Chronobiology International, Early Online: 1–13, (2014)

! Informa Healthcare USA, Inc.

ISSN: 0742-0528 print / 1525-6073 online
DOI: 10.3109/07420528.2014.957305

ORIGINAL ARTICLE

Investigation of the effectiveness of a split sleep schedule in

sustaining sleep and maintaining performance

Melinda L. Jackson1, Siobhan Banks2, and Gregory Belenky3

1
Melbourne School of Psychological Sciences, The University of Melbourne, Melbourne, Australia, 2Centre for Sleep
Research, University of South Australia, Adelaide, Australia, and 3Sleep and Performance Research Center,
Washington State University, Spokane, WA, USA
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14

Shift work is common in today’s society, and is associated with negative health outcomes, and accidents and
incidents. These detrimental effects can be primarily attributed to sleeping and working at an adverse circadian time.
The aim of this study was to examine whether a split sleep schedule is as effective as a consolidated day shift or night
shift schedule for maintaining performance and sustaining sleep. Fifty-three healthy male volunteers (mean ± SD
age ¼ 26.51 ± 4.07 years) underwent a randomized three condition study design. A split sleep condition involving two
5-h sleeping opportunities in 24 h [time in bed (TIB) 0300 h–0800 h and 1500 h–2000 h] was compared to a 10-h
consolidated nighttime sleep (TIB 2200 h–0800 h) and 10-h consolidated daytime sleep (TIB 1000 h–2000 h). All
participants underwent a baseline period of 10 h of nocturnal time in bed (TIB) followed by a 5-d simulated workweek
spent in one of the three conditions. Polysomnography, psychomotor vigilance task, digit-symbol substitution task
and subjective state were assessed. During the 5-d simulated workweek, participants in the nighttime sleep condition
For personal use only.

slept the most (total sleep time per day (TST) 8.4 h ± 13.4 min), followed by the split sleep condition (TST
7.16 h ± 14.2 min) and the daytime sleep condition (TST 6.4 h ± 15.3 min). Subjective sleepiness was highest in the
daytime sleep condition and lowest in the nighttime sleep condition. No significant differences in performance were
observed between the conditions. Compared to a nighttime consolidated sleep opportunity or split sleep, placement
of a consolidated sleep opportunity during the day yielded truncated sleep and increased sleepiness. Further research
in real-world situations is warranted to fully assess the efficacy of alternative split sleep schedules for improving safety
and productivity.
Keywords: Fatigue risk management, scheduling, shift work, sleepiness, split sleep, vigilance

INTRODUCTION
Shift work and extended work hours are associated with
daytime sleepiness and insomnia, and reduced alert-
ness, work productivity and quality of life (Rajaratnam &
Arnedt, 2001). Shift work has also been linked to
negative health effects, including increased cardiovas-
cular morbidity and mortality, and increased accident
risk (Drake & Wright, 2011; Rajaratnam & Arnedt, 2001;
Wehrens et al., 2012). One key problem for shift workers,
from which many of these negative consequences stem,
is reduced sleep. The extensive literature on shift work
indicates that, for the same duration of consolidated
sleep opportunity, actual sleep obtained is critically
dependent on the placement of the sleep opportunity
with respect to the circadian phase (Åkerstedt, 2003;
Drake & Wright, 2011). Shift workers coming off duty in
the morning, with an 8–10 h consolidated sleep
opportunity, are typically only able to obtain about 5 h
of daytime sleep before their sleep is truncated by the
combination of decreasing homeostatic drive for sleep
and increasing circadian drive for wake (Goel et al.,
2011; Van Dongen et al., 2010). This leads to cumulative
sleep restriction over extended periods of time. Studies
of restricted sleep show that over days of sleep restric-
tion there is a cumulative sleep dose-dependent deg-
radation in alertness and performance, which can occur
even during mild sleep restriction (loss of 51 h of sleep/
night); and that 7–8 h of consolidated nocturnal sleep
in 24 h appear to sustain performance over multiple
days, if not indefinitely (Belenky et al., 2003; Van
Dongen et al., 2003).
When the main sleep period is restricted or absent,
adding supplementary naps improves alertness and
performance (Bonnet, 1991; Dinges et al., 1987;

Submitted February 11, 2014, Returned for revision July 27, 2014, Accepted July 30, 2014
Correspondence: Melinda L. Jackson, Melbourne School of Psychological Sciences, The University of Melbourne, Melbourne 3010,
Australia. Tel: 613 9035 6129. Fax: 613 9347 6618. E-mail: melinda.jackson@unimelb.edu.au

