Part I

Sleep Timing

13
Chapter 2

Comparison of the Munich Chronotype

Questionnaire with the Horne-Östberg’s
Morningness-Eveningness Score

Andrei Zavada, Marijke C. M. Gordijn, Domien G. M. Beersma,

Serge Daan and Till Roenneberg
Chronobiol. Int. 2005, 22, 267-78.

15
16 CHAPTER 2. A SURVEY OF CHRONOTYPES

Abstract
We report on results from an Internet survey of sleeping habits
in a Dutch population using the Munich Chronotype Question-
naire (MCTQ), supplemented with the Horne–Östberg Morningness-
Eveningness Questionnaire (MEQ). The MCTQ was filled out by 5055
respondents, of which 2481 also completed the MEQ. MEQ score cor-
related well with the MCTQ assessment of time of mid-sleep on free
days (MSF; r = −0.73) and on workdays (MSW; r = −0.61). MEQ
was more strongly correlated with MSF (50 % of sleep time) than with
sleep onset (0 %), rise time (100 %) or with any other percentile (10–
40, 60–90 %) of sleep on free days. The study shows that chrono-
type (based on MSF as measured by the MCTQ) strongly correlates
with morningness-eveningness (as measured by the MEQ). However,
the MCTQ collects additional detailed information on sleep-wake be-
haviour under natural conditions.
2.1. INTRODUCTION 17

2.1 Introduction
The sleep-wake cycle is the most prominent daily pattern in human
behaviour. Activity is generally confined to the natural day and early
evening while sleep occupies the remaining night hours. There is con-
siderable inter-individual variation in preferred timing of the sleep-
wake rhythm, with extreme morning and evening types often called
‘larks’ and ‘owls’. The timing of sleep is largely under control of
the central circadian pacemaker, located in the suprachiasmatic nu-
clei (SCN) of the hypothalamus (Daan et al., 1984). The other major
regulatory component to the steady sleep-wake maintenance is the
homeostatic component (Daan et al., 1984; Dijk and Czeisler, 1995).
Both a stable relationship between endogenous (or internal) time and
external time as well as a good and well-timed sleep are believed to
be essential for human health (Roenneberg et al., 2004b; Dijk and
Lockley, 2002; Rajaratnam and Arendt, 2001). Systematic investiga-
tions of individual circadian phase preferences have been stimulated by
the publication of the Morningness-Eveningness Questionnaire (MEQ)
by Horne and Östberg (1976). The MEQ score was correlated with
core parameters of human circadian organisation, such as sleep tim-
ing (Carrier et al., 1997; Andrade et al., 1992; Laberge et al., 2000)
and endogenous period (Duffy et al., 2001). The use of the MEQ
in inter-individual experiments has enabled consistent segregation of
putative ‘larks’ from ‘owls’ (Andrade et al., 1992; Duffy et al., 1999;
Baehr et al., 2000; Bailey and Heitkemper, 2001). The administration
of MEQ in large-scale epidemiological surveys allowed investigators
to probe circadian preference profiles of a variety of subclasses in a
population (such as age groups: Carrier et al. (1997); Laberge et al.
(2000); Giannotti et al. (2002); Ishihara et al. (1992); gender: Adan
and Natale (2002); social/professional groups: Park et al. (1998a,b)).
The MEQ contains 19 questions, aimed to determine where on a daily
temporal scale the respondent’s maximum propensity to be active lies.
Most questions are preferential, in the sense that the respondent is
asked to indicate when, for example, he/she would prefer to wake
up or start sleep, rather than when he/she actually does. Questions
are multiple-choice, each answer is assigned a value. Their sum gives
a score, ranging from 16 to 86, with lower values corresponding to
evening types.
A new questionnaire has recently been designed to collect informa-
tion about the actual timing of daily sleep (and activity): the Munich
18 CHAPTER 2. A SURVEY OF CHRONOTYPES

