This is the author’s version of a work that was submitted/accepted for pub-
lication in the following source:
Larue, Gregoire S., Rakotonirainy, Andry, & Pettitt, Anthony N. (2011) Driv-
ing performance impairments due to hypovigilance on monotonous roads.
Accident Analysis and Prevention, 43(6), pp. 2037-2046.
This file was downloaded from: http://eprints.qut.edu.au/42041/
c ? Copyright 2011 Elsevier
Notice: Changes introduced as a result of publishing processes such as
copy-editing and formatting may not be reflected in this document. For a
definitive version of this work, please refer to the published source:
Driving performance impairments due to hypovigilance on monotonous roads
Gr´ egoire S. LARUE∗,a, Andry Rakotonirainya, Anthony N. Pettittb
aCentre for Accident Research and Road Safety - Queensland , Queensland University of Technology, 130 Victoria Park road, Kelvin Grove 4059,
bSchool of Mathematical Sciences, Queensland University of Technology, Gardens Point, Brisbane 4000, Queensland, Australia
Drivers’ ability to react to unpredictable events deteriorates when exposed to highly predictable and uneventful driving tasks.
Highway design reduces the driving task mainly to a lane-keeping manoeuvre. Such a task is monotonous, providing little
stimulation and this contributes to crashes due to inattention. Research has shown that driver’s hypovigilance can be assessed
with EEG measurements and that driving performance is impaired during prolonged monotonous driving tasks. This paper
aims to show that two dimensions of monotony - namely road design and road side variability - decrease vigilance and
impair driving performance. This is the first study correlating hypovigilance and driver performance in varied monotonous
conditions, particularly on a short time scale (a few seconds). We induced vigilance decrement as assessed with an EEG
during a monotonous driving simulator experiment. Road monotony was varied through both road design and road side
variability. The driver’s decrease in vigilance occurred due to both road design and road scenery monotony and almost
independently of the driver’s sensation seeking level. Such impairment was also correlated to observable measurements from
the driver, the car and the environment. During periods of hypovigilance, the driving performance impairment affected lane
positioning, time to lane crossing, blink frequency, heart rate variability and non-specific electrodermal response rates. This
work lays the foundation for the development of an in-vehicle device preventing hypovigilance crashes on monotonous roads.
Key words: Monotony, Vigilance, Driving
Driving a car is one of the most common, though fairly
dangerous, tasks in industrial countries. Road crashes are
the main cause of premature death of people younger than
45. The burden of crashes is counted not only in lives, life
handicaps but also as a cost to the society. The road toll
in Australia in 2005 was 1 627 fatalities for an estimated
social cost of AUS $ 15 billion . These figures are still
of concern, though road safety interventions have improved
the situation. Human errors contribute to around 90% of
all crashes, and inattention to the forward roadway is often
a contributing factor to crashes . This suggests that the
major effort to improve road safety should target counter-
measures for driver inattention. However current measures
focus mainly on safety devices, improvement of the road
infrastructure, laws and regulations, but very little on hu-
man factors . In Queensland inattention and fatigue (re-
vealed in driver’s lapses in vigilance) contribute to 6% and
5% of fatal road crashes respectively. Furthermore, inat-
tention is the second contributing factor to all crash occur-
rences (12%) .
The driving task is complex and demanding and as a con-
sequence, infrastructure authorities have attempted to sim-
plify the driving task, for example through modifications to
highways, while vehicle manufacturers have enhanced car
design and equipment to improve comfort and safety (e.g.
ABS, cruise control, power steering).
duced to a lane-keeping task on highways and new types of
Email address: firstname.lastname@example.org (Gr´ egoire S. LARUE)
Preprint submitted to Accident Analysis & Prevention
crashes have emerged from these contemporary road safety
interventions . Drivers, particularly professional drivers,
may suffer from the monotony of the driving task resulting
in an increase in crash risk due to lapses of vigilance. It
has been shown that if the driving task is highly predictable
and uneventful then their ability to react to unpredictable
events deteriorates . Most fatigue crashes are recorded
on roads of good quality, with few curvatures and which
were supposedly designed to improve road safety [15, 32].
This raises the question of the role of monotony of the driv-
ing task in explaining vigilance related crashes.
The concept of monotony is typically entangled in the re-
search with fatigue and hypovigilance. Nevertheless Thif-
fault and Bergeron ’s experiment demonstrated that
driver vigilance decrement as indexed by subjective and
physiological measures is more frequent in a monotonous
environment, such as highways and rural roads. It is also
estimated that 27% of city drivers having reported fatigue
related to a road crash or incident did not feel tired prior
to the incident and 35% felt slightly tired. Most of these
drivers were driving on well known, repetitive routes. This
phenomenon appears even more on rural roads where 45%
of drivers involved in crashes reported that they were not
tired at all prior to the incident and where road geometry
are highly monotonous .