1
 2 M. L. Jackson et al.
Ficca et al., 2010; Sallinen et al., 1998; Takeyama et al.,
2005). The term ‘‘split sleep’’ means two or more sleep
opportunities in a 24-h period, ranging from a main
sleep and a supplemental nap (e.g. 6 and 2 h), through a
main sleep and several naps, to multiple naps with no
clear main sleep (Belenky et al., 2008, 2011; Bonnet &
Arand, 2003; Takeyama et al., 2005). Spilt sleep sched-
ules are common practice in a number of industries
including healthcare, maritime and transport. Split sleep
may restore alertness and performance as effectively as
consolidated sleep (Nicholson et al., 1985; Schweitzer
et al., 2006). Laboratory evidence to date suggests that
when the longer sleep period is placed at night, split
sleep is similar to consolidated sleep in both duration
and recuperative value, and that the critical factor in
sustaining performance is total sleep time in 24 h
(Belenky et al., 2008). For example, physicians working
day shifts and sleeping at night, versus working night
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
shifts and having their main sleep during the day
supplemented by on-shift nighttime naps, are able to
accumulate approximately 7 h of total sleep time over
24 h and perform equally well on the psychomotor
vigilance task (PVT) in both conditions (McDonald et al.,
2013).
Total sleep time is strongly influenced by the time of
day that the sleep opportunity occurs. A laboratory
For personal use only.
study of maritime workers found that sleep opportu-
nities that were placed during the day (e.g. 12:00 h–
18:00 h and 18:00 h–24:00 h) had, on average, 1–1.5 h less
total sleep time compared to sleep periods placed
during the night (Eriksen et al., 2006). While studies
support the idea of flexibility in splitting sleep for
maintaining performance (Folkard & Tucker, 2003;
Horne & Reyner, 1999; Jones et al., 2006), many have
failed to take into account whether the placement of the
split sleep opportunity was at an adverse circadian time
(e.g. the longer sleep was placed at night (nocturnal
anchor sleep) and the shorter sleep during the day). It is
therefore unclear whether split sleep per se degrades
performance. Further to this, it is important to also
compare the effect of a split sleep opportunity to the
same total duration of sleep opportunity consolidated at
night and during the day. The current study aims to
overcome these shortcomings by examining split sleep
opportunities placed during a period of high sleep
propensity, and by comparing split sleep to both a
consolidated daytime and nighttime sleep opportunity.
Data on the effectiveness of a split sleep schedule will
have particular relevance for occupational environ-
ments in determining the adequate duration and
effective placement of the sleep opportunities.
The aim of this study was to compare a sleep split
schedule, where the sleep opportunities are split evenly
into two sleep periods, to two conditions in which sleep
was consolidated into a single period (either daytime
consolidated sleep and or nighttime consolidated sleep),
to determine the effects of those sleep patterns on sleep,
performance and subjective state.
MATERIALS AND METHODS
Participant recruitment and screening
Participants were recruited from the general population
of healthy young men ranging in age from 22 to 40 years.
Blood samples were collected during the study to assess
metabolic measures (data not shown), therefore women
and obese men (BMI430) were excluded from the
study, as were participants whom would have difficulty
with intravenous (IV) catheters. Prospective participants
responded to advertisements in local newspapers and
on the internet and underwent an initial telephone
screening interview. They then attended two laboratory-
based screening sessions beginning with an informed
consent procedure, and included a physical exam, blood
and urine samples and a variety of questionnaires to
assess suitability for participation. Participants were
required to meet the following additional criteria to be
eligible for participation in the study: psychologically
and physically healthy, as assessed by the study phys-
ician and medical history; no clinically significant
abnormalities in blood and urine and free of traces of
drugs; free of traces of alcohol by breathalyzer; no
history of drug or alcohol abuse in the past year and no
history of methamphetamine abuse; not a current
smoker (by questionnaire); no history of a learning
disability or brain injury (by questionnaire); no previous
adverse reaction to sleep deprivation (by history and
questionnaire); not vision impaired unless corrected to
normal; no sleep or circadian disorder; good habitual
sleep, between 6 and 10 h in duration, verified by wrist
actigraphy and diary in the week before the study;
regular bedtimes, habitually getting up between 0600 h
and 0900 h; neither extreme morning- nor extreme
evening-type (by questionnaire); no travel across time-
zones or shift work within one month of entering the
study; native English speaker. This study was approved
by the Washington State University Institutional Review
Board (IRB) and all participants gave written informed
consent. The experimental protocol conformed to inter-
national ethical standards (Portaluppi et al., 2010).
Design
A three-condition design was developed – one experi-
mental condition (split sleep opportunity) and two
control conditions (consolidated nighttime sleep oppor-
tunity and consolidated daytime sleep opportunity). All
three conditions were continued for a simulated work-
week of 5 consecutive days. Participants in the split
sleep and consolidated nighttime sleep conditions
arrived in the laboratory at 0900 h on Day 1 and
completed data collection for the present study at
1400 h on Day 10. Participants in the daytime sleep
condition were part of a separate study that involved
two 5-d work periods, and lived in the same laboratory as
participants in the current study for 16-d (Van Dongen
et al., 2011); only data from the first 5-d work week from
that study, including baseline and recovery, was used in
Chronobiology International

this study. The same exclusion criteria were applied to
participants in both studies.
The study began with two baseline days with night-
time sleep, followed by a 5-d simulated workweek in one
of the three sleep opportunity conditions, and ended
with two recovery days, again with nighttime sleep
(Figure 1). The design held constant at 10 h the total
daily amount of time available for sleep for all three
conditions – split sleep opportunity (0300 h–0800 h and
1500 h–2000 h), consolidated nighttime sleep opportun-
ity (2200 h–0800h) and consolidated daytime sleep
opportunity (1000 h–2000 h). Thus, all three conditions
had the same 90-h total sleep opportunity (10 h per day)
across the 10-d study. A 10-h sleep opportunity was
provided in all conditions to allow participants to obtain
sufficient sleep throughout the study and to approxi-
mate a 14-h on duty/10-h off duty schedule (Banks et al.,
2010; Van Dongen et al., 2011). While this amount of
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
sleep would not typically occur in the real world, this
laboratory-based study allowed us to determine the
effect of different sleep schedules independent of sleep
restriction imposed by the study protocol.
The split sleep experimental condition was imple-
mented as two, 5-h daily sleep opportunities; one from
0300 h to 0800 h and the other from 1500 h to 2000 h.
This placed the sleep opportunity in the first instance at
For personal use only.
a time when sleep propensity is high and in the second
instance at a time when sleep propensity, at least later in
the interval, is low. The consolidated nighttime control
sleep condition was implemented as a 10-h daily
nighttime sleep opportunity from 2200 h to 0800 h.
This placed the sleep opportunity at times when it is
likely that homeostatic drive for sleep was high and the
circadian drive for wakefulness was low, promoting
sustained, consolidated sleep (Dijk & Czeisler, 1994).
Such a consolidated nighttime sleep opportunity would
be typically associated with day shift work. The con-
solidated daytime control sleep condition was imple-
mented as a 10-h daytime sleep opportunity from 1000 h
to 2000 h. This placed the sleep opportunity initially at a
time when sleep propensity is high due to high homeo-
static drive and subsequently at a time when sleep
propensity is generally low due to the increasing
circadian drive for wakefulness. In addition, transition
naps were implemented for both split sleep and daytime
sleep conditions to aid in the transition to the workweek
schedule at the beginning of the 5-d work week, and to
aid the switch back to nighttime sleep at the end of the
workweek (Figure 1).
During their days in-residence at the Sleep and
Performance Research Center, participants had no
contact with the outside world. They slept, ate and
completed performance tests within the confines of the
sleep laboratory. There was no cell phone contact,
no email, no visitors and no live television, radio or
internet. During scheduled sleep periods, bedroom
lights were turned off and participants were not
permitted to engage in any activities other than sleeping
! Informa Healthcare USA, Inc.
Efficacy of split sleep schedule on sleep and performance 3
or resting. Ambient temperature throughout the labora-
tory was kept at 72 ± 2 F, and lighting levels were fixed
at 50 lux during scheduled wakefulness. Participants
were served 3 calorie-controlled meals per day at fixed
intervals.
Measures
To assess effects of schedule on sleep and performance,
measurements of sleep, performance and subjective
state were made (Figure 1). During off-duty days,
performance testing was significantly reduced to
mimic rest days in the real world. Performance testing
practice occurred on Day 1; these practice sessions were
not used for analysis.
Sleep
During most scheduled sleep periods, PSG was recorded
using digital equipment (Nihon Kohden, Foothill Ranch,
CA). Electroencephalography (EEG) was recorded from
frontal (F3, F4), central (C3, C4) and occipital (O1, O2)
locations, referenced against the mastoids (M1, M2).
Electrooculography (EOG), electromyography (EMG) and
electrocardiography (ECG) were also recorded. Sleep
stages were scored using standard technical specifica-
tions and rules recommended by the American Academy
of Sleep Medicine (AASM; Iber & Ancoli-Israel, 2007). The
following PSG/Sleep variables were compared between
conditions, time in bed, total sleep time, N3 sleep time,
REM sleep time, N2 sleep time, N1 sleep time, sleep
latency, latency to N3 sleep, latency to REM sleep.
The sleep periods that were recorded and the com-
parisons made across conditions are shown in Figure 1.
Every third or fourth day, electrodes were removed to
give participants an opportunity to take a shower and to
heal any skin irritation caused by the electrodes.
Performance and subjective assessments
The PVT is a standard assay of vigilance used to assess
fatigue, and was used as the primary outcome measure
in the current study (Dorrian et al., 2005). The number
of performance lapses (reaction times greater than
500 ms) was extracted. The PVT has high sensitivity to
fatigue and favorable statistical properties (Lim &
Dinges, 2008). A 10-min PVT was administered alone
or in combination with the subjective state measures for
the three conditions. For all conditions across the
workweek, eight PVT sessions were performed in every
24-h period.
Secondary performance outcomes were derived from
a computerized neurobehavioral test battery, adminis-
tered either alone or in combination with a PVT. The
battery consisted of the computerized versions of the
Digit Symbol Substitution test (DSST), the Karolinska
Sleepiness Scale (KSS; Åkerstedt & Gillberg, 1990); a
visual analog scale of mood, from elated to depressed
(VAS-M; Van Dongen et al., 2004); and the Positive Affect
Negative Affect Schedule (PANAS; Watson et al., 1988),
which is a measure of positive and negative emotion.
 4 M. L. Jackson et al.
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
For personal use only.