ChronoType Questionnaire (MCTQ) (Roenneberg et al., 2003). It

essentially is a tool to collect primary sleep times, such as bed- and
rise-times, clock times of becoming fully awake, sleep latency and in-
ertia, in addition to other time points (e.g., siesta). The MCTQ also
asks the respondents to rate themselves as one of the seven chronotypes
(Extreme Early, Moderate Early, Slightly Early, Normal, Slightly Late,
Moderate Late, Extreme Late). This particular question thus contains
a relative judgment, i.e., a judgment on the subject’s own behaviour
in comparison to others. Subjects are also asked to judge their chrono-
type at different life stages, such as childhood, adolescence, etc., which
allows an intra-individual assessments of how chronotype changes with
age.
The MCTQ asks that all information should be specified as pre-
cisely as possible as they are in the respondent’s present circumstances,
i.e., regardless of what the respondents may consider preferential for
themselves. This enables the investigators to draw up a survey of
the sleeping habits in a given community, to follow the dynamics of
sleep parameters over the seasons or different ages (Roenneberg et al.,
2004b), and to conclude eventually on possible risk factors contingent
on insufficient sleep, fatigue resultant thereof, ways to improve public
services, etc. The MCTQ is currently being used in Groningen, Mu-
nich, and Basel. A particularly valuable feature of the MCTQ is the
separate treatment of workdays and free days. This division is left at
the respondent’s discretion, in the sense that they may consider Sat-
urday as a workday as long as they work on weekends. The timing of
daily activities is obviously different between workdays and free days,
and this difference is greater in later chronotypes (Roenneberg et al.,
2003).
It is obvious that individuals will commonly tend to follow their
individual preferences as long as this does not conflict with external
demands. It is of interest to what extent the MCTQ parameters,
reflecting actual behaviour, co-vary with the MEQ score, reflecting
preferred behaviour. Since both the MCTQ and the MEQ aim to
quantify ‘chronotype’ (Roenneberg et al., 2003) and since the latter
has been the standard instrument, it is useful to establish how well
indices produced by these two questionnaires correlate. This analysis
is based on 2481 respondents who filled out both questionnaires during
one online session. While going into separate analyses of parameters
contained in the MCTQ, we did not attempt to ‘disassemble’ the MEQ,
and restricted our investigation to its calculated overall score (for a
2.2. MATERIALS AND METHODS 19

principal component analysis of MEQ parameters, see Taillard et al.

(2004)). We also assess the extent to which the MCTQ self-ratings
are indicative of the tendency of the respondents to be ‘evening’ or
‘morning’ types, i.e., how well their perceived notions of ‘larks’ and
‘owls’ agree with the sleep timings they report.

2.2 Materials and Methods

We created a website (chrono.biol.rug.nl) on which both MCTQ
and MEQ were available for electronic submission. The English text of
the MCTQ was translated into Dutch. The translation was validated
by back-translating the Dutch version into English by an uninvolved
person. After the authors had verified the back-translation, the Dutch
version was posted on the site, along with the original English text.
The text of the MEQ was also available in both languages (the Dutch
version by Prof. Dr. G. A. Kerkhof, Amsterdam).
All students (ca. 20,000) of the University of Groningen (RuG)
were sent an email (containing a link to the MCTQ web page) on May
10, 2003 inviting them to participate in the study. This campaign was
also widely advertised via many different channels.
Upon submission of their MCTQ questionnaire, respondents auto-
matically received a report (to the supplied email address) containing
information on how the subject compares to the rest of the database.
The browser window subsequently displayed a page with a further link
to the MEQ page for those interested in filling this out as well. Thus,
all respondents to the MEQ also responded to the MCTQ, but not
vice versa.
Before being stored into the database, responses were checked for
consistency. These checks prevented, for example, wake-up times ear-
lier than sleep onset times, they ensured that ‘time-of-day’ fields were
within 0 and 24, etc. The criteria were selected just to filter out il-
logical responses while avoiding any constraint on the respondent’s
possible daily pattern. In the case any of these checks failed, the re-
spondent was given a message stating the error and asked to correct
it and resubmit the form. Eventually less than 0.1 % of completed
questionnaires were rejected.
From mid-May, 2003 till end of May, 2004, a total of 5055 sub-
jects responded, 49 % of which (2481) completed both questionnaires
(Fig 2.1). All MEQ score-derived correlations in the following sections
 tn
de
20 CHAPTER 2. A SURVEY OF CHRONOTYPES

o
p
A B
800

700
s

600

500
Number of res

co
re
400

300

200

100

0
10 20 30 40 50 60 70 80 21 30 39 48 57 66 75
Age, years MEQ s

Figure 2.1 Distribution of ages (A) and MEQ scores (B). n = 2481.

are based on this sub-sample. The predominant part of the respon-

dents were university students, and they were Dutch residents nearly
without exception.
Statistical analyses included the Pearson’s product-moment corre-
lation coefficient (r). To compare these, Fisher z transformation was
applied.

