostberg, 1976)
and Diurnal Type Scale (Torsvall and Akerstedt, 1980). The
CS yields scores on a single scale of morningness versus
eveningness. Higher scores indicate greater ?morningness?,
whereas lower scores represent greater ?eveningness?.
The Korean version of the CS has been extensively studied
and shown to be reasonably reliable and valid (Kook et al.,
1999; Yoon et al., 1997). In the present study, Cronbach a
reliabilities of the CS were 0.65 in preadolescents, 0.70 in
adolescents, and 0.72 in young adults.
To estimate genetic and environmental influences on M–E in
preadolescents, adolescents, and young adults, intraclass
correlations were computed for MZ and DZ twins in each
age group and biometrical model-fitting analyses were con-
ducted. Because previous studies (e.g., Vink et al., 2001) have
shown no gender differences for the magnitude of genetic and
environmental factors in M–E and because the sample size in
the present study is relatively small to perform analyses to
detect gender differences in genetic and environmental influ-
ences on M–E, correlational analyses and model-fitting ana-
lyses were carried out on the basis of the combined sample of
males and females. Prior to correlation and model-fitting
analyses, however, the scores of the CS were corrected for
gender using a regression procedure (McGue and Bouchard,
1984) to avoid possible bias arising from gender effects.
The twin intraclass correlation was calculated using the
formula r ¼ (MSB)MSW)/(MSB + MSW), where MSB and
MSW are, respectively, the mean squares between and within
pairs estimated in a one-way analysis of variance (anova). The
hypotheses that the twin correlations were higher for MZ than
for DZ twins and equal across the three age groups (pre-
adolescents, adolescents, and young adults) were tested on the
basis of the Fisher z-transformation method that produces a
heterogeneity chi-square test statistic (Donner and Rosner,
MZ twins share identical genes, whereas DZ twins share on
average 50% of their segregating genes. For this reason,
additive genetic effects are indicated if the correlation for MZ
pairs is greater than DZ pairs; and the importance of shared
environmental effects is indicated if the correlation for DZ
pairs is greater than half the correlation for MZ pairs. Because
non-additive genetic effects involve allelic interactions, the DZ
twin correlation would be less than half the MZ twin
correlation if non-additive genetic effects are important for a
trait. Non-shared environmental effects are manifested as
within MZ pair differences (i.e., 1)rMZ), because the effects
always operate to make MZ twins dissimilar.
Using the software package Mx (Neale, 1999), additive genetic
(A), shared environmental (C), non-additive genetic (D), and
non-shared environmental variance and measurement error
Table 1 Sample divided by age and zygosity
Characteristics Preadolescent Adolescent Young adult Total
No. of MZ pairs 132
No. of DZ pairs
Numbers of opposite-sex DZ twin pairs are in parenthesis. MZ,
monozygotic twins; DZ, dizygotic twins.
Stability of genetic influence on morningness–eveningness 19
? 2007 European Sleep Research Society, J. Sleep Res., 16, 17–23
(E) parameters were estimated by the maximum likelihood
method. Fig. 1 illustrates a univariate behavioral genetic
model that includes A, C, D, and E parameters. On the basis
of the quantitative genetic theory, the A factors correlate 1.0
and 0.5 for MZ and DZ twins, respectively, whereas the D
factors correlate 1.0 and 0.25 for the corresponding twins.
Because all twins were reared in the same family, the
correlation for the C factors is 1.0 for both MZ and DZ
twins. Twins are not correlated for the E factors for non-
shared environmental factors are unique to each member of a
twin pair. Because all four parameters (A, C, D, and E) cannot
be estimated in the same model, the ACE model and the ADE
model were tested separately (Neale and Cardon, 1992).
Models were fit to the raw data.
Two steps were taken to complete the biometrical model-
fitting analyses. Initially, models were fit to the twin data of all
three age groups separately in order to find the best-fitting,
most parsimonious model within each age group. Five
different models (ADE, ACE, AE, CE, and E) were tested
against a null model within each age group.1The null model
was a baseline model where variances of MZ and DZ twins
were allowed to differ, while variances of the first and the
second twins were set to be equal. This model can serve as a
null model because in all of the five models (ADE, ACE, AE,
CE, and E), variances of MZ and DZ twins as well as those of
the first and the second twins within each zygosity group were
constrained to be equal.