FIGURE 1. PSG recording, Psychomotor Vigilance Test (***), Neurobehavioral Test Battery (X) schedule for the consolidated nighttime
sleep (top panel), split sleep (middle panel) and consolidated daytime sleep (bottom panel) conditions. Baseline periods included PSG A
and B for all conditions. For the nighttime and daytime sleep conditions, the work week sleep periods included PSG D and E, and for the
split sleep condition the work period included PSG E, F, G and H. The recovery sleep periods included PSG F and G for the nighttime sleep
condition, PSG I and J for the split sleep condition, and PSG G and H for the daytime sleep condition.

Chronobiology International

For each, an overall score was extracted, except for the
PANAS, for which both positive and negative affect
scores were determined. The DSST is a performance test
involving matching numbers to symbols (Wechsler,
1997). The computer screen showed a key with a set of
nine symbols, each with a corresponding digit (1–9).
When given a symbol in another, fixed location on the
screen, participants were required to type its corres-
ponding number. After the response, a new symbol was
immediately presented. The number of correct
responses in the 3-min task duration was extracted,
yielding a measure of cognitive throughput. The DSST is
sensitive to acute total sleep deprivation and chronic
sleep restriction (Van Dongen et al., 2003).
Statistical analysis
The primary statistical design involved a mixed model
analysis of variance (ANOVA) comparison of the within-
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
subject effects of condition (split sleep, nighttime sleep
and daytime sleep) on the between-subjects factors of
PSG variables, performance (PVT lapses and DSST) and
subjective measures (KSS, PANAS and VAS-M). For the
PSG variables, the secondary statistical analyses involved
a comparison of within-subjects effects of condition by
sleep period (baseline sleep period 1 [BL1], baseline sleep
period 2 [BL2], work period 1 [W1], work period 2 [W2],
For personal use only.
recovery [R]) mixed model ANOVAs. PSGs A and B were
used in the analysis to compare baseline sleep across
conditions (Figure 1). PSGs C and D for the nighttime and
daytime sleep conditions, and PSGs E, F, G and H in the
split sleep condition were used in the analysis comparing
sleep during the work week across conditions. For the
analysis comparing recovery sleep across conditions
included only one equivalent PSG in each condition;
PSG G for the nighttime sleep condition, PSG J for the
split sleep condition and PSG H for the daytime sleep
condition (Figure 1). A sub analysis of the six
PSG-recorded napping opportunities in the split sleep
condition was also examined using a mixed model
approach, comparing a within-subject factor (nap type:
afternoon nap, morning nap), with post-hoc t-tests.
In order to determine whether performance and sub-
jective measures changed across work days and within
work days between groups, secondary statistical analyses
involved condition by workday (Workday 1–5), and
condition by time of day (Session 1–8 for PVT and
Session 1–4 for all other measures), mixed model
ANOVAs. Planned group comparisons were performed
for all analyses.
RESULTS
Participants
Fifty-three male participants completed the study
(mean ± SD age ¼ 26.51 ± 4.07 years). Nineteen partici-
pants were randomized to the consolidated nighttime
sleep condition, 17 were randomized to the split sleep
condition, and 17 were randomized to the consolidated
! Informa Healthcare USA, Inc.
Efficacy of split sleep schedule on sleep and performance 5
TABLE 1. Demographic details of the participants in the con-
solidated nighttime sleep, split sleep and consolidation daytime
sleep conditions.
Nighttime sleep Split sleep Daytime sleep
Age (years) 25.8 26.3 27.5
BMI 24.4 23.9 24.8
MEQ 36.8 38.4 36.9
ESS 4.2 3.6 4.6
PQSI 2.2 2.3 2.5
BMI, body mass index; MEQ, morningness–eveningness ques-
tionnaire; ESS, Epworth Sleepiness Scale; PQSI, Pittsburgh Sleep
Quality Index.
daytime sleep condition. One participant in the night-
time sleep condition was excluded from the sleep and
performance analysis due to a suspected sleep disorder,
leaving 18 participants in the consolidated nighttime
sleep condition. There was no significant difference in
age of the participants across the conditions (F2,48 ¼
0.74; p ¼ 0.48). Table 1 displays the demographic infor-
mation of participants in each group.
Sleep
Two participants in the daytime sleep condition were
excluded from the sleep analysis – one due to poor sleep
efficiency throughout the study (due to flu-like symp-
toms) and another due to light exposure during the
sleep periods. This left 15 participants in the daytime
sleep condition, 18 participants in the nighttime sleep
condition and 17 participants in the split sleep condition
for the PSG analysis.
The PSG-recorded sleep periods for the three sleep
conditions – nighttime sleep, the split sleep and the
daytime sleep for baseline (two recordings; BL1, BL2),
work period (2 recordings; W1, W2) and recovery (one
recording; R) – were compared. By design, time in bed
(TIB) was equivalent for the three conditions across the
baseline, work and recovery periods. For the PSG
recordings, total TIB was 20 h for the two baseline
recordings, 20 h for the two work period recordings and
10 h for the single recovery recording.
The main effect of condition was significant for
minutes of total sleep time (TST) (F2,204 ¼ 12.49,
p50.001). Participants in the nighttime sleep condition
obtained significantly more sleep and participants in the
daytime and split sleep conditions, and participants the
split sleep condition obtained significantly more sleep
than participants in the daytime sleep condition
(p50.05; Figure 2A). There was a significant interaction
between condition and sleep period for TST
(F8,192 ¼ 5.92, p50.001). During the work period, the
nighttime sleep condition obtained significantly more
TST than participants in the split and daytime sleep
conditions (ps50.05). During recovery, participants in
split sleep condition obtained more TST than partici-
pants in daytime sleep condition (p50.05). There was
 6 M. L. Jackson et al.
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
For personal use only.