The aim of this study is to assess the effects of road
monotony on driver vigilance and correlate it to impaired
driving performance. This study focuses on two factors
that decrease vigilance: (i) the road design (in terms of
predictability) and (ii) the roadside environment (in terms
of variability). An electroencephalograph (EEG) is used
to monitor driver vigilance throughout a simulator experi-
June 20, 2011
ment. The effects on vigilance decrement are then corre-
lated to driving performance measurements.
In this experiment a simulated understimulating driving
task was designed to isolate the effects of road monotony
on driver vigilance. A driver simulator is used in order to
ensure the safety of participants as well as control that par-
ticipants are driving in the same conditions. This is done
by creating four different road scenarios which enable all
different combinations of high/low road design (geometry)
and roadside variability. Road geometry is varied through
vironment is varied through road signs, buildings and traffic
frequency. Driving more than 30 minutes on monotonous
roads has been shown to induce impairments similar to the
one observed when fatigued. On the other hand fatigue
due to time-on-task is observed after longer (more than one
hour) driving tasks [32, 34, 11]. In this study each driving
scenario is a short driving task of approximately 40 min-
utes. This disentangles monotony effects and fatigue due to
time-on-task. Such a design enables us first to isolate and
quantify effects of road monotony on driver vigilance. Then
vigilance levels are correlated with measurements from the
driver, the car and the environment. Such knowledge can
then be combined with appropriate mathematical models in
order to provide accurate real-time estimates of driver vigi-
lance on monotonous roads. Various mathematical models
can be used and a range of promising approaches can be
found in Larue et al. ’s study modelling vigilance evo-
lution during a short vigilance task.
2.1. Effects of monotonous driving on vigilance
Driving requires sustained vigilance, i.e. the ability to
maintain sustained attention within the road environment
.A lack of visual, motor or cognitive stimuli can
alter the ability to sustain vigilance. Drivers experience
vigilance decrement more frequently in monotonous envi-
ronments, especially when driving on highways at night.
Monotony has a mainly psychological effect on the driver
 through its effects on alertness and results in a vigi-
lance decrement. In fact the driver experiences boredom,
drowsiness or loss of interest quite rapidly when perform-
ing an uneventful driving task. Monotony on its own can
lead to vigilance lapses and to poor driving performance,
independent of other factors .
2.2. Consequences of monotonous driving
Monotony related crashes occur mainly on highways
(predictable, straight lanes) at night . This can be ex-
plained by the fact that a hypovigilant driver is unable to
react on time (or react at all) to critical events  such as
going off the road. This occurs rapidly and thirty minutes of
monotonous driving has been shown to be enough to induce
vigilance impairment .
Both performing a monotonous task and driving in
a monotonous environment have consequences on the
driver’s ability to drive. Indeed under such conditions
the driver may quickly lose the motivation to perform
the task and then become less vigilant . Particularly
the monotony of the road can result in passive fatigue,
i.e. fatigue due to the demands of the driving task itself.
The experiment by Oron-Gilad  shows that underload
situations such as monotonous situations lead to fatigue
symptoms and impaired driving performance although the
drivers were neither tired nor sleep deprived prior to the
Driving performance is most seriously affected by short
episodes of sleep occurring when the individual tried to stay
awake, episodes called microsleeps . However, decre-
ment in performance occurs during reduced level of vigi-
lance without microsleeps .
2.3. Individual differences in coping with monotonous
There is some variability between individuals’ capacity
to sustain vigilance. The profile of drivers more likely to be
involved in vigilance-related crashes have been determined
in a simulator experiment . Extravert drivers need more
stimulation changes, and as a consequence, they have less
disappears with time . Sensation seeking drivers need
varied, complex stimuli and experiences. They take physi-
cal and social risks to reach such experiences. This sensa-
tion seeking level can be more or less developed but leads
to risk taking driving, and negative reactions to monotonous
driving . Sensation seeking seems to be a good indica-
tor of the driver’s ability to focus on a monotonous task
2.4. Measuring effects of monotony on driving
2.4.1. Physiological measures
Effects of monotony are largely an impairment of cogni-
tive functions of the driver and they can be detected through
various physiological measures. Different types of tech-
nologies are used to detect such variations. Most of them
are based on physiological measurements, such as the EEG
. They are presented in the following paragraphs.
