Content uploaded by Remy Schmitz
Author content
All content in this area was uploaded by Remy Schmitz on Mar 06, 2018
Content may be subject to copyright.
Content uploaded by Remy Schmitz
Author content
All content in this area was uploaded by Remy Schmitz on Jan 02, 2018
Content may be subject to copyright.
© Sleep Research Society 2017. Published by Oxford University Press on behalf of the Sleep Research Society.
All rights reserved. For permissions, please e-mail journals.permissions@oup.com.
1
Submitted: 25 November, 2016; Revised: 17 September, 2017
O A
Sleep Deprivation Triggers Cognitive Control
Impairments in Task-Goal Switching
HichemSlama, PhD1,2, Daphne OliviaChylinski, MS1, GaétaneDeliens, PhD1,3,
RachelLeproult, PhD1, RémySchmitz, PhD1, PhilippePeigneux, PhD1
1UR2NF – Neuropsychology and Functional Neuroimaging Research Group at CRCN – Center for Research in
Cognition and Neurosciences and UNI – ULB Neurosciences Institute, Université Libre de Bruxelles (ULB),
Brussels, Belgium; 2UNESCOG – Research Unit in Cognitive Neurosciences at CRCN – Center for Research in
Cognition and Neurosciences and UNI – ULB Neurosciences Institute, Université Libre de Bruxelles (ULB),
Brussels, Belgium; 3ACTE – Autism in Context: Theory and Experience/Langage & Esprit, Université libre de
Bruxelles (ULB), Brussels, Belgium
Work Performed: Université Libre de Bruxelles (ULB)
Corresponding Author: Hichem Slama, PhD, Center for Research in Cognition & Neurosciences, ULB, Campus du Solbosch, Avenue F.D. Roosevelt 50, CP
151, 1050 Bruxelles, Belgium. Telephone: 32 (0)2 650 26 40; Fax: 32(0)2 650 22 09; E-mail: Hichem.Slama@ulb.ac.be.
Abstract
Study Objectives: This study investigates the impact of sleep deprivation (SD) on task-goal switching, a key component of
cognitive exibility.
Methods: Task-goal switching performance was tested after one night of regular sleep (n=17 participants) or of total SD
(n=18). To understand the relationships between task-switching performance and other cognitive processes following SD,
participants were tested for other key attentional (alertness and vigilance) and executive (inhibition and working memory)
functions. Spontaneous eye blink rate (EBR) was also measured as an indirect marker of striatal dopaminergic function.
Results: SD negatively affects task-goal switching as well as attentional and inhibition measures, but not working memory.
Changes in task-goal switching performance were not signicantly correlated with changes in objective and subjective
markers of fatigue and sleepiness, response inhibition, or spontaneous EBR.
Conclusions: Altogether, our results show differentiated effects of SD on key executive functions such as working memory,
inhibition, and task-goal switching.
Key words: sleep deprivation; task switching; cognitive exibility; cognitive control; dopamine; eye blink
Statement of Signicance
This study is the rst to evidence impact of sleep deprivation on task-goal switching, a key component of cognitive
exibility and cognitive control. Moreover, alterations in task-goal switching were not associated with changes in
spontaneous eye blink rate as well as attentional and executive functions. These results are of signicance because
sleep debt is more and more present in modern society, and cognitive exibility subtends rapid adaptation to constantly
changing situations or environments, a condition usually encountered in everyday life.
17September2017
SLEEPJ, 2018, 1–12
doi: 10.1093/sleep/zsx200
Advance Access Publication Date: 13 December 2017
Original Article
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018
Introduction
Cognitive control (also called executive functions or executive
control [1]) refers to the mental ability to regulate thoughts and
actions in accordance with internally represented goals [1–3].
Cognitive control allows individuals to dene a goal, choose a
strategy to achieve it, and monitor its execution [1]. Specically,
cognitive control enables an adaptation to novel situations
when overlearned routines are insufcient. Cognitive exibility
is a crucial component of cognitive control that subtends rapid
adaptation to constantly changing situations or environments,
a condition usually encountered in everyday life [4].
Cognitive exibility is usually investigated in laboratory con-
ditions using task-switching paradigms (for reviews, see Refs. 5
and [6]) in which participants repeatedly switch between two
tasks or more. For instance, participants can be presented with
colored shapes and instructed to decide between colors (red vs.
blue) or shapes (circle vs. square) according to a cue presented
previously. Switching between tasks leads to more errors and
increased processing time compared with performing the same
task [6–8]. The additional time to respond (or higher error rate)
in a shifting task condition is called a switch cost. It is usually
computed by subtracting the mean reaction time (or error rate)
for repeated conditions from the reaction time (or error rate) for
switch conditions.
Importantly, task switching in itself is not a unitary pro-
cess. It is rather a complex mechanism that involves several
components. Imagine you are photocopying different docu-
ments. Some of them are printed double-sided but the others
are single-sided and need to be printed onto a single sheet of
paper. In addition, consider that the documents are mixed in
a xed order, which you cannot change (i.e., you cannot pho-
tocopy the double-sided documents rst and then the others,
meaning you have to switch between conditions). Succeeding
in this situation need at least two distinct sets of operations.
First, for each document, you have to keep in mind whether
you are photocopying a double-sided document. In the task-
switching literature, this stage has been labeled “task-goal
activation.” [9, 10] Second, you have to program the settings of
the photocopier to a “duplex printing” or a “convert to duplex
printing” mode, according to the task goal. This stage has been
labeled “task-rule activation” and requires retrieving the rules
that allow achieving the goal (i.e., the photocopier instruction
manual in this example) in long-term memory [9, 10]. In sum,
for any given task, preparation requires setting a goal (“what
to do?”; i.e., task-goal activation) followed by the activation of
the rules (“how to do the task?”; i.e., task-rule activation). The
task-rule activation stage depends on the complexity of the
instructions. Indeed, in the example mentioned above, the
task-rule activation depends on the complexity of the photo-
copier instruction manual. In simple experimental tasks like
color and shape discrimination, the task-goal activation is
the task to perform (e.g., color discrimination), and the task-
rule activation stage is most often achieved by activating the
necessary stimulus–response (S–R) mappings (i.e., the S–R
instructions) to execute the task [9, 10]. For instance, a produc-
tion rule for response selection in a color discrimination task
might have the following form [10]:
“IF ((GOAL IS TO DO COLOUR-DISCRIMINATION TASK) AND
(STIMULUS COLOUR IS RED)) THEN (PRESS RIGHT INDEX-FINGER
KEY).”
As can be seen from this instruction set, correct task goal
activation is needed to adequately implement the task rule.
If you experiment problems or fail in one of these two stages,
you will be slower or commit errors, particularly when you
have to switch between tasks. However, you could fail because
you struggle to retrieve the goal but apply the correct rule, or
because you retrieve the correct goal but struggle to apply the
rule. This is why the distinction between task-goal and task-
rule is essential in the task switching literature. In a switch-
ing situation, task-goal switching refers to the rst component,
whereas task-rule switching refers to the second one. Rubinstein
and colleagues [10] proposed that, during task-goal switching,
the current goal is inserted into declarative working memory
(WM), and the previous goal is deleted. They also suggested
that, during task-rule switching, S–R rules for the current task
are loaded into procedural WM. Therefore, a way to investigate
the dissociation between task-goal and task-rule switching is
to compare tasks involving a strong S–R mapping (memory)
load and tasks involving a weak load or an absent one. For
instance, Ravizza and colleagues [11, 12] conducted several
experiments investigating the impact of S–R rules on task
switching. They used odd-man-out test situations (i.e., par-
ticipants had to nd the stimulus that did not match the oth-
ers). In the weak S–R load condition, responses were spatially
congruent to the target stimulus location, that is, participants
did not need to learn arbitrary S–R rules. On the contrary, in
the strong S–R load condition, participants were required to
respond according to an arbitrary rule that had previously
been memorized. Because arbitrary S–R mapping involves the
learning of a response code (i.e., a rule) that often does not
match the target dimension (e.g., red is arbitrarily associated
with the left response key, blue with the right response key),
switching conditions particularly rely on task-rule switching
[5]. In addition, arbitrary S–R mapping gives rise to poten-
tial sources of interference as both learning and keeping or
retrieving the arbitrary rule in long-term memory are needed
to perform the task [13]. By contrast, in nonarbitrary mapping
conditions, the response is naturally associated with the tar-
get’s features, which decreases the impact of task-rules and
emphasizes task-goal activation.