FIGURE 2. Polysomnographic measures across two baseline sleep periods (BL1, BL2), two sleep periods during the work week (W1, W2),
and one recovery sleep period (R) for the nighttime sleep (black circles), split sleep (grey squares) and daytime sleep (white triangles)
conditions. Panel A depicts total sleep time (TST), panel B depicts stage N2 sleep, panel C depicts REM sleep and panel D depicts sleep
onset latency in minutes across the five sleep periods. Error bars represent standard errors.

also a main effect of sleep period (F4,192 ¼ 21.07,
p50.001). Participants obtained more sleep during the
baseline period compared to the work period and
recovery (p50.05).
For minutes of stage N2 sleep there were no signifi-
cant differences between conditions (F2,204 ¼ 1.11,
p ¼ 0.33; Figure 2B). There was a significant interaction
between sleep period and condition for stage N2 sleep
(F8,192 ¼ 6.10, p50.001). During the work period, par-
ticipants in the daytime sleep condition obtained less
N2 sleep than participants in the nighttime sleep
condition. There was a significant main effect of sleep
period (F4,192 ¼ 22.97, p50.01), with the most N2
sleep obtained during baseline period and the least
obtained during the work period.
Overall, REM sleep differed significantly between
all three conditions (F2,204 ¼ 16.52, p50.001; Figure
2C). Participants in the split sleep condition obtained
significantly more REM sleep than the daytime sleep
condition, however, still obtained significantly less than
the nighttime sleep condition, despite having equal
amounts of baseline time in bed. There was a significant
interaction between condition and sleep period for REM
sleep (F8,192 ¼ 5.08, p50.001). During the work period,
participants in the nighttime sleep condition obtained
more REM sleep than the split and daytime sleep
conditions. There was also a significant main effect of
sleep period (F4,192 ¼ 3.67, p50.01).
For sleep onset latency (SOL; in minutes), there were
no significant differences across the conditions
(F2,192 ¼ 1.94, p ¼ 0.15; Figure 2D). There was a signifi-
cant interaction between sleep period and condition for
SOL (F8,192 ¼ 4.75, p50.001). During the work period,
participants in the nighttime sleep condition had
Chronobiology International

FIGURE 3. Average total sleep time in minutes during the after-
noon naps (1500 h–2000 h; grey bars) and morning naps (0300 h–
0800 h; black bars) in the split sleep condition. Error bars represent
standard errors.
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
shorter SOL than participants in the daytime sleep
condition. There was a significant main effect of sleep
period (F4,192 ¼ 3.23, p50.05). For latency to stage N3
sleep, there were no significant differences between
conditions (F2,48 ¼ 1.57, p ¼ 0.22), however there was a
significant interaction between sleep period and condi-
tion (F8,192 ¼ 3.70, p50.001; Figure 2E). During the
For personal use only.
work period, participants in the split sleep condition
displayed significantly longer latency to stage N3 than
participants in the nighttime sleep condition. During the
recovery period, participants in the daytime sleep
condition displayed longer latency to stage N3 than
participants in the nighttime and split sleep conditions.
There was also a significant main effect of sleep period
(F4,192 ¼ 3.58, p50.01).
For latency to REM sleep, stage N1 sleep or stage N3
sleep, there were no significant differences between
conditions, or were there any interactions between
condition and sleep period (all p40.05).
Sleep in the split sleep condition
A sub-analysis was performed on the split sleep condi-
tion data, comparing the afternoon naps and morning
naps. There was a significant main effect of nap type for
TST (Figure 3; F5,80 ¼ 29.94; p50.001). TST was signifi-
cantly longer in the morning nap opportunities
(M ¼ 260.2 ± 7.56 min) than in the afternoon nap oppor-
tunities (M ¼ 154.3 ± 6.63 min; ps50.001). There was
significantly more N3 in the morning naps compared
to the afternoon naps (58.1 ± 4.3 min versus
33.3 ± 3.3 min; F5,80 ¼ 7.66; p50.001). Similarly, more
REM sleep was observed in the morning naps compared
to the afternoon naps (69.7 ± 3.1 min versus
34.2 ± 3.0 min; F5,80 ¼ 19.39; p50.001). Stage N1 sleep
differed significantly between the nap periods
(F5,80 ¼ 4.09; p50.01). The first morning nap and the
last afternoon nap had less stage N1 sleep than the other
nap periods. Stage N2 sleep also differed significantly
between the nap periods (F5,80 ¼ 4.09; p50.01), with
! Informa Healthcare USA, Inc.
Efficacy of split sleep schedule on sleep and performance 7
the morning naps having more N2 sleep than the
afternoon naps (123.4 ± 4.2 min versus 78.4 ± 5.4 min).
There was no difference in sleep latency between the
nap periods (p ¼ 0.48).
Performance
Psychomotor vigilance task
The primary performance outcome measures for the
study were the number of lapses (RTs4500 ms) and
fastest 10% of RTs on the PVT. Performance data from
the two participants who were excluded from the sleep
analyses was examined and compared to the mean PVT
performance of the participants who were included in
the sleep analysis, and they did not significantly differ.
Therefore, the performance data for these participants
was included in the analyses. Two participants in the
nighttime sleep condition and two participants in the
split sleep condition were excluded from the PVT
analysis as they were found to be noncompliant on
this task. These participants exhibited an overall average
of 8.82 (SD 4.87) lapses on the PVT, whereas the other
participants had an overall average of only 1.99 (SD
1.56). This level of PVT performance is worse than that
reported in severely sleep restricted participants
(Belenky et al., 2003). This left 17 participants in the
daytime sleep condition, 16 participants in the night-
time sleep condition and 15 in the split sleep condition
for the PVT analysis.
The primary statistical analysis of the effect of
condition was not statistically significant for PVT
lapses (F2,1677 ¼ 0.94, p ¼ 0.39). In order to examine
changes in performance over days within each condi-
tion, a further analysis of PVT lapses investigated the
interaction of condition by work day (Figure 4A). This
two-way interaction was not significant (F8,1665 ¼ 0.95,
p ¼ 0.47), but the main effect of work day was statistic-
ally significant (F4,1665 ¼ 4.72, p ¼ 0.001). A secondary
analysis of PVT lapses investigated the interaction of
condition by session (time of day), (collapsed over days).
The two-way interaction was statistically significant
(F14,1656 ¼ 6.90, p50.001; Figure 5A). In sessions 1
and 2, participants in the daytime sleep condition had
significantly fewer lapses compared to the nighttime
sleep condition (post-hoc ps50.05). In session 3, par-
ticipants in the daytime sleep condition had signifi-
cantly fewer lapses compared to the split sleep
condition (p50.05). In sessions 4 and 5, participants
in the split sleep condition had more lapses than the
other conditions (ps50.05). In session 6, participants in
the split sleep condition had more lapses than partici-
pants in the daytime sleep condition (p50.05). In
session 7, participants in the split sleep condition had
more lapses than the nighttime sleep condition
(p50.05). In session 8, participants in the nighttime
sleep condition had more lapses than the daytime sleep
condition (p50.05). The main effect of session was also
significant, with lapses increasing across the work shift
(F7,1656 ¼ 16.40, p50.001).
 8 M. L. Jackson et al.
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
For personal use only.
FIGURE 4. Lapses (Panel A) and fastest 10% of reaction times (Panel B) on the 10-min Psychomotor Vigilance Test (PVT), Karolinska
Sleepiness Scale (KSS) score (Panel C) and Visual Analogue Mood rating scale (Panal D) as a function of days at baseline and in the 5-d work
period in the nighttime sleep (black circles), split sleep (grey squares) and daytime sleep (white triangles) conditions. Error bars represent
standard errors.
The effect of condition for PVT fastest 10% of RTs was
not statistically significant (F2,1677 ¼ 0.90, p ¼ 0.41), or
was the two-way interaction of condition by work day,
collapsed over session, was not significant
(F8,1665 ¼ 1.41, p ¼ 0.34). However, the main effect of
work day was statistically significant (F4,1665 ¼ 2.45,
p ¼ 0.04), with a slight increase in RTs towards the end
of the work week (Figure 4B). Secondary analysis of
interaction of condition by session, collapsed over days
for PVT fastest 10% of RTs revealed a significant
interaction (F14,1656 ¼ 5.50, p50.001; Figure 5B). The
main effect of session was also statistically significant
(F7,1656 ¼ 7.24, p50.001).
Digit-symbol substitution test
For the number of correct responses on the DSST, the
effect of condition was not statistically significant
(F2,968 ¼ 1.93, p ¼ 0.15), nor was the interaction
between condition and work day (F8,956 ¼ 1.34,
p ¼ 0.22). However, there was significant main effect of
work day (F4,956 ¼ 22.66, p50.001), indicative of the
known learning effect that occurs on this task. Analyses
of condition by session showed a significant interaction
(F6,959 ¼ 4.87, p50.001; Figure 5C). Participants in the
nighttime sleep condition had significantly higher DSST
performance than participants in the split and daytime
sleep conditions in sessions 3 and 4 (p’s50.05).
Subjective measures
Karolinska sleepiness scale
For the KSS, the primary analysis focusing on the main
effect of condition yielded a significant effect (F2,968 ¼
5.50, p ¼ 0.004). The two-way interaction of condition by
work day was also statistically significant, with partici-
pants in the daytime sleep condition reporting signifi-
cantly greater sleepiness on work days 1–4 (F8,956 ¼
3.11, p ¼ 0.002; Figure 4C). There was no significant
main effect of work day for KSS scores (p ¼ 0.74). The
analysis of the interaction of condition by session
was statistically significant (F6,959 ¼ 22.60, p50.001;
Figure 5D). Daytime sleep participants reported higher
levels of sleepiness towards the end of their work shift
relative to the other two conditions. There was also a
main effect of session for KSS scores (F3,959 ¼ 25.78,
p50.001).
Mood rating scales
For the VAS-M, the primary analysis focusing on the
effect of condition yielded a trend to significance
Chronobiology International

Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
For personal use only.
FIGURE 5. Lapses on the Psychomotor Vigilance Test (PVT;
Panel A), PVT fastest 10% of reaction times (Panel B), Digit
Symbol Substitution Test (DSST) correct responses (Panel C) and
Karolinska Sleepiness Scale (KSS) scores (Panel D) at each test
session per work day, collapsed over the 5-d work period for the
nighttime sleep (black circles), split sleep (grey squares) and
daytime sleep (white triangles) conditions. Error bars indicate
standard error. Note that sessions were conducted at different
times of day within each condition. For the PVT, the nighttime
sleep condition testing was at 0900, 0930, 1200, 1230, 1500, 1530,
1800 and 1830; split sleep condition testing was at 2100, 2130, 0000,
0030, 0900, 0930, 1200 and 1230, and the daytime sleep condition
testing was at 2100, 2130, 0000, 0030, 0300, 0330, 0600 and 0630.
For KSS and DSST, the nighttime sleep condition testing was at
0900, 1200, 1500 and 1800, the split sleep condition testing was at
2100, 0000, 0900 and 1200, and the daytime sleep condition, testing
was at 2100, 0000, 0300 and 0600.
! Informa Healthcare USA, Inc.
Efficacy of split sleep schedule on sleep and performance 9
(F2,48 ¼ 2.83, p ¼ 0.07), as did the two-way interaction of
condition by work day (F8,957 ¼ 1.77, p ¼ 0.08;
Figure 4D). Participants in the split sleep condition
rated their mood lower than the other two conditions
across all work days (ps50.05). The main effect of work
day was not significant (F4,957 ¼ 1.79, p ¼ 0.13). In order
to investigate subjective mood as a function of session,
a further analysis of the VAS-M examined the inter-
action of condition by session. The two-way interaction
was not significant (F2,960 ¼ 1.69, p ¼ 0.12). There was,
however, a significant main effect of session (F3,960 ¼
4.92, p ¼ 0.002).
For the positive affect scale of the PANAS, the
primary analysis focusing on the main effect of condi-
tion yielded non-significance (F2,48 ¼ 2.09, p ¼ 0.13).
Interactions of condition by work day trended towards
significance (F8,956 ¼ 1.86, p ¼ 0.06), with a significant
main effect of work day emerging (F4,956 ¼ 5.45,
p50.001). Examination of PANAS positive affect as a
function of session showed a significant interaction of
condition by session (F6,959 ¼ 4.34, p50.001). Across
most of the sessions within each work day, participants
in the split sleep condition had significantly lower
positive affect scores than participants in the nighttime
and daytime sleep conditions (ps50.05). There was also
a significant main effect of session (F3,959 ¼ 12.73,
p50.001), with positive affect scores decreasing signifi-
cantly across the work period (ps50.05).
For the negative effect scale of the PANAS, the
primary analysis focusing on the main effect of condi-
tion yielded non-significance (F2,968 ¼ 0.87, p ¼ 0.42).
The interaction between condition and work day was
not significant (F6,956 ¼ 0.81, p ¼ 0.59), however there
was a significant main effect of work day (F4,956 ¼ 2.67,
p ¼ 0.03). There was no significant interaction of condi-
tion by session (F6,959 ¼ 0.92, p ¼ 0.48), nor was there
was a significant main effect session (F3,959 ¼ 1.31,
p ¼ 0.27) for negative affect scores.
DISCUSSION
This study examined whether a split sleep schedule was
as effective as a consolidated period of sleep, placed
either during the day or during the night, for maintain-
ing sleep and performance across a simulated week of
work. Overall, participants in the nighttime and split
sleep conditions obtained significantly more total sleep
time than participants in the daytime sleep condition.
While there was no clear effect of sleep schedule on
performance measures, participants in the daytime
sleep condition reported higher levels of sleepiness
both within the work shift and across the simulated
working week compared to the other two conditions,
with no difference between the split and nighttime sleep
conditions. Results of the present laboratory study
suggest that split sleep schedules may be a good
alternative to a consolidated daytime sleep in industries
that allow for this kind of flexibility in their scheduling.
 10 M. L. Jackson et al.
Despite the daily sleep opportunity of 10 h, TST, a
determinant of recuperation (Banks et al., 2010;
Wesensten et al., 1999), differed across the three con-
ditions in a manner consistent with our knowledge of
the human circadian rhythm in sleep propensity, and
what is observed in shift workers in the real world
(Åkerstedt, 2003; Goel et al., 2011). In a healthy person
with a consolidated 8 h sleep opportunity at a circadian
phase with high sleep propensity, the normal sleep
latency is typically between 10 and 12 min. For a
consolidated sleep opportunity, the sleep latency
reflects only a small shortening of actual sleep relative
to sleep opportunity. With sleep split into two or more
bouts, one goes through the process of falling asleep at
least two times. As such, a split sleep opportunity could
degrade alertness and performance only in so far as it
might reduce TST in comparison to a consolidated sleep
opportunity (Wesensten et al., 1999). TST in the night-
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
time and split sleep conditions in the current study were
in the normal range of 7–9 h, compared to TST in the
daytime sleep condition, which was in the mildly sleep-
restricted range of 6–7 h. Similarly, REM sleep was
reduced in the daytime sleep condition, with no effect
on N3 sleep, consistent with previous field studies in
shift worker populations (Foret & Benoit, 1974). REM
sleep appears to be influenced by the circadian timing of
For personal use only.
the sleep period, which varied by condition, whereas N3
sleep, influenced by the homeostatic drive accumulated
from prior wakefulness, was equivalent across condi-
tions, as has been found in previous forced desynchrony
studies (Paech et al., 2012). Importantly for recovery
following shift work, the split sleep condition achieved
more TST on the recovery night compared to the
daytime sleep condition. This suggests that participants
in the daytime sleep condition experienced more circa-
dian adaptation, and thus found it difficult to obtain
adequate sleep when attempting to switch back to a
nighttime sleep schedule, which was not observed
following a 5-d split sleep schedule.
Our primary performance measure, PVT lapses, did
not differ between the split sleep conditions and the two
control sleep conditions, and PVT lapses were overall
relatively infrequent across work days. The split sleep
group did exhibit a higher number of lapses towards the
end of their night work period and at the start of their
morning work period. However, their performance
plateaued relative to the daytime sleep group, whom
exhibited increasing impairment across the work shift.
Note that the sleep opportunity was 10 h per day for all
days in all conditions. Thus, even for the participants in
the daytime sleep condition (who had slightly restricted
sleep), performance did not degrade significantly. In a
study with a similar degree of mild daily sleep restriction
over 7 d, only a small increase in PVT lapses was seen
(Belenky et al., 2003). In addition, the experimental
manipulation lasted 5 d in the current study; perhaps
not sufficient to increase PVT lapses. For the DSST,
there was evidence of improvement in performance
across all conditions over the course of the study,
representing the well-known practice effects of this task
(Van Dongen et al., 2003). Note these results must be
taken with caution, as by design, the nighttime sleep
condition had an extra DSST session prior to their
working week, and therefore had additional practice on
the task relative to the other conditions. Thus, perform-
ance was relatively unaffected by the split sleep sched-
ule relative to the two other sleep conditions of what
was an extended sleep opportunity, despite the clear
circadian effects on actual sleep. However, if the sleep
opportunity was restricted in the daytime sleep or split
sleep conditions, as occurs more frequently in the real
world, these results could be very different. Night shift
workers typically obtain between 5–6 h of sleep between
shifts (Foret & Benoit, 1974); 1–2 h less than the
participants in the current study. Whether adequate
sleep would be obtained under a split sleep schedule in
the real world is unknown, and if not, performance may
deteriorate to a greater extent than that observed in the
current study.
Subjective sleepiness on the KSS differed significantly
among the three conditions, with participants in the
daytime sleep condition reporting significantly more
sleepiness than participants in the nighttime or split
sleep conditions. As has been documented previously
(Santhi et al., 2007), there appeared to be a ‘‘first-night
effect’’ in the daytime sleep condition, with KSS scores
highest on the first work day. Subjective sleepiness was
also increased towards the end of the shift in the
daytime sleep condition, consistent with data in workers
on nightshift (Tilley et al., 1982). Overall KSS scores were
in the low to moderate sleepiness range (Åkerstedt &
Gillberg, 1990). In contrast to the KSS, there were no
main effects of condition on mood. However, partici-
pants in the split sleep condition reported lower mood
and less positive affect across their work shift compared
to the other conditions. This may have been due to
baseline differences in mood ratings between the three
groups. Given the link between circadian disruption and
mood disturbance, close examination of the effects of
split sleep schedules on mood should be considered in
future studies.
The findings of the current study are in line with
previous studies examining different types of split sleep
schedules (Mollicone et al., 2007, 2008). For example,
Mollicone et al. (2007) described the effect on total sleep
time of various combinations of nocturnal anchor sleep
and diurnal nap. They found that sleep efficiency,
performance on the PVT and subjective sleepiness
were all found to be primarily a function of total time
in bed regardless of how the sleep was divided between
nocturnal anchor sleep and diurnal nap (Mollicone
et al., 2007, 2008). The maintenance of performance
observed in the split sleep condition in the current study
may simply be due to the 10 h TIB in each 24 h sleep
opportunity provided and the amount of sleep obtained
by the participants, which was equivalent to the
Chronobiology International