Brain activity. The most reliable and reproducible way to
observe psychological effects of monotonous driving is to
use an EEG [18, 23, 33]. EEG signals are analysed in the
frequency domain, and four different bands contain the in-
formation: α, β, θ and δ. It has been observed that EEG
θ and α frequencies rhythms increase during monotonous
tasks [18, 30]. The most reliable method to measure this
variation is to use the following algorithm:
creasing, this ratio between slow and fast wave activities
indicate a decrement of alertness [20, 3]. Bursts are also of
interest to detect increments in bands occurring relatively
sparsely. It can particularly be used to detect microsleep
following alpha and theta activities .
An EEG device cannot be used in a vehicle for at least
three reasons: (i) the inconvenience for the driver, (ii) the
prohibitive cost, and (iii) the noise introduced due to elec-
tromagnetic field interferences. Nevertheless, such a device
can be used in a laboratory-based experiment so that cor-
relation with driving performance (observed variables from
the driver the car and the environment) can be isolated and
β. When in-
Eye activity. Driving is, to a very large extent, a visual vig-
ilance task  so that various eye activity measures are
correlated to the observed changes in EEG signals during
monotonous driving. Research has shown that disappear-
ance of blinks, mini-blinks and relative quiescence in eye
movement are the earliest reliable signs of drowsiness, pre-
ceding slow eye movement and EEG alpha frequency and
amplitude changes . Blink duration and frequency in-
crease during fatigue . The PERcentage of eye CLOSure
(PERCLOS) measures the percentage of time during which
a driver has his eyes closed over a window of several min-
utes (usually 1 to 3 minutes). An eye closure is usually
characterised by an 80% (sometimes 70%) closure of the
eye compared to its nominal size. Fast blinks are removed
(time < 0.25s) from the computation of the PERCLOS.
Driver hypovigilance is then detected using a maximal eye
closure threshold .
Eye activity can be followed using systems based on
cameras. These systems are easily implemented in cars as
they are unobtrusive (as compared to EOGs).
Heart activity. Heart rate, measured by electrocardiogra-
phy (ECG), can be monitored to assess the individual phys-
iological level of workload. Most studies show that the
metric heart rate, if it changes at all, increases and the met-
ric heart rate variability decreases during effortful mental
processing . It has also been shown that heart rate de-
creases significantly during amonotonous driving task .
Electrodermal activity. Electrodermal activity (EDA) is
frequently used as an indirect measure of attention, cog-
nitive effort, or emotional arousal . EDA can be distin-
guished into tonic and phasic parts. The skin conductance
level (SCL) is the tonic value and shows the continuity of
activity over time. The skin conductance response (SCR)
is the phasic part and reveals changes in skin conductance
within a short time period . SCR can be due to stim-
ulus or non-specific causes. An increase in tonic EDA in-
dicates readiness for action and an increase of phasic EDA
indicates that one’s attention is directed toward a stimulus
[27, 29]. Skin conductance, in both tonic and phasic parts,
is therefore expected to decrease during monotonous tasks.
2.4.2. Driving performance
Impairments of driver’s cognitive abilities result in im-
pairment of driving performance.
through various metrics related to the vehicle and the en-
vironment. Driving experiments have highlighted the driv-
ing measures that are the most impaired during monotonous
Steering Wheel Movement (SWM) can be used to anal-
yse the lateral control of the vehicle. SWM correlates with
the effect of the monotony of the environment, particularly
its standard deviation . Another interesting indicator
is the ability of the driver to position their car on the road
in terms of lateral position (through its standard deviation)
. Another indicator of reduced vigilance is the diffi-
culty of maintaining constant speed. This can be checked
through the following metrics from the vehicle dynamics on
the road: the average speed and the standard deviation of
speed. It is also possible to assess it by the driver behaviour
through the actions on pedals . The limitation of such
This can be detected
indicators is that they do not take into account the environ-
ment (road geometry) which has an impact on the steer-
ing wheel pattern. Thus such systems are limited to sim-
ple driving contexts and should be relevant in the case of a
monotonous road design where the road is mainly straight.
In the case of more complex road geometry, this indicator
can be used on the road sections which are straight.