Besides S–R mappings, the type of cues indicating the
task to perform is another element that inuences task-rule
switching. Depending on its nature, the cue itself can repre-
sent an additional rule to learn [5, 6, 14]. Acue can be strong
(i.e., transparent, e.g., the word “color” used to signal a color
judgment task) or weak (i.e., arbitrary, e.g., a triangle indicates
the color task and a diamond indicates the shape task). Weak
cues increase the involvement of task-rule activation because
the information provided by the cue is not sufcient in itself
to dene the task to perform, and participants must retrieve
the meaning of the cue (i.e., the rule) in long-term memory in
addition to the main task [5, 6]. Therefore, the use of transpar-
ent cues also emphasizes task-goal activation and decreases
the impact of task-rules. As expected, arbitrary (weak) cues
give rise to higher switch costs than transparent (strong) ones
[5, 14, 15].
The functional dissociation between task-goal and task-
rule switching has been evidenced using functional neu-
roimagery and latent factor analyses [16–18]. Task-goal and
task-rule switching have been shown to rely on distinct neu-
ral processes and substrates. Event-related potentials studies
2 | SLEEPJ, 2018, Vol. 41, No. 2
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018
identied an early parietal and frontal positivity associated
with task-goal activation, and late parietal positivity and
frontal negativity associated with task-rule activation [16].
Functional magnetic resonance imaging studies also found
that left anterior regions are differentially involved in task-
goal and task-rule representations with the ventrolateral
prefrontal cortex (PFC) involved in abstract rules representa-
tion, the presupplementary motor area (pre-SMA) involved
in rules suppression, and the inferior frontal junction (IFJ)
involved in task-goal representation. Posterior regions [i.e.,
the posterior parietal cortex (PPC) and the intraparietal sul-
cus (IPS)] are more involved in S–R rules and response-sets
representations [16, 17]. In a recent study [18] using latent fac-
tor analyses, 20 task pairs were administered to 119 young
adults to assess ve proposed components of mental set
shifting. Task-goal switching was labeled Judgment shifting and
required participants to switch between varying classication
tasks. For example, participants had to determine either the
color or the shape of objects presented. Task-rule switching
was labeled mapping shifting, and participants had to switch
between S–R mappings. Three other components, which we
will not manipulate in this study, were also assessed, namely,
dimension shifting, response set shifting, and stimulus set
shifting. Modeling latent factors for each of the components
revealed that a model with ve separate yet correlated factors
ts the data best. Importantly, task-goal switching was more
consistently associated with a separate factor than task-rule
switching and could not fully be accounted for by a general
shifting factor, conrming the importance of dissociating
these twostages.
Sleep deprivation (SD) is well known to exert a deleterious
impact on various cognitive domains (e.g., Refs. 19–23), but only
a few studies investigated its impact on cognitive exibility
[24–27]. These studies found an increased switch cost after a
night of total SD. However, several executive functions were
mixed in one study (i.e., response inhibition, task switching,
and task strategy [27]), which results in difcult comparisons
and interpretations. The three other studies used task-switch-
ing paradigms that involved arbitrary S–R mapping (e.g., left
button for red or circle shape). Therefore, increased SD-related
switch costs reported in prior studies [24–26] might be due to (or
aggravated by) SD-related impairments in memory load capaci-
ties eventually hampering cognitive exibility. In addition,
weak cues were used in all SD studies on task switching also
increasing the memory load [24–26]. In sum, these studies argu-
ably evidenced task-rule switching decits after SD. However,
it remains disputable whether task-goal switching in itself is
impaired after SD, or if higher switch costs after SD are the con-
sequence of SD-related difculties to switch the rules during
task performance.
To the best of our knowledge, the specic impact of SD on
the task-goal component remains unexplored. This information
is of importance because task-goal activation and maintenance
are key features of most of the theoretical frameworks on cog-
nitive control [3, 28, 29]. During task switching, the goal of the
task is generally a more transient representation than the rule,
the latter being usually xed at the beginning of the experiment
during instructions. As discussed above, task-goal and task-rule
switching processes rely on distinct neural substrates [16, 17].
Likewise, SD does not similarly affect neural activity in all brain
regions [30, 31]. It is therefore possible that SD does not affect
to the same extent task-goal and task-rule switching processes.
In a prior study [32], we measured task-goal switching using a
cued match-to-sample task in which the response mapping was
congruent with the target location (i.e., nonarbitrary mapping)
and the cues were words indicating the task to perform (i.e.,
transparent cues). Results disclosed improved accuracy switch-
cost scores after a short nap, indicating an effect of sleep on
task-goal switching. These ndings also suggest that SD may
have an opposite effect and deteriorate task-goal switching and
consecutive accuracy switch-cost scores compared with regular
sleep(RS).
Finally, cognitive control abilities are tightly related to cen-
tral dopaminergic activity. Striatal dopamine is thought to
operate as a gating signal that triggers the updating of WM and
increases cognitive instability or exibility [28, 29]. According
to the work of Dreisbach etal [33]., dopamine plays a central
role in the stability–exibility dilemma. In other words, to fol-
low a goal-directed behavior, a compromise should be reached
between maintaining the current goal (i.e., keeping away from
distraction) and updating information (i.e., adapting our behav-
ior). Spontaneous eye blink rate (EBR) has been described as an
indirect marker of central dopaminergic function [34] linked
to cognitive exibility [33, 35–37]. For instance, healthy people
with higher EBR were shown to exhibit increased cognitive ex-
ibility but also reduced cognitive stability [33, 36]. Interestingly,
spontaneous EBR increases after SD and positively correlates
with sleepiness, which has been interpreted as increased cen-
tral dopamine activity to counteract the sleep drive [38, 39]. EBR
is also a marker of drowsiness or sleepiness [40] and arousal
levels [41]. However, recent results [42] have questioned the
plausibility of a dopamine increase after SD. Indeed, SD has
been found to downregulate D2 and D3 receptors. Because
spontaneous EBR primarily relates to cognitive function via
D2-driven modulation [37], one should thus expect an EBR
decrease after SD, which is not the case [38, 39]. Furthermore,
SD does not affect the impact of methylphenidate, a DA trans-
porter blocker, on D2/D3 receptors [42], indicating that a dopa-
mine increase after SD is also rather improbable. Therefore,
if the EBR increase observed after SD is not due to dopamine
increase, it should not be associated with a task-goal switching
modulation afterSD.
In this framework, the present study investigates the impact
of one night of total SD on cognitive exibility using a task-
goal switching paradigm with nonarbitrary mapping and
strong (transparent) cue conditions that minimize task-rule
cognitive load. We predicted that SD would aggravate switch
costs, indicating a task-goal switching alteration. We also pre-
dicted a spontaneous EBR increase after SD. However, as noted
above, this increase might not be associated with the task-goal
switching alteration. Importantly, many studies have provided
evidence for a deleterious impact of SD on alertness, vigi-
lance, WM, and other mental abilities [19–23]. In this respect,
SD-related impairments in cognitive exibility might be an
indirect consequence of the deterioration of these other cogni-
tive functions. Therefore, we tested participants for other key
attentional (alertness and vigilance) and executive (inhibition
and WM) functions [4, 43].
Slama etal. | 3
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018
Material and Methods
Participants
Thirty-eight French-speaking participants gave their written
informed consent to participate in this study approved by the
Ethics Committee of the Faculty of Psychological Sciences at the
Université libre de Bruxelles (ULB). Data for three of them were
discarded due to a technical failure during the switching task.
The remaining 35 participants (19 females) had a mean age of
21.94±2.52years old (range=18–26years). All participants were
right-handed, had no history of medical, neurological, or psychi-
atric disorders, were free of any medication or drug, and with-
out depression signs (13-item Beck Depression Inventory [44];
score range=0–5; cutoff score=8). Participants’ chronotype was
neutral (n=29), moderate evening (n= 4), or moderate morn-
ing (n = 2) types according to the Morningness–Eveningness
Questionnaire [45] (score range= 35–63). Habitual fatigue level
was below the cutoff score on the Fatigue Severity Scale [46]
(FSS; score range=1.33–5.11; cutoff score=5.5). There were four
smokers in the SD group and two smokers in the RS group. All
smokers smoked less than 10 cigarettes perday.