nighttime sleep condition. An equal division of the sleep
opportunities, including a period of sleep placed at an
adverse circadian phase, did not have a significant effect
on performance, but may still be more beneficial for
subjective sleepiness than consolidated daytime sleep,
where individuals are awake for the whole night. The
within-session results of the current study support this;
sleepiness was relatively stable across the final daytime
test bouts within each work day in the split sleep
condition, whereas a dramatic increase in attentional
lapses and KSS scores towards the end of the shift was
observed in the daytime sleep condition. In the split
sleep condition, part of the work period was conducted
during the day (e.g. between 0900 h and 1400 h in the
current study). This potentially reduced the perform-
ance impairment that can occur when the work period is
placed at an adverse circadian time. The current study
design for the split sleep group effectively reduced time
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
awake/the length of the work period relative to the other
conditions. Thus, a relative difference between the
conditions may reflect the shortened wake opportunity
in the split sleep condition. This raises the possibility of
whether shift workers can get by with shorter sleep
durations, so long as the intervening wake period is
also short; thereby preventing the accumulation of
homeostatic pressure and concomitant performance
For personal use only.
deterioration.
Of further interest is the comparison of the two sleep
opportunities in the split sleep condition. Participants in
the split sleep condition obtained substantially more
sleep during the morning sleep opportunity (0300 h–
0800 h) when sleep propensity was higher than dur-
ing the afternoon/evening sleep opportunity (1500 h–
2000 h). This highlights the benefits of placing at least
some of the available sleep opportunity during periods
of high circadian sleep propensity. Both the duration
and timing of rest opportunities play a role in determin-
ing the amount and quality of sleep that shift workers
will obtain (Roach et al., 2003). Between shifts, night
shift workers often have truncated diurnal sleep periods
(Pilcher et al., 2000), presumably due to circadian drive
for wakefulness occurring towards the end of the sleep
period. With regards to work scheduling, there is a
strong emphasis on number of work hours, shift length
and types and speed of shift rotation, but, there is less
focus on timing of rest opportunities. Scheduled rest
breaks that do not take into account time of day
influences on sleep may not allow shift workers suffi-
cient sleep between shifts, regardless of the minimum
break requirement and opportunity to recover. Overall,
the findings of the current study have direct implica-
tions for rest break policy, for example, sleeper berth use
in the trucking industry and rest breaks in aviation,
mining and health care industries. Hours of service
regulations could be adapted to allow for greater
flexibility of the timing and duration of sleep, and also
be more directed by the individual. Field studies of
experienced nightshift workers have demonstrated that
! Informa Healthcare USA, Inc.
Efficacy of split sleep schedule on sleep and performance 11
naps during the shift, although taken opportunistically,
often occur during times when sleep propensity is high
(Mollicone et al., 2008), suggesting that workers have a
good understanding of when they can maximize their
sleep opportunity. Further research is needed to inves-
tigate whether the results of the current study are
replicable under real world circumstances, and the
efficacy of different split sleep schedules.
There are some limitations that should be noted.
First, for participants in the nighttime consolidated
sleep condition and the split sleep condition, the stay in
laboratory ended on Day 10, while participants in the
daytime sleep condition were from a longer 16-d study
(Van Dongen et al., 2011). All groups knew how long
they would be in the laboratory and knew their assigned
sleep condition after the first baseline night. The
prospect of additional time in the laboratory could
have affected the participants in the daytime sleep
condition, and perhaps this anticipation accounted for
the changes observed. However, the findings with
respect to total sleep time fit well with our knowledge
of the effect of circadian rhythms on sleep propensity,
thus making it less likely that they resulted from an
anticipatory effect. Further, with the exception of
differences in sleepiness (which were in accord with
the differences in total sleep time), there were no
differences among the conditions on the other subject-
ive measures, one of which (the PANAS) included
ratings of anxiety and another (the VAS-M) assessed
depression. Thus, it seems reasonable to take the
findings of the present study as a direct effect of the
sleep condition manipulation (nighttime sleep, split
sleep or daytime sleep) rather than any differences in
the length of stay in the laboratory. The participants in
this study were all young healthy males who were not
shift workers. Given that it would be unlikely that shift
workers in the real world would have 10-h opportunity
for sleep on a regular basis, the results may be different
in a chronically sleep restricted group. Similarly, in
practice, a split schedule such as the one examined in
the current study would require a second shift to work
the opposite schedule (i.e. 0800 h–1500 h and 2000 h–
0300 h) so that 24-h coverage can be achieved. The
practicality and efficacy of this alternate schedule
requires further examination.
Shiftwork is common in today’s society, and is
associated with adverse health and safety outcomes as
a result of poor sleep and impaired performance. It is
important to examine alternative options for scheduling
that may reduce this burden. Further research in shift
work populations and in real-world situations is war-
ranted to fully assess the efficacy of split sleep
schedules.
ACKNOWLEDGEMENTS
We thank the peer review committee of the project for
their valuable comments and suggestions: Janet M.
 12 M. L. Jackson et al.
Mullington, PhD, Goran Kecklund, PhD, and Nancy J.
Wesensten, PhD (Chair). We thank Hans P.A. Van
Dongen, PhD, for help with study design and for
making available data and experimental materials of