After the best-fitting model within each age group was
selected, the model was used as the basis for the construction
of the full and constrained models for the cross-sectional
analyses. In the full model, the parameters were allowed to
differ across age groups, whereas in the constrained model, the
parameters were set to be equal. Because the full model
implicates that genetic and environmental variances vary with
age, whereas the constrained model indicates stability with age,
the comparison of the fit of the full and constrained models
will tell us whether genetic and environmental influences on
M–E are stable or change from preadolescence to young
Two criteria were used in deciding on the best-fitting model:
the chi-square difference test and the Akaike information
criterion (AIC ¼ v2) 2df). When the raw data are used in
model-fitting analyses, Mx computes minus twice the log-
likelihood of the data ()2LL), with an arbitrary constant that
is a function of the data. If two models are nested, differences
in )2LL between nested models are distributed as a chi-square,
with degrees of freedom as follows: dfk+1) dfk, where k is
number of degrees of freedom (Bollen, 1989). A significant
increase in chi-square in the constrained model when com-
pared with the full model would suggest that the constrained
model fit the data less well than the full model. A non-
significant change in chi-square would indicate that the full
model and the constrained model are equally acceptable.
When competing models were equally acceptable on the
basis of the chi-square difference test, a model that yielded the
lowest AIC was chosen as the best-fitting model. AIC
quantifies the information content of a model in terms of the
joint criterion of fit and parsimony. Because the model-fitting
that results in the most information is that for which AIC is
minimum (Akaike, 1987), the model that produces the lowest
AIC was considered the best-fitting model.
The scores of the CS showed normal distribution (skew-
ness ¼ 0.09; curtosis ¼ 0.02) in the present sample. Analyses
of variance were conducted to test effects of age, gender,
zygosity, and their interactions. Prior to anovas, the total
sample was split into two groups, each containing one twin
from every pair, to correct for non-independence of observa-
tions in twin pairs. In both twin groups, only age effects
attained statistical significance at P value of less than 0.001.
The mean level of the CS progressively declined from
preadolescents to young adults, suggesting that twins become
more evening-oriented with age. These age effects were
consistent with the results from previous studies of M–E
mentioned earlier (Gau and Soong, 2003; Kerkhof, 1985; Park
et al., 2002; Roenneberg et al., 2004; Shinkoda et al., 2000).
Twin correlations across age and zygosity
Table 2 presents MZ and DZ twin intraclass correlations for
the CS separately by age group as well as pooled over age
r =1.0 for MZ, r =0.5 for DZ
r =1.0 for MZ, r =0.25 for DZ
r =1.0 for MZ & DZ
Twin 1Twin 2
Figure 1. A univariate behavioral genetic model. Twin 1 and Twin 2
represent scores of the CS for the first and the second twin, respect-
ively. A, D, C, and E are latent variables representing, respectively,
additive genetic, non-additive genetic, shared environmental, and non-
shared environmental variance. Non-shared environmental variance
includes measurement error. The curved, two-headed arrows indicate
correlations between the variables they connect, and the one-headed
arrows represent path, standardized partial regression of the measured
variable on the latent variable.
1The DE model was not tested because it has been argued that
dominance effects alone are not enough to explain the very low DZ
correlation when compared with MZ correlation and that overdom-
inance occurs only in extreme gene frequencies (Eaves, 1988).
20 Y.-M. Hur
? 2007 European Sleep Research Society, J. Sleep Res., 16, 17–23
groups. Also reported in the table are the results of the
statistical comparisons of the correlation across age groups
and between the two zygosity groups. In none of the age-group
comparisons did the chi-square test statistics achieve signifi-
cance; that is, there is no evidence that twin similarity for the
CS varies with age. In contrast, the MZ correlation exceeded
the DZ correlation in all four groups (i.e., preadolescents,
adolescents, young adults, and pooled group) with three of the
four comparisons being statistically significant. MZ correla-
tions higher than twice the DZ correlations suggested the
existence of non-additive genetic influence on individual
difference in the CS. Taken together, analyses of twin
correlations showed that genetic influences on the variation
of the CS tend to be constant among all three age groups.