2.3 Results
The range of MEQ scores observed among 2481 respondents was 17–78
(of a potential 16–86). Based on multiple questions, the MEQ total
score results in a continuous measure leading to an almost Gaussian
distribution.
In view of the well-known age dependence of sleep timing (Park
et al., 1998a,b; Dijk et al., 2000), further analyses were carried out
for four age groups (I, <25 years; II, 25–30; III, 30–40; IV, >40)
separately. Table 2.1 provides statistics on the MEQ scores and several
2.3. RESULTS 21

d
n
o
p
ents
Number of res
350

300

250

200

150
A

4:57
B
3:47

100

50

0
2 4 6 8 10 2 4 6 8 10

Local time, h

Figure 2.2 Distribution of mid-sleep times, on free days (A, msf) and on workdays
(B, msw). n = 2481. Clock times near peaks are means of respective midsleep times.

variables derived from the MCTQ for both work days and free days.
The results show a progressive advance of sleep-times with increasing
age which is consistent with the results of another MCTQ-based study
involving a larger population sample (n ≈ 25, 000) with a broader age
distribution (Roenneberg et al., 2004b). We calculated the correlations
with the individual MEQ score for the following MCTQ variables:
sleep duration, sleep onset time (SO, i.e., bedtime plus sleep latency),
rise time (RT ), and midpoint of sleep (M S). This was done separately
for work days and free days (indicated by W and F , respectively),
since sleep timing is conspicuously different between work and free
days (Fig. 2.2). The results are summarized in Table 2.2.
The strongest correlation (|r| > 0.7) was found between MEQ and
the midpoint of sleep on free days (MSF; see also Roenneberg et al.
(2003)). Sleep onset correlated with the MEQ score usually slightly
better than sleep-end times, especially on work days; we discuss this in
more detail below. Sleep duration was not significantly correlated with
MEQ in our sample as was reported before (Roenneberg et al. (2003)
CHAPTER 2. A SURVEY OF CHRONOTYPES

Table 2.1 Average values of the MCTQ parameters included in the analysis (MEQ subsample), per age group. In parentheses
are standard deviations (h).

Age groups MEQ MCTQ Parameters

(years) n Score workdays
sleep onset mid end duration
I (<25) 1342 47.0 (10.5) 00:00 (1.10) 04:02 (0.95) 08:04 (1.10) 8.07 (1.06)
II (25–29) 372 47.8 (11.9) 23:48 (1.08) 03:43 (0.95) 07:38 (1.07) 7.85 (0.99)
III (30–39) 285 50.2 (12.2) 23:34 (1.03) 03:20 (0.83) 07:06 (0.90) 7.53 (0.99)
IV (>39) 482 54.5 (11.9) 23:42 (1.22) 03:23 (0.92) 07:04 (1.00) 7.36 (1.23)
all groups 2481 48.9 (11.5) 23:52 (1.12) 03:47 (0.98) 07:43 (1.15) 7.84 (1.11)

free days
sleep onset mid end duration
I 01:01 (1.20) 05:23 (1.23) 09:45 (1.43) 8.73 (1.34)
II 00:48 (1.35) 05:00 (1.23) 09:13 (1.45) 8.43 (1.30)
III 00:20 (1.15) 04:25 (1.15) 08:29 (1.45) 8.15 (1.25)
IV 00:19 (1.12) 04:04 (1.08) 07:48 (1.42) 7.50 (1.32)
all groups 00:46 (1.33) 04:57 (1.32) 09:09 (1.63) 8.39 (1.40)
22
 2.3. RESULTS
Table 2.2 Correlation coefficients (Pearson’s r) of some of the MCTQ parameters with the MEQ score (MEQ scores increases
from ‘eveningness’ ‘ to ‘morningness’). All correlations significant at the 0.01 level, two-tailed. For the total sample, the
correlations were tested by applying Fisher transformation and comparing those pairwise (zsleeponset vs zmidsleep , zmidsleep
vs zrisetime , and zsleeponset vs zrisetime ). All compared correlations were significantly different (p < 0.01) except for free-day
zsleeponset vs zrisetime .