Participants were required to maintain a RS-wake cycle for
3days prior to the study (i.e., sleep at least 7 hours per night, go
to bed before 01:00 am, wake up before 10.30 am, and no naps
during the day). Compliance was controlled using actigraphic
recordings (ActiGraph wGT3X-BT Monitor, Pensacola, FL).
Participants were also asked to complete the St. Mary Hospital
questionnaire [47] and a sleep diary after each night of sleep.
Procedure
Participants were assigned to one out of two conditions. They
either slept at home between the evening and morning testing
sessions (RS; 17 participants) or stayed awake in the laboratory
during the entire night (SD; 18 participants). For organizational
reasons, participants knew before the testing session in which
condition they were included. No difference was found between
the two groups for all the above-mentioned inclusion criteria
(two-tailed t-tests for independent samples; all p-values >.26),
except for a trend [t(33)=−2.01; p=.052] on habitual fatigue lev-
els (FSS scores), on average higher in the SD (3.21±1.01) than in
the RS (2.56±0.91) group. However, all scores remained below
cutoff pathological values.
As illustrated in Figure1, both SD and RS groups were rst
tested in the evening starting at 18:00. Participants completed
the following tasks, always in the same order: psychomotor vig-
ilance task (PVT), subjective sleepiness (Karolinska Sleepiness
Scale, KSS; Visual Analogue Scales for Sleepiness, VAS-S) and
fatigue (Visual Analogue Scales for Fatigue, VAS-F) scales, WM
(N-back), and response inhibition (Stop-signal) executive tasks
(see below for a description of these tasks and scales). At 19:00,
spontaneous EBRs were recorded for 3 minutes followed by the
task-switching protocol. Afterward, the RS group went home for
a regular night of sleep, whereas the SD group stayed in the test-
ing room during the whole night under the constant supervision
of two experimenters. During the SD period, participants were
asked to remain quiet and seated most of the time. They were
allowed to engage in calm activities (e.g., reading, playing soci-
ety games, and watching movies). Water was available ad libi-
tum, and isocaloric meals were offered hourly. Participants had
to refrain from stimulant drinks (e.g., coffee, tea, or energizers)
or smoking. The PVT and the KSS were administered hourly to
document changes in objective and subjective vigilance levels,
respectively, over the night. At 09:00 am, the second testing ses-
sion started with the exact same protocol administered on the
previous evening. Participants assigned to the RS group came
to the laboratory in the morning at 09:00 for the second testing
session.
Fatigue, Sleepiness, Alertness and Vigilance, andEBR
At the beginning of each session, participants’ subjective level
of sleepiness and fatigue were assessed using the KSS [48] and
VAS-F/S [49].
Objective measures of alertness and vigilance were obtained
using the 10-minute version of the PVT [50]. In the PVT, par-
ticipants were instructed to press a key as fast as possible
whenever a millisecond countdown appeared in the middle of
a computer screen. Stimuli were randomly presented with an
inter-stimuli interval ranging from 2 to 10 seconds. Alertness
or PVT speed was estimated using the median reaction times
(RTs), less affected by extreme values than the mean [25, 51].
Figure1. Schematic representation of the experimental protocol (evening, night, and morning sessions). KSS=Karolinska Sleepiness Scale; VAS-F/S=Visual Analogue
Scales for Fatigue/Sleepiness; PVT=Psychomotor Vigilance Task; RS=regular Sleep; SD=sleep deprivation; EBR=spontaneous eye blink rate recording.
4 | SLEEPJ, 2018, Vol. 41, No. 2
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018
PVT variability as a marker of vigilance [52] was measured using
the RTs’ coefcient of variation, i.e., standard deviation divided
by mean [53].
To record spontaneous EBR, participants were seated in a
separate room 1 m away from a wall with a xation cross at
the level of the eyes. They were instructed to relax, sit comfort-
ably, and xate the cross on the wall. They were allowed to move
slightly and to blink but not to turn their head to the right or left.
This information helped diverting participants’ attention from
their blinks. Eye movements and blinks were video-recorded
for 3 minutes. Participants were seated in front of the xation
cross during the instructions, and the rst 10 seconds of record-
ing were not analyzed to allow for adaptation to the environ-
ment. Video-recordings were visually inspected a posteriori to
count the number of eye blinks. Ablink was dened as a rapid
closing and opening of the eyelid: slow eye closures were not
counted as blinks. Two independent evaluators counted the
EBR, and results were compared. In case of disagreement, the
video was recounted for a third time and an agreement decided.
Spontaneous EBR was computed as the mean EBR per minute
during each recording session and was entered in statistical
analyses like in prior studies [54, 55].
WM and InhibitionTasks
WM N-BackTask
The WM N-back task [56] was adapted from a previous study
from our lab [57]. In the 0-back (N0) condition, they had to press
the response key only when the digit « 2 » appeared on the
screen, i.e., a simple detection task. In the 2-back (N2) condi-
tion, they had to press the space bar when the digit currently
displayed matched the digit presented two steps earlier in the
sequence. Therefore, the task required for successful compari-
son both the maintenance and the updating of a series of items
in WM. Participants performed ve blocks of each condition
with alternation of 0- and 2-back conditions (see Supplementary
Material for a detailed description of the task procedure). RTs
and corrected accuracy scores (hits minus false detections) were
averaged over the ve blocks per session for the N2 and N0 con-
ditions. WM performance reecting the updating process (i.e.,
difference between N2 and N0 conditions) was computed on
corrected accuracy scores (WM accuracy= accuracy N0– accu-
racy N2) and on mean RTs for correct responses (WM speed=RT
N2– RT N0). Lower scores reect smaller differences between N2
and N0 conditions, indicating a better performance in the WM
updating process.
Inhibition Stop SignalTask
The Stop Signal task evaluating response inhibition abili-
ties was adapted from a previous study from our lab [58] (see
Supplementary Material for a detailed description). Participants
were presented with arrows pointing to the left or to the right
(go signals) at random intervals. They were instructed to answer
as quickly and as accurately as possible to which side the arrow
was pointing by pressing the corresponding key. In 20% of the
trials, a stop signal (vertical arrow) was displayed shortly after
the horizontal arrow (see Supplementary Figure S1). In this case,
participants were instructed to try not responding, i.e., inhibiting
the response initiated after presentation of the go signal. If the
participant managed successfully to refrain from responding,
then the stop signal delay (SSD) was lengthened in the next stop
trial, resulting in a harder response inhibition. If he or she failed
(i.e., he or she provided an answer even though the stop signal
instructed not to), the SSD on the next stop trial was shortened
to facilitate the response inhibition. This SSD adaptive method
guarantees a similar task difculty across participants who
work at the edge of their inhibitory capacities. The stop signal
reaction time (SSRT) was calculated for both morning and even-
ing sessions. The SSRT is the maximal time delay between the
go and the stop signals at which the participant still successfully
inhibit the initiated response (SSRT=mean RT – SSD). SSRT pro-
vides an individual index of inhibitory control, with longer SSRT
indicating poorer response inhibition [59].
Task-Goal Switching
Task-goal switching was assessed using a cued match-to-
sample task from our lab [32]. The task design is illustrated in
Figure2. At the beginning of each trial, an instruction cue was
displayed in the center of the screen for 250 milliseconds, fol-
lowed by a short (0 milliseconds) or a long (1800 milliseconds)
cue-stimulus interval (CSI). Participants had to then match the
stimulus presented in the upper part of the screen with one of
the two stimuli presented in the lower part of the screen accord-
ing to the instruction cue. Responses were always spatially con-
gruent to the target location, i.e., left key for left answer and
right key for right answer.
The task-switching protocol was programmed using
Psyscope X software [60, 61] and run on Mac Mini computers.