Table 3 summarizes the results of the model-fitting analyses
within each age group. The first model is the null model where
variances were allowed to differ between MZ and DZ twins.
When variances were restricted to be equal between MZ and
DZ twins, non-significant changes in v2occurred in the ADE,
ACE, and AE models in preadolescents and adolescents,
indicating that these three models are better than the null
model. The fact that the CE and E models were rejected in
preadolescents and adolescents on the basis of the chi-square
difference test suggests that environmental effects alone cannot
explain individual differences in M–E in preadolescents and
adolescents. Among the ADE, ACE, and AE models, AIC was
minimized in the ADE model in preadolescents, and in the AE
model, in adolescents. These results indicate that the best-
fitting model is the ADE model for preadolescents, and the AE
model for adolescents.
In young adults, non-significant changes in v2occurred in
the CE model in addition to the ADE, ACE, and AE models
when variances were fixed to be equal between MZ and DZ
twins. It appears that due to the small sample size in young
adults, the superiority in fit between the AE and CE models
could not be clearly determined by the chi-square difference
test alone. However, in the young adult group, the lowest
AIC was found in the AE model, suggesting that the AE
model is better than any other competing models for young
On the basis of the conclusions drawn from the model-fitting
analyses within each age group, the full cross-sectional model
was constructed with the A, D, and E parameters. Table 4
provides the results of fitting the full model and its submodels
to the cross-sectional data, and genetic and environmental
estimates and their 95% confidence intervals for various
models. The full model (Model 1) included the A, D, and E
parameters for preadolescents, with A and E parameters
varying for adolescents and young adults. Next model (Model
2) restricted the A and E parameters to be the same for
adolescents and young adults, while maintaining the A, D, and
E parameters for preadolescents. This procedure yielded a
non-significant change in v2, suggesting that the magnitudes of
A and E effects are constant across adolescents and young
adults. Next, the D parameter was eliminated from Model 2
and the A and E parameters were constrained to be equal
across the three age groups (Model 3). Again, the change in v2
was not significant. Model 3 also had the lowest AIC value of
all models tested. Thus, Model 3 where additive genetic and
non-shared environmental variances were constrained to be
equal from preadolescence to young adulthood was chosen as
the best, most parsimonious cross-sectional model for M–E. In
Model 3, the additive genetic and non-shared environmental
influences were 45% (95% CI: 39–50%) and 55% (95% CI:
50–61%), respectively. The results of model-fitting analyses
were in line with the conclusions drawn from an examination
of the twin correlations.
Table 2 Intraclass correlations of the sex-corrected scores of the
Composite Scale for MZ and DZ twins in three age groups
Age group correlation
PreadolescentAdolescent Young adult
**P < 0.01. MZ, monozygotic twins; DZ, dizygotic twins.
?Age group heterogeneity test (df ¼ 2).
?Zygosity group heterogeneity test (df ¼ 1).
Table 3 Results of model-fitting analyses within each age group
PreadolescentsAdolescents Young adults
*P < 0.05; **P < 0.01. AIC ¼ v2)2(Ddf). A ¼ additive genetic effects, D ¼ non-additive genetic effects, C ¼ shared environmental effects,
E ¼ non-shared environmental effects and measurement error. The best-fitting, most parsimonious model in each age group is indicated in
Stability of genetic influence on morningness–eveningness 21
? 2007 European Sleep Research Society, J. Sleep Res., 16, 17–23
The present study is the first examination of the developmental
nature of genetic and environmental effects on M–E on the
basis of South Korean twin sample. The results of the present
study demonstrate that genetic influences on M–E emerged
from preadolescent children as young as 9 years old and
persisted in adolescents and young adults. The estimates of
genetic and non-shared environmental influences on M–E in
preadolescents, adolescents, and young adults in the present
study were similar to those found in late adolescent and adult
Caucasian twins. Broad sense-heritability that includes both
additive and non-additive genetic effects was 45% and the
remaining variance of M–E was largely explained by non-
shared environmental effects (55%).