the study ‘‘Duration of Restart Period Needed to Recycle
with Optimal Performance: Phase II’’ from which the
data for the daytime sleep condition in this study were
drawn. We also wish to thank the staff and students at
Sleep and Performance Research Center, Washington
State University, Spokane who assisted with data
collection.
DECLARATION OF INTEREST
This was not an industry supported study. Dr Belenky
has received research funding from the Unites States
Department of Transportation Federal Motor Carrier
Safety Administration, Continental Airlines and United
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
Airlines. The other authors have reported no conflicts of
interest. This study was supported by the Federal Motor
Carrier Safety Administration.
REFERENCES
Åkerstedt T, Gillberg M. (1990). Subjective and objective sleepiness
For personal use only.
in the active individual. Int J Neurosci. 52:29–37.
Åkerstedt T. (2003). Shift work and disturbed sleep/wakefulness.
Occup Med. 53:89–94.
Banks S, Van Dongen HPA, Maislin G, Dinges DF. (2010).
Neurobehavioral dynamics following chronic sleep restriction:
Dose-response effects of one night for recovery. Sleep. 33:
1013–26.
Belenky G, Wesensten NJ, Thorne DR, et al. (2003). Patterns of
performance degradation and restoration during sleep restric-
tion and subsequent recovery: A sleep dose-response study.
J Sleep Res. 12:1–12.
Belenky G, Hursh SR, Fitzpatrick J, Van Dongen HPA. (2008). Split
sleeper berth use and driver performance: A review of the
literature and application of a mathematical model predicting
performance from sleep/wake history and circadian phase.
Spokane, WA: Washington State University.
Belenky G, Wu LJ, Jackson ML. (2011). Occupational sleep medi-
cine. Practice and promise. In Van Dongen HPA, Kerkhof GA,
eds. Progress in brain research. Oxford, UK: Elsevier BV,
pp. 189–203.
Bonnet MH, Arand DL. (2003). Clinical effects of sleep fragmen-
tation versus sleep deprivation. Sleep Med Rev. 4:297–310.
Bonnet MH. (1991). The effect of varying prophylactic naps on
performance, alertness and mood throughout a 52-hour con-
tinuous operation. Sleep. 14:307–15.
Dijk D-J, Czeisler CA. (1994). Paradoxical timing of the circadian
rhythm of sleep propensity serves to consolidate sleep and
wakefulness in humans. Neurosci Lett. 166:63–8.
Dinges DF, Orne MT, Whitehouse WG, Orne EC. (1987). Temporal
placement of a nap for alertness: Contributions of circadian
phase and prior wakefulness. Sleep. 10:313–29.
Dorrian J, Rogers NL, Dinges DF. (2005). Psychomotor vigilance
performance: Neurocognitive assay sensitive to sleep loss.
In Kushida C, ed. Sleep deprivation. Clinical issues,
Pharmacology, and Sleep Loss Effects. New York: Marcel
Dekker, pp. 39–70.
Drake CL, Wright KP. (2011). Shift work, shift work disorder,
and jet lag. In Kryger R, Dement W, eds. Principles and
practice of sleep medicine. 5th ed. St. Louis, MO: Elsevier, pp.
784–98.
Eriksen CA, Gillberg M, Vestergren P. (2006). Sleepiness and sleep
in a Simulated ‘‘Six Hours On/Six Hours Off’’ sea watch system.
Chronobiol Int. 23:1193–202.
Ficca G, Axelsson J, Mollicone DJ, et al. (2010). Naps, cognition and
performance. Sleep Med Rev. 14:249–58.
Folkard S, Tucker P. (2003). Shift work, safety and productivity.
Occup Med. 53:95–101.
Foret J, Benoit O. (1974). Sleep patterns of workers on rotating
shifts. Electroencephalogr Clin Neurophysiol. 37:377–44.
Goel N, Van Dongen HPA, Dinges DF. (2011). Circadian rhythms in
sleepiness, alertness, and performance. In Kryger R, Dement W,
eds. Principles and practice of sleep medicine. 5th ed. St. Louis,
MO: Elsevier, pp. 445–55.
Horne J, Reyner L. (1999). Vehicle accidents related to sleep:
A review. Occup Environ Med. 56:289–94.
Iber CS, Ancoli-Israel S. (2007). The AASM manual for the scoring
of sleep and associated events: Rules, terminology and tech-
nical specifications. Westchester, IL: American Academy of
Sleep Medicine.
Jones C, Dorrian J, Jay S, et al. (2006). Self-awareness of impair-
ment and the decision to drive after an extended period of
wakefulness. Chronobiol Int. 23:1253–63.
Lim J, Dinges DF. (2008). Sleep deprivation and vigilant attention.
Ann NY Acad Sci. 1129:305–22.
McDonald J, Potyk D, Fischer D, et al. (2013). Napping on the night
shift: A naturalistic study of sleep, performance, and learning in
physicians-in-training. J Grad Med Educ. 5:634–8.
Mollicone DJ, Van Dongen HPA, Dinges DF. (2007). Optimizing
sleep wake schedules in space: Sleep during chronic nocturnal
sleep restriction with and without diurnal naps. Acta Astronaut.
60:354–61.
Mollicone DJ, Van Dongen HPA, Rogers NL, Dinges DF. (2008).
Response surface mapping of neurobehavioral performance:
Testing the feasibility of split sleep schedules for space
operations. Acta Astronaut. 63:833–40.
Nicholson AN, Pascoe PA, Roehrs T, et al. (1985). Sustained
performance with short evening and morning sleeps. Aviat
Space Environ Med. 56:105–14.
Paech GM, Ferguson SA, Sargent C, et al. (2012). The relative
contributions of the homeostatic and circadian processes to
sleep regulation under conditions of severe sleep restriction.
Sleep 35:941–8.
Pilcher JJ, Lambert BJ, Huffcutt AI. (2000). Differential effects
of permanent and rotating shifts on self-report sleep length:
A meta-analytic review. Sleep. 23:155–63.
Portaluppi F, Smolensky MH, Touitou Y. (2010). Ethics and
methods for biological rhythm research on animals and
human beings. Chronobiol Int. 27:1911–29.
Rajaratnam SM, Arendt J. (2001). Health in a 24-h society. Lancet.
358:999–1005.
Roach GD, Reid KJ, Dawson D. (2003). The amount of sleep
obtained by locomotive engineers: Effects of break duration and
time of break onset. Occup Environ Med. 60:e17. doi:10.1136/
oem.60.12.e17.
Sallinen M, Harma M, Akerstedt T, et al. (1998). Promoting alertness
with a short nap during a night shift. J Sleep Res. 7:240–7.
Santhi N, Horowitz TS, Duffy JF, Czeisler CA. (2007). Acute sleep
deprivation and circadian misalignment associated with tran-
sition onto the first night of work impairs visual selective
attention. PLoS One. 2:e1233.
Schweitzer PK, Randazzo AC, Stone K, et al. (2006). Laboratory and
field studies of naps and caffeine as practical countermeasures
for sleep-wake problems associated with night work. Sleep. 29:
39–50.
Takeyama H, Kubo T, Itani T. (2005). The nighttime nap strategies
for improving night shift work in workplace. Ind Health. 43:
24–9.
Chronobiology International