Participants answered on the computer’s keyboard using their
right and left hand forengers. RT was recorded for each trial
as the time elapsed from target onset to response (in millisec-
onds). The cue was a written word of the relevant dimension (in
French): COULEUR (color), FORME (shape), NOMBRE (number), or
CONTOUR (outline). For each trial, the three stimuli were pre-
sented on a black screen, each within a rectangle [9 cm wide
Figure2. Cued match-to-sample switching task. For each trial, a xation cross
is followed by an instruction cue (“shape,” “color,” “number,” or “outline”), after
which three gures are presented. Participants are asked to decide which one
of the two lower gures matches the top one, depending on the instruction cue,
and to press the corresponding key (left key for left gure and right key for right
gure).
Slama etal. | 5
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018
(visual angle= 8.58°) and 5 cm tall (4.77°)] in a triangular dis-
position, with the target picture on the top (Figure2). The geo-
metrical gures were 2.5cm wide (2.39°) and 1.5cm tall (1.43°).
The distance between the screen and the participants’ eyes was
approximately 60cm.
Prior to the task, participants received written instruc-
tions that were repeated by the investigator before the session
started. They were instructed to press a key (“q” for left and “m”
for right on an azerty keyboard) corresponding to the position
(left or right, respectively) of the bottom picture that matched
the target (top picture) according to the cued dimension. Asam-
ple trial gure was included as an example. The participants
were instructed to perform as fast as possible while minimiz-
ing the number of errors. Each trial began with a xation cross
displayed at the center of the screen during 450 milliseconds,
followed by the cue presented for 250 milliseconds. Short and
long CSI conditions were presented in a counterbalanced order
across participants. Beside warm-up trials, each CSI condition
comprised two blocks of 96 trials (with 48 switch and 48 repeat
trials in each block). Repeat and switch trials were presented in
a pseudo-randomized order controlling for transitions between
dimensions and congruency effects. Each trial involved two
dimensions that differ (e.g., shape and color) and two dimen-
sions that were kept constant (e.g., number and outline). Trials
were all incongruent ones (i.e., each dimension was associated
with a different response) so that it was always possible to know
whether the participant responded correctly according to the
target dimension. Incongruency was, thus, kept constant dur-
ing the entire task. The order of the dimensions was pseudo-
randomized to control for transitions and to ensure an equal
number of presentations for each dimension. In the short CSI
(0 milliseconds) condition, the cue was directly followed by the
target and the two potential matching gures until a response
was provided. In the long CSI (1800 milliseconds) condition, a
black screen followed the cue during 500 milliseconds was then
replaced by a point in the center of the screen for 250 millisec-
onds. The black screen–point sequence was repeated twice, fol-
lowed by another black screen for 300 milliseconds before the
presentation of the target and the two potential matching g-
ures. The intertrial interval was set to 300 milliseconds in the
long CSI condition and to 2100 milliseconds in the short CSI
condition to balance the total time for each trial. Both short
and long CSIs were preceded by a training block of 10 trials and
administered within a single session that lasted approximately
30 minutes, with a short break between each intervaltype.
Task-goal switching analyses are based on the percentage
of errors (task-switching accuracy) and on mean RTs for trials
with correct responses only (task-switching speed). Both meas-
ures are complementary and necessary to assess a potential
task-switching modulation [62]. The rst three trials in each
block were considered as warm-up trials and excluded from the
analyses. Trials immediately following an error were also dis-
carded from the analyses (9% of trials). RTs’ outliers were identi-
ed for each participant, each CSI duration (short vs. long), and
each task-switching condition (repeat vs. switch trials) using the
generalized extreme studentized deviate (GESD) test [62]. This
procedure led to discard 3.4% of trials.
Data Analyses
Data are expressed as mean ± standard deviation, unless
mentioned otherwise. Signicance level was set at p < .05.
Partial etas-squared was calculated as a measure for effect
size. Actigraphic data were analyzed using ActiLife 6 software
(ActiGraph, 2014). Bedtime was visually determined and com-
pared with the information reported by the participants in
their sleep diary. Sleep duration was computed using the Cole
Kripke algorithm [63]. Separate mixed-design ANOVAs were
conducted with between-subject factor Group (RS vs. SD) and
within-subject factor Session (evening vs. morning). For task-
goal switching accuracy and speed, mixed-design ANOVAs
were computed with between-subject factor Group and within-
subject factors Session, CSI Duration (short CSI vs. long CSI),
and Task Switching (repeat trials vs. switch trials). Interactions
with task switching were decomposed using planned com-
parisons according to our hypotheses (i.e., increased switch
costs in the SD group in the morning compared with the even-
ing and the RS group). Interactions were decomposed using
Tukey’s post hoc comparisons when applicable. The relation-
ships between SD-related changes in task-goal switching per-
formance and changes in other cognitive measures or scales
were investigated using Pearson correlations and Bayesian
correlations using JASP-software [64]. Changes in fatigue and
sleepiness subjective scores (VAS-S, VAS-F, and KSS), task per-
formance (PVT, N-back, and Stop Signal tasks), and spontane-
ous EBR were estimated by subtracting the results obtained
during the evening session from the results obtained during
the morning session [ΔSession = morning session – evening
session]. p-Values were adjusted for multiple comparisons
using Bonferroni correction per variable type compared with
the measures of task switching. Correlations were corrected
for the three questionnaires scores (VAS-S, VAS-F, and KSS),
two PVT measures (PVT speed and variability), and two N-back
performances (WM speed and WM accuracy). The difference
between two correlation coefcients was computed using the
r-to-Fisher-z transformation: rʹ =.5*[ln(1+r) – ln(1−r)], where rʹ
is the Fisher-z transformed (to a normally distributed variate)
Pearson correlation coefcient, r and r is the standard Pearson
correlation coefcient. The signicance of the difference
between two correlation coefcients is computed as follows:
d = r1ʹ–r2ʹ, where d is the difference between the two Fisher
z-transformed correlation coefcients; sd=Square Root ((n1 +
n2−6)/((n1−3) × (n2−3))), where sd is the standard error of the
difference between the two normalized (Fisher-z transformed)
correlation coefcients, n1 and n2 are the two sample sizes (for
r1 and r2, respectively). The test statistic d/sd is then evaluated
against the t distribution with df= n1 + n2 − 4 degrees of free-
dom. The one-sided and two-sided p values are computed as
usual, by considering either both sides or only one side of the
t-distribution. Here, we chose the two-sided p value.
Results
Sleep Prior to the Experiment
Mean sleep duration during Nights 1 to 3 preceding the experi-
mental night averaged, respectively, 428 ± 59, 428 ± 57, and
474± 69 minutes in the RS group and 429±74, 433± 71, and
444±61 minutes in the SD group. Amixed-design ANOVA com-
puted on sleep duration with between-subject factor Group (RS
vs. SD) and within-subject factor Night (1 to 3)disclosed a main
effect of Night [F(2,66)= 3.317, p=.042, η2
p= .091]. Participants
in both groups tended to sleep more during night 3 than dur-
ing nights 1 (p=.065) and 2 (p=.098), without any signicant
6 | SLEEPJ, 2018, Vol. 41, No. 2
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018
difference between nights 1 and 2 (p=.982). The main effect of
Group and the Group × Night interaction failed to reach signi-
cance (all p-values >.357). Actimetric measures indicate that the
RS group slept 397±40 minutes during the experimental night,
whereas the SD group did not sleep.
Sleepiness, Fatigue, Alertness, and Vigilance
Figure3 presents the results and post hoc comparisons for the
KSS, VAS-S, VAS-F, and PVT. Means and standard deviations are
presented in Supplementary Table S1. Interactions were signi-
cant for all variables (all ps < .001; see Supplementary Table S2).
Figure3. Results for KSS, VAS-S, and VAS-F scores, PVT speed and variability, N-back task, Stop signal task, EBR, and task-goal switching for both groups (RS, SD) and
both sessions (evening, morning). RS=regular sleep group; SD= sleep deprivation group; KSS= Karolinska Sleepiness Scale; VAS-F/S= Visual Analogue Scales for
Fatigue/Sleepiness; PVT=Psychomotor Vigilance Task; PVT speed=PVT median of RTs; PVT variability=PVT coefcient of variation of RTs; EBR/minute=spontaneous
eye blink rate per minute; WM (working memory N-Back) speed=RT N2– RT N0; WM accuracy=accuracy N0– accuracy N2; SSRT=Stop Signal Reaction Time; Accuracy
and Latency switch costs=switch trials – repeat trials. *p < .05; **p < .01; ***p < .001.