There was some indication of non-additive genetic influence
on M–E, especially in the preadolescent group. However, one
should note that although the non-additive genetic estimates in
the preadolescent group in Models 1 and 2 in Table 4 were
large (46%), they did not attain statistical significance. Given
that a very large sample is required to reliably detect non-
additive genetic influences (Martin et al., 1978), this failure in
attaining statistical significance seems due to the relatively
small sample size of the present study.
In an attempt to resolve the magnitude of additive and non-
additive genetic influences on M–E across the three age
groups, we fit ADE model where A, D, and E parameters were
restricted to be the same across the three age groups. This
model showed additive and non-additive genetic and non-
shared environmental effects to be 18%, 27%, and 55%,
respectively, suggesting that genetic influences estimated in the
best-fitting model (Model 3 in Table 4) might consist of 18%
of additive and 27% of non-additive effects. Consistent with
previous studies, the present sample also showed negligible
influences of rearing environmental factors on individual
difference in M–E.
It was interesting to note that in spite of differences in
culture and frequencies of genes between Caucasians and East
Asians, heritability estimates of M–E were similar in the two
groups. Cross-sectional twin studies have documented that for
many human traits, genetic factors tend to increase with age,
whereas the effects of common rearing tend to diminish
throughout the developmental period because as children
become old, they tend to become progressively independent
from their parents and select activities on the basis of their own
genetically influenced preferences (McCartney et al., 1990;
McGue and Bouchard, 1998). It appears that M–E does not
follow this developmental pattern. It is very likely that the
levels of environmental constraints imposed on sleep/wake
time and patterns are different among preadolescents, adoles-
cents, and young adults. Lifestyles are also likely to vary
among these three age groups. However, heritability estimates
of M–E were constant across age groups, and shared
environmental factors were consistently near zero in all three
age groups. These results indicate that even before adolescence
genetic factors powerfully suppress the effect of rearing
environments on the variation of M–E.
In support of twin studies of M–E, recent molecular studies
have found genes associated with M–E. For example, the 3111
C allele in the 3¢-flanking region of the CLOCK gene has been
shown to be associated with increased eveningness tendencies
and delayed sleep timing in Caucasian sample (Katzenberg
et al., 1998) and also in Japanese sample (Mishima et al.,
2005). It was remarkable that genetic factors in M–E were
detected in preadolescent children as young as 9 years old in
the present study. This finding suggests that M–E is a genetic
disposition that crystallizes relatively early in life. To date, the
majority of molecular studies that aimed at identification of
the genes involved in M–E employed adult samples. The
present findings suggest that the preadolescent sample can be
used as well to discover genes involved in M–E.
There are limitations of the present study that need to be
addressed. First, the present study is based on cross-sectional
twin data. Therefore, it is important that the results of the
present study should be replicated by research using longitud-
inal twin designs. Secondly, one should note that the finding
that heredity makes a large contribution to the variance of M–
E in various age groups does not necessarily mean that
Table 4 Results of model-fitting analyses for the best-fitting, most parsimonious cross-sectional model for the Composite Scale for three age groups
Parameter estimates (%)
Goodness of fit indicesPreadolescentsAdolescents Young adults
Ddf AICADEA D EA D E
1. ADE for Pre.,
and AE vary
for Adol. & YA.
2. ADE for Pre.
AE same for
Adol. & YA.
3. AE same across
three age groups
5385.4 1944 0 (0–51) 46 (0–63) 54 (43–68) 44 (34–55) –56 (49–65) 47 (33–64) –53 (42–65)
5385.7 1946 0.3 2
)3.70 (0–49) 46 (0–58) 54 (42–68) 45 (38–52) –55 (48–62) 45 (38–52) –55 (48–62)
5388.7 1949 3.3 5
)6.7 45 (39–50)– 55 (50–61) 45 (39–50) –55 (50–61) 45 (39–50) –55 (50–61)
AIC ¼ v2)2(Ddf). A ¼ additive genetic effects, D ¼ non-additive genetic effects, E ¼ non-shared environmental effects and measurement error.