Age groups workdays free days

(years) sleep onset mid end duration sleep onset mid end duration
I (<25) −0.61 −0.60 −0.43 0.18 −0.60 −0.72 −0.66 −0.09
II (25–29) −0.61 −0.61 −0.46 0.18 −0.66 −0.74 −0.66 −0.05
III (30–39) −0.56 −0.59 −0.45 0.18 −0.64 −0.71 −0.63 −0.15
IV (>39) −0.53 −0.58 −0.44 0.17 −0.67 −0.71 −0.56 −0.04
total −0.59 −0.61 −0.48 0.11 −0.64 −0.73 −0.66 −0.16

23
24 CHAPTER 2. A SURVEY OF CHRONOTYPES

and, in more detail, Roenneberg et al. (2004b)). Other variables, such

as sleep latency, time spent outdoors, time of peak alertness, least-
awake time and time spent reading before falling asleep, did not reach
significant correlation (|r| < 0.1 throughout) with MEQ score (and are
therefore not reported).

2.3.1 Association of sleep timing with MEQ score

Sleep duration showed a gradual reduction from younger to older re-
spondents both on workdays and free days (Table 2.1), but its correla-
tion with the MEQ score remained low in all age groups, from −0.15
in age group 30–39 y to 0.18 in people younger than 39 y (Table 2.2).
The correlation of sleep duration to MEQ score changes sign between
workdays and free days.
Workday sleep onset became progressively earlier from age group
I (<25) to age group IV (>39): the advance was on average 26 min
(Table 2.1). The oldest group (IV) has an 8-min delay relative to age
group III. As this delay is not counterbalanced by a corresponding
change in sleep-end time, resulting mid-sleep time in age group IV is
also slightly later than in age group III. The correlation of workday
mid-sleep time with the MEQ score (Table 2.2) stays at the same level
(−0.6) in all age groups, and is fairly similar to the correlation for
workday sleep onset and MEQ.
Sleep onset and rise times on free days follow the age-dependent
trend towards earlier clock time seen on workdays, except that the
age-dependent advance of sleep timing on free days is about twice as
large (79 vs 39 min based on MS; Table 2.1). Sleep onset advanced
by 18 min from age group I to IV on workdays vs 42 min on free days.
This difference is even more pronounced for sleep-end times (60 min
and 117 min, respectively).
Mid-sleep, rather than sleep onset or end, gave the best corre-
lation with the MEQ score. This can also be seen in the narrower
scatter for MEQ vs MSF compared to sleep onset or rise time on
free days (Fig. 2.3). MEQ scores correlated better with MSF (−0.73)
compared to MSW (−0.61). This MSF–MEQ correlation appeared
higher than the correlation between MEQ score and SOF (p < 0.001
for all correlation comparisons mentioned). This makes free-day sleep
timing dissimilar from work days, in which sleep onset and mid-sleep
are associated similarly well with the MEQ score (−0.61 and −0.59,
respectively).
2.3. RESULTS 25

Figure 2.3 Mean values of primary sleep timings (A, standard deviations in paren-
theses, min; upper bar, workdays; lower bar, free days). Association of MEQ score
with workday (B) and free-day (C) sleep timings. Two-tailed, all significant at 0.01
level. Lines are reduced major axes.
26 CHAPTER 2. A SURVEY OF CHRONOTYPES

0.74 76
75 78

Correlation with MEQ (–r)

75
0.72 81

76 84
0.70

dtio
ra
u d
nen
88
0.68
77

0.66 92
80
0.64

0.62

0 20 40 60 80 100
sleep onset % sleep sleep

Figure 2.4 Pearson’s correlation coefficients (r) of MEQ score and free-day sleep
timepoints (φx , see text). The largest value, observed at φ50 (i.e., msf), was 0.73.
Labels to points are standard deviations (min) for the corresponding φx .