Tilley AJ, Wilkinson RT, Warren PS, et al. (1982). The sleep and
performance of shift workers. Hum Factors. 24:629–41.
Van Dongen HPA, Maislin G, Mullington JM, Dinges DF. (2003).
The cumulative cost of additional wakefulness: Dose-response
effects on neurobehavioral functions and sleep physiology from
chronic sleep restriction and total sleep deprivation. Sleep. 26:
117–26.
Van Dongen HPA, Baynard M, Maislan G, Dinges D. (2004).
Systematic interindividual differences in neurobehavioral
impairment from sleep loss: Evidence of trait-like differential
vulnerability. Sleep. 27:423–33.
Van Dongen HPA, Jackson ML, Belenky G. (2010). Duration of
restart period needed to recycle with optimal performance:
Phase II. Technical Report No. FMCSA- RRR-10-062.
Washington, DC: Federal Motor Carrier Safety Administration.
Chronobiol Int Downloaded from informahealthcare.com by Dicle Univ. on 11/12/14
For personal use only.
! Informa Healthcare USA, Inc.
Efficacy of split sleep schedule on sleep and performance 13
Van Dongen HPA, Belenky G, Vila BJ. (2011). The efficacy of a
restart break for recycling with optimal performance depends
critically on circadian timing. Sleep. 34:917–29.
Watson D, Clark LA, Tellegen A. (1988). Development and valid-
ation of brief measures of positive and negative affect: The
PANAS scales. J Personality Social Psychol. 54:1063–70.
Wechsler D. (1997). WAIS-III administration and scoring manual.
San Antonio, TX: The Psychological Corporation.
Wehrens SMT, Hampton SM, Kerkhofs M, Skene DJ. (2012). Mood,
alertness, and performance in response to sleep deprivation
and recovery sleep in experienced shiftworkers versus non-
shiftworkers. Chronobiol Int. 29:537–48.
Wesensten NJ, Balkin T, Belenky G. (1999). Does sleep deprivation
impact recuperation? A review and reanalysis. J Sleep Res.
8:237–45.