Slama etal. | 7
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018
Post hoc tests showed that scores were similar in RS and SD
groups in the evening, but signicantly differed in the morning
with the SD group being more sleepy, more fatigued, and exhib-
iting decreased alertness and vigilance.
WM and Inhibition
WM N-BackTask
Figure3 presents the results and post hoc comparisons for WM
speed and accuracy. Means and standard deviations are pre-
sented in Supplementary Table S3. The mixed-design ANOVA
disclosed a main effect of Group [F(1,33)=6.06, p=.019, η2
p=.155],
with RS participants exhibiting a higher accuracy than the SD
group. Session and Group × Session interaction effects failed to
reach signicance (all p-values >.258). The mixed-design ANOVA
computed on WM speed disclosed a main effect of Session
[F(1,33) = 7, p = .012, η2
p = .175], with participants being faster
in WM in the morning than in the evening. Group and Group ×
Session interaction effects failed to reach signicance (all p-val-
ues >.174).
Inhibition Stop SignalTask
Figure3 presents the results and post hoc comparisons for the
Stop Signal task. Means and standard deviations are presented
in Supplementary Table S3. The mixed-design ANOVA com-
puted on SSRT disclosed a main effect of Group [F(1,33)=6.22,
p=.018, η2
p=.158] and a signicant Group × Session interaction
[F(1,33)= 17.23, p < .001, η2
p= .343]. The Session effect was not
signicant [F(1,33)=.69, p=.412, η2
p=.02]. Post hoc comparisons
indicated that, in the SD group, SSRT increased in the morning
compared with the evening (p=.006). SSRT in the evening and
morning sessions did not differ signicantly from those in the
RS group (p=.115). SSRT was signicantly higher in the SD group
than in the RS group in the morning (p < .001) but not in the
evening (p=.996).
SpontaneousEBR
Figure 3 presents the results and post hoc comparisons for
spontaneous EBR. Means and standard deviations are presented
in Supplementary Table S3. The mixed-design ANOVA com-
puted on mean EBR disclosed a signicant main effect of ses-
sion [F(1,33)= 7.47, p=.01, η2
p= .185] and a signicant Group ×
Session interaction [F(1,33)=10.85, p=.002, η2
p=.247]. The Group
effect was not signicant [F(1,33)=.41, p=.525, η2
p=.012]. Post
hoc comparisons indicated that, in the SD group, EBR increased
in the morning compared with the evening (p < .001) session. In
the RS group, EBR did not differ between evening and morning
sessions (p < .979). No other comparison reached statistical sig-
nicance (all p-values >.149).
The Pearson correlation coefcient between EBR in the even-
ing and morning sessions in the RS group was highly signicant
(r=0.903, p < .001). This correlation was not signicant in the SD
group (r= 0.365, p=.137). Correlation coefcients were signi-
cantly different between the RS and SD groups (p=.007).
Task-Goal Switching
For task-goal switching accuracy, the mixed-design ANOVA com-
puted on percentages of errors disclosed main effects of Group
[F(1,33)=6.97, p=.013, η2
p=.174], Session [F(1,33)=7.49, p=.01,
η2
p=.185], CSI Duration [F(1,33)= 10.45, p=.003, η2
p=.241], and
Task Switching [F(1,33) = 51.37, p < .001, η2
p = .609], as well as
Group × Session [F(1,33)=20.61, p < .001, η2
p=.384], Group × Task
Switching [F(1,33)=5.77, p=.022, η2
p= .149], Group × Session ×
Task Switching [F(1,33)= 6.55, p =.015, η2
p= .166], and Group ×
Session × CSI Duration [F(1,33)=5.79, p= .022, η2
p= .149] inter-
action effects. No other interaction reached signicance (all
p-values >.328). Planned comparisons for the Group x Session x
Task Switching interaction indicated that accuracy switch-cost
scores were signicantly higher in the SD group than in the RS
group in the morning (p=.005) but not in the evening (p=.689).
Furthermore, accuracy switch-cost scores increased in the SD
group in the morning compared with the evening (p = .041).
Figure 3 presents accuracy switch-cost scores for both groups
and both sessions. Figure4 presents accuracy scores for repeat
and switch trials, for both groups, both sessions, and both CSI
durations. Supplementary Table S4 presents RTs and switch-cost
scores for both groups, both sessions, and both CSI durations.
Post hoc tests for the Group × Session × CSI Duration inter-
action indicated that participants in the SD group made more
errors in the morning than in the evening for both short
(p= .019) and long (p < .001) CSIs. They also made more errors
in the morning than participants in the RS group for long CSI
(p= .003). No other comparison reached statistical signicance
(all p-values >.127).
Concerning task-goal switching speed, the mixed-design
ANOVA computed on RTs disclosed a main effect of Task
Switching [F(1,33)=4.77, p =.036, η2
p= .126] with participants
being faster on repeat trials than on switch trials (878±167 vs.
893±184 milliseconds). The Group × Session interaction was sig-
nicant [F(1,33)=5.06, p=.031, η2
p=.133], but post hoc compari-
sons failed to disclose signicant differences (all p-values >.175).
No other comparison reached statistical signicance (all p-val-
ues >.16). Figure3 presents accuracy switch-cost scores for both
groups and both sessions. Supplementary Table S5 presents RTs
Figure 4. Task-goal switching accuracy in the cued match-to-sample task by
session (evening, morning) and trials (repeat, switch) for RS and SD groups;
RS=regular sleep group; SD=sleep deprivation group. The accuracy switch costs
are the difference between switch and repeat trials. **p < .01; *p < .05.
8 | SLEEPJ, 2018, Vol. 41, No. 2
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018
and switch-cost scores for both groups, both sessions, and both
CSI durations.
Relationships Between Impairments in Task-Goal
Switching and Other Variables
In the evening session, we observed a signicant negative cor-
relation between EBR and task-goal switching accuracy (across
all participants). Participants with lower spontaneous EBR dis-
played higher accuracy switch costs. This relationship was
signicant at long CSI (r = −0.353; p = .037) but did not reach
signicance at short CSI (r = 0.265; p = .124). However, in the
SD group, correlations between ΔSession accuracy switch-cost
scores and ΔSession VAS-S, VAS-F, KSS, EBR, PVT speed, PVT
variability, Stop signal task SSRT or WM speed, and accuracy
did not reach statistical signicance (all p-values >.213). These
results indicate that increases in accuracy switch-cost scores
(i.e., worse performance) from morning to evening after SD were
not signicantly associated with the alterations of other vari-
ables after SD (for detailed values, see Supplementary TableS6).
Finally, in the SD group, we assessed the relationships
between SD effects on EBR and other cognitive measures or
scales (for detailed values, see Supplementary Table S6). No sig-
nicant correlations were evidenced between ΔSession EBR and
ΔSession for the other variables (all p-values >.167).
We conducted Bayesian correlational analyses [64] in the
SD group to investigate the statistical signicance of the null
hypothesis (i.e., an absence of association between the vari-
ables). Correlations between ΔSession accuracy switch-cost
scores and ΔSession VAS-S, VAS-F, KSS, EBR, PVT speed, Stop sig-
nal task SSRT, or WM accuracy were considered signicantly in
favor of the null hypothesis (Bayes Factors <0.333). Correlations
between ΔSession accuracy switch-cost scores and WM speed or
PVT variability were considered as inconclusive (Bayes Factors
comprised between 0.333 and 3). Finally, correlations between
ΔSession EBR and KSS, WM accuracy, or Stop signal task SSRT
were also in favor of the null hypothesis while they were consid-
ered inconclusive for the other variables.
Discussion
The current study investigated the impact of one night of total
SD on task-goal switching in healthy participants. Our results
evidence deteriorated task-goal switching performance after SD,
characterized by higher accuracy switch costs in the morning
compared with the evening in SD participants, and higher accu-
racy switch costs than in the RS group in the morning but not in
the evening. SD also deteriorated objective and subjective mark-
ers of fatigue and sleepiness, increased spontaneous EBR, and
decreased response inhibition in the Stop Signal task. However,
we did not nd signicant correlations between changes in task-
goal switching performance and changes in objective and sub-
jective markers of fatigue and sleepiness, response inhibition,
or spontaneous EBR. Finally, SD did not signicantly affect WM
updating measures in the N-backtask.