95% CI are in parenthesis. The best-fitting, most parsimonious model in each age group is indicated in boldface. Pre, preadolescents; Adol.,
adolescents; YA, young adults.
? 2007 European Sleep Research Society, J. Sleep Res., 16, 17–23
observed differences in the mean level of M–E among different
age groups are due to genetic factors. It would be interesting in
the future research to investigate genetic and environmental
origins of the developmental change in the mean level of M–E.
Appropriate methods to study this research question have been
suggested (Dolan et al., 1989, 1994). Third, the sample size of
the present study was too small to reliably discriminate
between additive and non-additive genetic effects. To reach a
firm conclusion about additive and non-additive genetic
components in M–E from preadolescence to young adulthood,
the present study needs to be replicated with a larger sample,
and in the meantime, one should interpret the genetic estimate
(45%) found in the present study as a sum of both additive and
non-additive genetic effects. Finally, the present findings may
not generalize to the South Korean population at large. The
conclusions drawn from this study are necessarily limited to
preadolescents through young adults in Seoul, South Korea
because majorities of the twins in the present sample are
residents of Seoul.
This study was supported in part by funding from the Institute
of Human Behavioral Medicine, Medical Research Center,
Seoul National University in 2004. I would like to thank the
twins who participated in the study.
Akaike, H. Factor analysis and AIC. Psychometrica, 1987, 52: 317–
Bollen, K. A. Structural Equations with Latent Variables. John Wiley &
Sons, New York, 1989.
Carskadon, M. A., Vieira, C. and Acebo, C. Association between
puberty and delayed phase preference. Sleep, 1993, 16: 258–262.
Dolan, C. V., Molenaar, P. C. M. and Boomsma, D. I. LISREL
analysis of twin data with structured means. Behavior Genetics, 1989,
Dolan, C. V., Molenaar, P. C. M. and Boomsma, D. I. Simultaneous
genetic analysis of means and covariance structure: Pearson–Lawley
selection rules. Behavior Genetics, 1994, 24: 17–24.
Donner, A. and Rosner, B. On inferences concerning a common
correlation coefficient. Applied Statistics, 1980, 29: 69–76.
effects of depression and age on the Horne–Ostberg morningness–
eveningness score. Journal of Affective Disorder, 1991, 23: 93–98.
Eaves, L. J. Dominance alone is not enough. Behavior Genetics, 1988,
Gau, S. F. and Soong, W. T. The transition of sleep–wake patterns in
early adolescence. Sleep, 2003, 26: 449–454.
Horne, J. and Ostberg, O. A self-assessment questionnaire to deter-
mine morningness–eveningness in human circadian rhythms. Inter-
national Journal of Chronobiology, 1976, 4: 97–110.
Hur, Y.-M. and Kwon, J. S. Changes in twinning rates in South
Korea: 1981–2002. Twin Research and Human Genetics, 2005, 8: 76–
Hur, Y.-M., Bouchard, T. J., Jr and Lykken, D. T. Genetic and
environmental influence on morningness–eveningness. Personality
and Individual Differences, 1998, 25: 917–925.
Hur, Y.-M. Shin, J. S., Jeong, H.-U. and Han, J. Y. The South Korean
Twin Registry. Twin Research & Human Genetics, (in press).
Katzenberg, D., Young, T., Finn, L., Lin, L., King, D. P., Takahashi,
J. S. and Mignot, E. A CLOCK polymorphism associated with
human diurnal preference. Sleep, 1998, 21: 569–576.
Kerkhof, G. A. Inter-individual differences in the human circadian
system: a review. Biological Psychology, 1985, 20: 83–112.
Klei, L., Reitz, P., Miller, M., Wood, J., Maendel, S., Gross, D.,
Waldner, T., Eaton, J., Monk, T. H. and Nimgaonkar, V. L.
Heritability of morningness–eveningness and self-report sleep meas-
ures in a family-based sample of 521 hutterites. Chronobiology
International, 2005, 22: 1041–1054.