2.3.2 MSF correlates better with the MEQ scores than any other
sleep-related phase marker
The different degrees of correlation between various sleep-related refer-
ence points, (sleep onset, mid-sleep, rise time) led us to perform a more
detailed examination, by dividing the sleep episode into 10-percentiles
φx (for x = 10, 20, . . . , 90). Figure 2.4 shows that on both sides of
MSF Pearson’s r gradually decreases towards the beginning and end
of the sleep period, strongly indicating that mid-sleep is (a) the best
time point for correlations between the MCTQ and the MEQ and (b)
may be the best marker for sleep-based assessments of chronotype.
The highest correlation between MSF and MEQ-scores is not merely
a consequence of reduced estimation error. Although the standard de-
viations observed for MSF are smaller than for sleep onset and rise
time, φ30 and φ40 showed even smaller standard deviations (Fig. 2.4)
but worse correlations to MEQ scores.
2.4. DISCUSSION 27

2.4 Discussion
Not surprisingly, sleep is on average earlier on workdays than on free
days, and there is less inter-individual variation in its timing. Several
authors (Roenneberg et al., 2003; Taillard et al., 1999; Dijk et al.,
2000) have observed that later chronotypes suffer more from sleep
loss caused by their work schedules. The larger discrepancy between
the individual preference for specific sleep times and the obligations of
social schedules (e.g., school or work begin) lead to larger differences in
sleep durations between workdays and free days. While sleep onset is
later in owls than in larks, sleep end on work days is largely dictated by
the alarm clock. For these reasons MEQ may be correlated positively
with sleep duration on work days, and negatively with sleep duration
on free days (Table 2.2). This means that earlier chronotypes get more
sleep on work days than later chronotypes whereas the opposite is true
for free days (Roenneberg et al., 2003, 2004b).
The clear result that MSF, rather than other phase reference points
of sleep shows the best correlation with MEQ-scores may reflect the
considerable variation in individual sleep duration. On work days,
sleep timing (especially sleep end) is largely dictated by social timing,
and for later chronotypes, sleep duration is kept near its minimum
(depending on age: 7.36–8.07 h). On free days, sleep duration varies
within a broader range (averages for different age groups: 7.50–8.73 h).
While sleep duration on work days is directly influenced by the alarm
clock, it may still be influenced on free days by the work schedules
as an after-effect because working people, particularly later chrono-
types, have to compensate for a sleep debt accumulated during the
work week. Since sleep duration shows little systematic variation with
the MEQ-score nor with the MCTQ-derived chronotype (Roenneberg
et al., 2003, 2004b), each chronotype category contains a similar por-
tion of short and long sleepers. Thus, short sleepers appear to go to
bed later and rise earlier around their individual mid-sleep on free days
while long sleepers go to bed earlier and rise later around their MSF.
A physiological circadian phase marker, the dim-light melatonin
onset (DLMO), also shows higher correlation with mid-sleep time than
with either sleep onset or rise time (Martin, Eastman, 2002; Terman
et al, 2001). A correlation between dim-light melatonin offset, cir-
cadian preference and sleep timing has also been reported (Laberge
et al, 2000; this study, however, did not specifically report mid-sleep
correlations. MSF also shows high correlations with the minimum of
28 CHAPTER 2. A SURVEY OF CHRONOTYPES

the daily cortisol rhythm assessed in a constant routine (Roenneberg

et al., 2004b).
Our analysis demonstrates that an estimate of the actual timing of
sleep on free days obtained with the MCTQ is strongly related to the
MEQ score, and that the timing of mid-sleep on free days in particu-
lar is a good predictor for chronotype (as judged by sleep preferences).
Detailed information on the timing of human behaviour under natu-
ral circumstances and a refined description of individual chronotype
are prerequisites for future research on the mechanisms of circadian
rhythm generation and entrainment. For example, possibilities to gain
reliable insights into the genetic basis of a given trait are greatly im-
proved by refined phenotyping.

2.5 Acknowledgements
This work is supported by the BrainTime program (5th European
Framework Program, grant QLG3-CT-2002-01829). We are grateful
to the University of Groningen, and Drs. J. A. den Boer for help with
addressing all university students, and Bram van Bunnik for technical
assistance.