To the best of our knowledge, the current study is the rst
to demonstrate that SD negatively affects the task-goal com-
ponent of task switching. Indeed, prior studies used paradigms
investigating task-rule switching [24–26], a related but distinct
component of cognitive exibility [9, 10]. The present ndings
are also consistent with our previous study that evidenced a
benecial effect of a nap on task-goal switching [32]. In the pre-
sent study, SD resulted in the reverse effects of napping, sup-
porting the hypothesis of a relationship between sleep-related
restorative processes and task-goal switching performance.
Task-goal activation and maintenance are key features in most
theoretical frameworks of cognitive control [3, 28, 29]. Therefore,
task-goal switching alterations observed after SD suggest that
failure to adjust or maintain the task goal could be one of the
reasons for a deterioration of cognitive control after SD. Our
results also extend previous research on task-rule switching
[24–26] in showing that both task-goal and task-rule switching
can be affected bySD.
The effects of SD on task-goal switching were found on ac-
curacy but not on response latencies. A signicant Session ×
Group interaction was observed on latency switch costs with a
pattern similar to accuracy switch costs (Figure3), but post hoc
tests failed to disclose any signicant difference. The latency
switch cost usually observed in task switching studies largely
relies on the learning and rehearsal of arbitrary associations be-
tween stimuli and responses [65, 66], and the preparation mainly
benets the rehearsal or the reconguration of S–R mappings
(i.e., task-rules). The proposal that retrieving information from
long-term memory plays a central role in task switching was
developed in the context of the retrieval-demand hypothesis [66].
The authors observed that preparatory effects on switch costs
and the retrieval-demands were eliminated when relevant task
rules (i.e., the critical S–R associations) were directly provided by
the task cues. They also suggested that episodic retrieval might
be sufcient to explain the endogenous component (recongur-
ation) of task switching. According to them, it would be impos-
sible to represent more than one task rule in WM. Consequently,
after a switch, the next task rule has to be retrieved from
long-term memory. They also suggested that in some cases the
failure to prepare (see also Ref. 67) requires retrieving S–R map-
pings when the targets are displayed, thus leading to a residual
switch cost. This proposal was supported by the fact that par-
ticipants can only partially prepare, maybe for one S–R associ-
ation at a time [65]. In line with these elements, we propose that
the latency switch cost is mainly affected by task-rule retrieval
in episodic memory because these task rules are encoded in
long-term memory and their retrieval is predominantly a time
consuming process affecting latency switch cost. In contrast, as
evidenced by this experiment and previous work from our team
[32], the accuracy switch cost appears to be more related to task-
goal switching. This is presumably due to the fact that it involves
an online maintenance and adaptation of the goal in WM, and
less long-term memory retrieval. Failure to maintain or adjust
the goal would therefore be characterized by errors and an ac-
curacy switch-cost. The state-instability hypothesis [20] posits
that SD induces instability in the capacity to maintain attention
and alertness due to the growing inuence of sleep initiating
mechanisms. Amajor component of state instability might be
task-set instability. The task sets and task-goals can be subject
to temporary breakdowns, an effect actually close to the notion
of goal neglect [68]. However, the notion of task-set instability is
more restricted than the notion of state instability and allows
clearer predictions. In particular, it can be predicted that the
consequences of SD will be stronger with additional conditions
promoting task-set instability, like in a task-goal switching para-
digm. Indeed, shifting between tasks by denition involves a
Slama etal. | 9
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018
temporary task-set instability. Consequently, task-set instability
after SD should be stronger in switch than in repeat trials [26], a
prediction supported by ourdata.
It is well known that insufcient sleep leads to a general slow-
ing of response speed and increased variability in performance,
particularly for simple measures of alertness and vigilance [22,
69]. Accordingly, we evidenced SD-related changes in subjec-
tive and objective measures of sleepiness and fatigue (includ-
ing alertness and vigilance). RS and SD groups exhibited similar
levels in the evening, but differed signicantly in the morning,
the SD group being sleepier, more fatigued, less alert, and vigi-
lant. In addition, spontaneous EBR were higher after SD. These
results are in line with previous ndings [38–40, 70]. However,
the increase in EBR was not signicantly correlated with modu-
lation of subjective and objective markers of sleepiness or vigi-
lance, indicating that SD-related processes causing EBR increase
might differ from those affecting alertness and vigilance.
Task switching is one of the cognitive control components
that has been most convincingly associated with spontaneous
EBR [33, 35, 36, 55]. In this study, we found a negative association
between EBR and accuracy switch costs. This association was
signicant at long CSI, indicating that preparation processes
associated with goal adaptation and goal maintenance (pro-
active control) are associated with EBR and, putatively, with stri-
atal dopamine. Indeed, in previous research, increased EBR was
also associated with increased cognitive exibility and lower
switch costs. It was concluded that higher dopamine activity
favors an easier disengagement from the previous task [33]. Our
results are in line with this proposal. However, the EBR increase
observed after SD was not signicantly associated with task-
goal switching impairment, suggesting that different SD-related
processes affect spontaneous EBR and cognitive exibility. This
result is compatible with previous works that have questioned
the plausibility of a dopamine increase after SD [42]. Therefore,
if the increase in EBR observed after SD is not due to dopamine
increase, it should not be strongly associated with a task-goal
switching modulation after SD, which is the case in the present
study. Our results, therefore, support the idea that EBR increase
after SD is not or only partially related to a dopamine increase
or cognitive exibility impairment.
Previous studies have shown that vigilance is much more
affected by SD than executive functions [22, 71], and some stud-
ies have failed to disclose a specic effect of SD on executive
functions [72]. Executive functions are by essence multidimen-
sional, and operate on and through other cognitive processes.
Therefore, any cognitive task recruiting executive functions
will recruit other, more basic, processes as well, a phenomenon
known as the task impurity problem [73]. It cannot be excluded
that low scores previously observed on executive function tasks
after SD were due to decits in nonexecutive components of
task performance, like alertness or vigilance, more than decits
in executive functioning per se. Controlling for nonexecutive
components in executive tasks can be achieved using control
conditions and composite scores such as switch costs, SSRT, or
WM updating score. In the present study, WM updating meas-
ures were not affected by SD, in agreement with prior studies
disclosing null [74, 75] or small effects [76] of SD in a N-back task.
On the other hand, SD exerted an impact on response inhibition
in the Stop Signal task, which is in agreement with prior nd-
ings [77, 78].
The lack of eye tracking or electrooculography (EOG) is a limi-
tation of our study, as this would have allowed the evaluation of
pupillary dilation, and trial-to-trial changes in EBR during task,
to track potential uctuations in dopamine related to ongoing
task demands [37]. For instance, both EBR during task and pu-
pillary dilation have been associated with cognitive processes
such as cognitive inhibition [79] and mental fatigue [80]. Further
studies should investigate these aspects. Another limitation of
our study is the absence of a direct comparison between task-
rule and task-goal switching. Future SD studies should use ex-
perimental designs allowing a direct comparison between these
two components of task switching.
In conclusion, we showed here that total SD negatively
affects both task-goal switching and response inhibition meas-
ures, but not WM updating. Furthermore, spontaneous EBR
increase after SD was not signicantly associated with impaired
task-goal switching. Altogether, our results conrm the impor-
tance of considering key executive functions such as WM updat-
ing, inhibition, and task switching as, at least partially, separated
[4, 43] and having a relative sensitivity to the effects of SD.
Supplementary Material
Supplementary material is available at SLEEP online.
Acknowledgments
The authors would like to thank Guillermo Borragan Pedraz and
Ivan Ronse for their help in the acquisition of the data. HS and
DOC contributed equally to this work.
Funding
The study was in part supported by a FRSM – Fonds de la Recherche
Scientique et Médicale grant #7020836 (to PP).
Disclosure Statement
GD is a postdoctoral researcher supported by Jean-François
Peterbroeck’s foundation. RS is a Postdoctoral Researcher of the
FRS-FNRS Fonds National de la Recherche Scientique, Belgium.