Kook, S.-H., Yoon, J.-S. and Lee, H.-Y. Cross validation of
the Korean translation of Composite Scale (KtCS) to measure
Association, 1999, 38: 297–305.
Martin, N. G., Eaves, L. J., Kearsey, M. J. and Davies, P. The power
of the classical twin study. Heredity, 1978, 40: 97–116.
McCartney, K., Harris, M. J. and Bernieri, F. Growing up and
growing apart: a developmental meta-analysis of twin studies.
Psychological Bulletin, 1990, 107: 226–237.
McGue, M. and Bouchard, T. J., Jr. Adjustment of twin data for the
effects of age and sex. Behavior Genetics, 1984, 14: 325–343.
McGue, M. and Bouchard, T. J., Jr. Genetic and environmental
influences on human behavioral differences. Annual Review of
Neuroscience, 1998, 21: 1–24.
Merrow, M., Spoelstra, K. and Roenneberg, T. The circadian cycle:
daily rhythms from behaviour to genes. EMBO Report, 2005, 6:
Mishima, K., Tozawa, T., Satoh, K., Saitoh, H. and Mishima, Y. The
3111T/C polymorphism of hClock is associated with evening
preference and delayed sleep timing in a Japanese population
sample. American Journal of Medical Genetics, Part B. Neuro-
psychiatric Genetics, 2005, 133: 101–104.
Neale, M. C. Mx: Statistical Modeling, 5th edn. Department of
Psychiatry, Medical College of Virginia, Richmond, VA, USA, 1999.
Neale, M. C. and Cardon, L. R. Methodology for Genetic Studies of
Twins and Families. Kluwer, London, 1992.
Park, Y. M. Matsumoto, K., Seo, Y. J., Kang, M. J. and Nagashima,
H. Changes of sleep or waking habits by age and sex in Japanese.
Perceptual and Motor Skills, 2002, 94: 1199–1213.
Plomin, R., DeFries, J. C. and McClearn, G. E. Behavior Genetics: A
Primer. Freeman, New York, 1990.
Roenneberg, T., Kuehnle, T., Pramstaller, P. P., Ricken, J., Havel, M.,
Guth, A. and Merrow, M. A marker for the end of adolescence.
Current Biology, 2004, 14: R1038–R1039.
Shinkoda, H., Matsumoto, K., Park, Y. M. and Nagashima, H. Sleep–
wake habits of schoolchildren according to grade. Psychiatry and
Clinical Neurosciences, 2000, 54: 287–289.
Smith, C. S., Reilly, C. and Midkiff, K. Evaluation of three circadian
rhythm questionnaires with suggestions for an improved measure of
morningness. Journal of Applied Psychology, 1989, 74: 728–738.
Taillard, J., Philip, P., Chastang, J. F., Diefenbach, K. and Bioulac, B.
Is self-reported morbidity related to the circadian clock? Journal of
Biological Rhythms, 2001, 16: 183–190.
Torsvall, L. and Akerstedt, T. A diurnal type scale. Construction,
consistency and validation in shift work. Scandinavian Journal of
Work, Environment, & Health, 1980, 6: 283–290.
Vink, J. M., Groot, A. S., Kerkhof, G. A. and Boomsma, D. I. Genetic
analysis of morningness and eveningness. Chronobiology Inter-
national, 2001, 18: 809–822.
Yang, C. K., Kim, J. K., Patel, S. R. and Lee, J. H. Age-related
changes in sleep/wake patterns among Korean teenagers. Pediatrics,
2005, 115(Suppl. 1): 250–256.
Yoon, J.-S., Shin, S.-M., Kook, S.-H. and Lee, H.-Y. A preliminary
study on the Korean translation of Composite Scale (KtCS) to
measure morningness–eveningness. Journal of Korean Neuropsychi-
atric Association, 1997, 36: 122–131.
of Korean Neuropsychiatry
Stability of genetic influence on morningness–eveningness23
? 2007 European Sleep Research Society, J. Sleep Res., 16, 17–23