RL was a Research Associate supported by a grant Brains Brack
to Brussels from INNOVIRIS, the Brussels Institute for Research
and Innovation, Région Bruxelles-Capitale, Belgium. PP was
Francqui Research Professor, 2013–2016.
References
1. Diamond A. Executive functions. Annu Rev Psychol. 2013; 64:
135–168. doi: 10.1146/annurev-psych-113011-143750.
2. Braver TS. The variable nature of cognitive control: a
dual mechanisms framework. Trends Cogn Sci. 2012; 16(2):
106–113.
3. Miller EK, Cohen JD. An integrative theory of prefrontal cor-
tex function. Annu Rev Neurosci. 2001; 24(1): 167–202.
4. Miyake A, Friedman NP. The nature and organization of
individual differences in executive functions: four general
conclusions. Curr Dir Psychol Sci. 2012; 21(1): 8–14.
10 | SLEEPJ, 2018, Vol. 41, No. 2
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018
5. Kiesel A, Steinhauser M, Wendt M, etal. Control and inter-
ference in task switching–a review. Psychol Bull. 2010; 136(5):
849–874.
6. Vandierendonck A, Liefooghe B, Verbruggen F. Task switch-
ing: interplay of reconguration and interference control.
Psychol Bull. 2010; 136(4): 601–626.
7. Jersild AT. Mental set and shift. Archives of Psychology. 1927;
14(89): 1–81.
8. Spector A, Biederman I. Mental set and mental shift revis-
ited. Am J Psychol. 1976; 89(4): 669–679.
9. De Baene W, Albers AM, Brass M. The what and how compo-
nents of cognitive control. Neuroimage. 2012; 63(1): 203–211.
10. Rubinstein JS, Meyer DE, Evans JE. Executive control of cog-
nitive processes in task switching. J Exp Psychol Hum Percept
Perform. 2001; 27(4): 763–797.
11. Ravizza SM, Carter CS. Shifting set about task switching:
behavioral and neural evidence for distinct forms of cogni-
tive exibility. Neuropsychologia. 2008; 46(12): 2924–2935.
12. Hakun JG, Ravizza SM. Cognitive control: preparation of
task switching components. Brain Res. 2012; 1451: 53–64.
doi: 10.1016/j.brainres.2012.02.046.
13. Dreisbach G, Goschke T, Haider H. The role of task rules and
stimulus-response mappings in the task switching para-
digm. Psychol Res. 2007; 71(4): 383–392.
14. Logan GD, Schneider DW. Interpreting instructional cues in
task switching procedures: the role of mediator retrieval.
J Exp Psychol Learn Mem Cogn. 2006; 32(2): 347–363.
15. Logan GD, Schneider DW. Priming or executive control?
Associative priming of cue encoding increases “switch
costs” in the explicit task-cuing procedure. Mem Cognit.
2006; 34(6): 1250–1259.
16. De Baene W, Brass M. Dissociating strategy-depend-
ent and independent components in task prepar-
ation. Neuropsychologia. 2014; 62: 331–340. doi: 10.1016/j.
neuropsychologia.2014.04.015.
17. Ruge H, Jamadar S, Zimmermann U, Karayanidis F. The many
faces of preparatory control in task switching: reviewing a
decade of fMRI research. Hum Brain Mapp. 2013; 34(1): 12–35.
18. von Bastian CC, Druey MD. Shifting between mental sets:
An individual differences approach to commonalities and
differences of task switching components. J Exp Psychol Gen.
2017; 146(9): 1266–1285.
19. Chee MW, Choo WC. Functional imaging of working mem-
ory after 24hr of total sleep deprivation. J Neurosci. 2004;
24(19): 4560–4567.
20. Doran SM, Van Dongen HP, Dinges DF. Sustained attention
performance during sleep deprivation: evidence of state
instability. Arch Ital Biol. 2001; 139(3): 253–267.
21. Durmer JS, Dinges DF. Neurocognitive consequences of
sleep deprivation. Semin Neurol. 2005; 25(1): 117–129.
22. Killgore WD. Effects of sleep deprivation on cogni-
tion. Prog Brain Res. 2010; 185: 105–129. doi: 10.1016/
B978-0-444-53702-7.00007-5.
23. Lim J, Dinges DF. Sleep deprivation and vigilant attention.
Ann N Y Acad Sci. 2008; 1129(1): 305–322.
24. Bratzke D, Rolke B, Steinborn MB, Ulrich R. The effect of 40h
constant wakefulness on task-switching efciency. J Sleep
Res. 2009; 18(2): 167–172.
25. Couyoumdjian A, Sdoia S, Tempesta D, etal. The effects of
sleep and sleep deprivation on task-switching perform-
ance. J Sleep Res. 2010; 19(1 Pt 1): 64–70.
26. Heuer H, Kleinsorge T, Klein W, Kohlisch O. Total sleep
deprivation increases the costs of shifting between simple
cognitive tasks. Acta Psychol (Amst). 2004; 117(1): 29–64.
27. Jennings JR, Monk TH, van der Molen MW. Sleep deprivation
inuences some but not all processes of supervisory atten-
tion. Psychol Sci. 2003; 14(5): 473–479.
28. O’Reilly RC. Biologically based computational models of
high-level cognition. Science. 2006; 314(5796): 91–94.
29. O’Reilly RC, Frank MJ. Making working memory work: a
computational model of learning in the prefrontal cortex
and basal ganglia. Neural Comput. 2006; 18(2): 283–328.
30. Basner M, Rao H, Goel N, Dinges DF. Sleep deprivation and
neurobehavioral dynamics. Curr Opin Neurobiol. 2013; 23(5):
854–863.
31. Chee MW, Chuah LY. Functional neuroimaging insights into
how sleep and sleep deprivation affect memory and cogni-
tion. Curr Opin Neurol. 2008; 21(4): 417–423.
32. Slama H, Deliens G, Schmitz R, Peigneux P, Leproult R.
Afternoon nap and bright light exposure improve cognitive
exibility post lunch. PLoS One. 2015; 10(5): e0125359.
33. Dreisbach G, Müller J, Goschke T, etal. Dopamine and cog-
nitive control: the inuence of spontaneous eyeblink rate
and dopamine gene polymorphisms on perseveration and
distractibility. Behav Neurosci. 2005; 119(2): 483–490.
34. Slagter HA, Georgopoulou K, Frank MJ. Spontaneous eye
blink rate predicts learning from negative, but not posi-
tive, outcomes. Neuropsychologia. 2015; 71: 126–132. doi:
10.1016/j.neuropsychologia.2015.03.028.
35. Müller J, Dreisbach G, Brocke B, Lesch KP, Strobel A, Goschke
T. Dopamine and cognitive control: the inuence of spon-
taneous eyeblink rate, DRD4 exon III polymorphism and
gender on exibility in set-shifting. Brain Res. 2007; 1131(1):
155–162.
36. Tharp IJ, Pickering AD. Individual differences in cognitive-
exibility: the inuence of spontaneous eyeblink rate, trait
psychoticism and working memory on attentional set-
shifting. Brain Cogn. 2011; 75(2): 119–125.
37. Jongkees BJ, Colzato LS. Spontaneous eye blink rate as pre-
dictor of dopamine-related cognitive function-A review.
Neurosci Biobehav Rev. 2016; 71: 58–82. doi: 10.1016/j.
neubiorev.2016.08.020.
38. Barbato G, De Padova V, Paolillo AR, Arpaia L, Russo E, Ficca
G. Increased spontaneous eye blink rate following pro-
longed wakefulness. Physiol Behav. 2007; 90(1): 151–154.
39. Barbato G, Ficca G, Beatrice M, Casiello M, Muscettola G,
Rinaldi F. Effects of sleep deprivation on spontaneous eye
blink rate and alpha EEG power. Biol Psychiatry. 1995; 38(5):
340–341.
40. Cafer PP, Erdmann U, Ullsperger P. The spontaneous eye-
blink as sleepiness indicator in patients with obstructive
sleep apnoea syndrome-a pilot study. Sleep Med. 2005; 6(2):
155–162.
41. Bohne A, Savage CR, Deckersbach T, etal. Visuospatial abili-
ties, memory, and executive functioning in trichotillomania
and obsessive-compulsive disorder. J Clin Exp Neuropsychol.
2005; 27(4): 385–399.
42. Volkow ND, Tomasi D, Wang GJ, etal. Evidence that sleep
deprivation downregulates dopamine D2R in ventral stri-
atum in the human brain. J Neurosci. 2012; 32(19): 6711–6717.
43. Miyake A, Friedman NP, Emerson MJ, Witzki AH, Howerter
A, Wager TD. The unity and diversity of executive func-
tions and their contributions to complex “Frontal Lobe”
tasks: a latent variable analysis. Cogn Psychol. 2000; 41(1):
49–100.
44. Beck AT, Steer RA, Garbin MG. Psychometric properties of
Beck Depression Inventory: twenty-ve years of evaluation.
Clin Psychol Rev. 1988; 8(1): 77–100.
Slama etal. | 11
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018
45. Horne JA, Ostberg O. A self-assessment questionnaire to
determine morningness-eveningness in human circadian
rhythms. Int J Chronobiol. 1976; 4(2): 97–110.
46. Krupp LB, LaRocca NG, Muir-Nash J, Steinberg AD. The
fatigue severity scale. Application to patients with multiple
sclerosis and systemic lupus erythematosus. Arch Neurol.
1989; 46(10): 1121–1123.
47. Ellis BW, Johns MW, Lancaster R, Raptopoulos P,
Angelopoulos N, Priest RG. The St. Mary’s Hospital sleep
questionnaire: a study of reliability. Sleep. 1981; 4(1):
93–97.
48. Akerstedt T, Gillberg M. Subjective and objective sleepiness
in the active individual. Int J Neurosci. 1990; 52(1-2): 29–37.
49. Lee KA, Hicks G, Nino-Murcia G. Validity and reliability of a
scale to assess fatigue. Psychiatry Res. 1991; 36(3): 291–298.
50. Dinges DI, Powell JW. Microcomputer analysis of per-
formance on a portable, simple visual RT task sustained
operations. Behav Res Methods Instrum Comput. 1985; 17(6):
652–655.
51. Schmidt C, Collette F, Leclercq Y, et al. Homeostatic sleep
pressure and responses to sustained attention in the
suprachiasmatic area. Science. 2009; 324(5926): 516–519.
52. van Zomeren AH, Brouwer WH. Clinical Neuropsychology of
Attention. Oxford: Oxford University Press, 1994.
53. Fiehmig HC, Steinborn M, Langner R, Scholz A, Westhoff
K. Assessing intraindividual variability in sustained atten-
tion: reliability, relation to speed and accuracy, and practice
effects. Psychol Sci. 2007; 49(2): 132–149.
54. Chermahini SA, Hommel B. The (b)link between creativity
and dopamine: spontaneous eye blink rates predict and
dissociate divergent and convergent thinking. Cognition.
2010; 115(3): 458–465.
55. Zhang T, Mou D, Wang C, et al. Dopamine and executive
function: Increased spontaneous eye blink rates correlate
with better set-shifting and inhibition, but poorer updating.
Int J Psychophysiol. 2015; 96(3): 155–161.
56. Owen AM, McMillan KM, Laird AR, Bullmore E. N-back work-
ing memory paradigm: a meta-analysis of normative func-
tional neuroimaging studies. Hum Brain Mapp. 2005; 25(1):
46–59.
57. Massat I, Slama H, Kavec M, etal. Working memory-related
functional brain patterns in never medicated children with
ADHD. PLoS One. 2012; 7(11): e49392.
58. Massat I, Slama H, Villemonteix T, et al. Hyperactivity in
motor response inhibition networks in unmedicated chil-
dren with attention decit-hyperactivity disorder. World J
Biol Psychiatry. 2016; 19(2): 101–111.
59. Rubia K, Smith AB, Brammer MJ, Toone B, Taylor E. Abnormal
brain activation during inhibition and error detection in
medication-naive adolescents with ADHD. Am J Psychiatry.
2005; 162(6): 1067–1075.
60. Cohen JD, MacWhinney B, Flatt M, Provost J. PsyScope: a
new graphic interactive environment for designing psych-
ology experiments. Behav Res Methods Instrum Comput. 1993;
25(2): 257–271.
61. International School of Advanced Studies (SISSA/ISAS).
PsyScope X project (version 77 build) [computer software].
62. Rosner B. Percentage points for a generalized ESD many-
outlier procedure. Technometrics. 1983; 25(2): 165–172.
63. Cole RJ, Kripke DF, Gruen W, Mullaney DJ, Gillin JC. Automatic
sleep/wake identication from wrist activity. Sleep. 1992;
15(5): 461–469.
64. JASP Team (2018). JASP (Version 0.8.5) [Computer software].
65. Lien MC, Ruthruff E, Remington RW, Johnston JC. On the
limits of advance preparation for a task switch: do people
prepare all the task some of the time or some of the task
all the time? J Exp Psychol Hum Percept Perform. 2005; 31(2):
299–315.
66. Mayr U, Kliegl R. Task-set switching and long-term memory
retrieval. J Exp Psychol Learn Mem Cogn. 2000; 26(5): 1124–1140.
67. De Jong R, Berendsen E, Cools R. Goal neglect and inhibi-
tory limitations: dissociable causes of interference effects
in conict situations. Acta Psychol (Amst). 1999; 101(2-3):
379–394.
68. Duncan J, Emslie H, Williams P, Johnson R, Freer C.
Intelligence and the frontal lobe: the organization of goal-
directed behavior. Cogn Psychol. 1996; 30(3): 257–303.
69. Killgore WD, McBride SA, Killgore DB, Balkin TJ. The effects
of caffeine, dextroamphetamine, and modanil on humor
appreciation during sleep deprivation. Sleep. 2006; 29(6):
841–847.
70. McIntire LK, McKinley RA, Goodyear C, McIntire JP. Detection
of vigilance performance using eye blinks. Appl Ergon. 2014;
45(2): 354–362.
71. Lim J, Dinges DF. A meta-analysis of the impact of short-
term sleep deprivation on cognitive variables. Psychol Bull.
2010; 136(3): 375–389.
72. Tucker AM, Whitney P, Belenky G, Hinson JM, Van Dongen
HP. Effects of sleep deprivation on dissociated components
of executive functioning. Sleep. 2010; 33(1): 47–57.
73. Phillips LH. Do ‘frontal tests’ measure executive functions?
Issues of assessment and evidence from uency tests. In:
Rabbitt P, ed. Methodology of Frontal and Executive Function.
Hove: Psychology Press, 1997:191–213.
74. Miyata S, Noda A, Ozaki N, etal. Insufcient sleep impairs
driving performance and cognitive function. Neurosci Lett.
2010; 469(2): 229–233.
75. Teran-Perez G, Ruiz-Contreras AE, Gonzalez-Robles RO, etal.
Sleep deprivation affects working memory in low but not in
high complexity for the N-back test. Neurosci Med 2012; 3(4):
380–386.
76. Lo JC, Groeger JA, Santhi N, etal. Effects of partial and acute
total sleep deprivation on performance across cognitive
domains, individuals and circadian phase. PLoS One. 2012;
7(9): e45987.
77. Acheson A, Richards JB, de Wit H. Effects of sleep depriv-
ation on impulsive behaviors in men and women. Physiol
Behav. 2007; 91(5): 579–587.
78. Chuah YM, Venkatraman V, Dinges DF, Chee MW. The neural
basis of interindividual variability in inhibitory efciency
after sleep deprivation. J Neurosci. 2006; 26(27): 7156–7162.
79. Oh J, Han M, Peterson BS, Jeong J. Spontaneous eyeblinks are
correlated with responses during the Stroop task. PLoS One.
2012; 7(4): e34871.
80. Hopstaken JF, van der Linden D, Bakker AB, Kompier MA.
A multifaceted investigation of the link between mental
fatigue and task disengagement. Psychophysiology. 2015;
52(3): 305–315.
12 | SLEEPJ, 2018, Vol. 41, No. 2
Downloaded from https://academic.oup.com/sleep/article-abstract/41/2/zsx200/4737239
by guest
on 25 February 2018