Twenty-five subjects, 7 males and 18 females aged be-
tween 18 and 49 (mean age = 29.1 years, SD = 8.3), vol-
unteered to participate in this study. Participants were re-
cruited from the Queensland University of Technology (10
students younger than 25 and 15 staff members older than
25). Participants had their licence for a minimum of two
years, drove a minimum of three days a week and drove
a minimum of 100 kilometres a week. This is similar to
previous research  so that potential differences due to
age cannot be attributed to inexperience. All subjects pro-
vided written consent for this study, which was approved
by the Queensland University of Technology ethics com-
mittee. Participants were paid AUS $80 for completing the
four driving sessions; students undertaking the first year
psychology subject received course credit for their partic-
The level of sensation seeking of participants was of in-
terest although participants were not specifically recruited
based on their level of sensation seeking. Of the 25 par-
ticipants, 16 were average sensation seekers, 6 were high
sensation seekers and 3 were low sensation seekers.
3.2. Experimental design
Monotony is multidimensional and mainly arises during
a driving task due to task monotony (no need to check mir-
rors, change gear, brake, etc.) and environment monotony.
Environment monotony results from a lack of stimulation
which can occur in a driving context due to the road design
(straight or presence of curves) and the roadside environ-
ment (quantity and variability of traffic signs, variability of
In this experiment, four different scenarios were run (see
Table I). In each experiment, the participant was asked to
drive and respect road rules for approximately 40 minutes.
The driving task was reduced to a lane keeping task to in-
duce task monotony:
• driving consisted in following a lane (no itinerary in-
volved) at constant speed (60 kilometres per hour),
without having to stop the car (no red traffic lights,
stops) or to brake a lot (no T intersections or perpen-
• no manual gear changes
• no need to change lane or indicate (turn signals)
• low traffic
Table 1: The four experiment scenarios
road design variability
Road geometry was varied through the curvature of the
road as well as its altitude (see Figure 1). In the road de-
sign with low variability, the road was essentially straight or
had few curves and was flat. Such a design was appropriate
to model highways and some rural roads. In the road de-
sign with high variability the road was a sequence of small
straightsections, significantcurvesandhills. Thismodelled
urban roads and some rural roads.
fatigue-related crashes occur (see Figure 1). The character-
istics of these spots as well as pictures have been obtained
from the Department of Main Roads Queensland (DTMR).
The roadside environment was varied in terms of road signs
frequency and variability and in terms of scenery (desert
with bushes along the road, urban highway, rural road with
houses, farms, industries etc.).
3.3.1. Driving simulator
Experimentation was conducted on the driving simulator
Scaner from OKTAL. The road and environment were de-
veloped to fit the study requirements in terms of monotony
(see section 3.2). The participant sat in front of a screen
where the simulator is played using an RGB video pro-
jector. The simulator displayed a view of the road with
a speedometer. The participant drove the simulator using
a modified computer steering wheel which provided force
feedback, and a two pedal set (brake and accelerator only).
Five speakers reproduced the inside sound of a car environ-
Any data related to the car or the environment was col-
lected by the simulator. These data were collected by the
use of a library of functions (speed, lane position, etc.)
available through a user graphical interface. Data related to
the driver required other sensors. Bioradio provided EEG
data (seven channels) and ECG data (one channel). Facelab
provided data related to the driver’s eyes (eye movements,
blinks, etc.). Biopac provided data related to skin conduc-
3.3.3. Synchronisation interface
Data collected from the simulator and the different sen-
sors were synchronised using RTmaps.
recorded and time stamped data from the different devices.
Different questionnaires were used to determine drivers’
profiles. The Eysenck Personality Questionnaire - Re-
vised (EPQR) was used to define the driver’s extroversion
level . The Sensation Seeking Scale - Form V (SSS)
was used to obtain the participant’s sensation seeking level
(Zuckerman, 1994). Finally a general background ques-
tionnaire was used to control driving experience, sleep pat-
tern and caffeine consumption.
Data extraction was performed with Matlab version
126.96.36.1997. Particularly, the EEGLAB v6.03b and Auto-
nomic Nervous System Laboratory toolboxes have been
used to analyse raw EEG, ECG and skin conductance data.
Statistical modelling was performed with the software R
Participants were tested individually in a quiet room in
four sessions lasting approximately one hour each. Each
participant drove in one of the four scenarios (randomly
assigned) in the simulator once a week at a fixed testing
time. Testing times were 9am (7 participants), 11am (5 par-
ticipants), 1pm (9 participants) and 3pm (4 participants).
Each participant chose a testing time for which they felt
they would be the most alert. During the first session par-
ticipants were asked to answer the questionnaires presented
before. Then, for each of the four scenarios participants
were given instructions about the nature of the experiment
(that is to drive following the road rules, at the speed limit,
familiarise the participant with the driving task on the sim-
ulator. Then the participant performed their scenario (about
forty minutes) at the end of which they answered questions
about their alertness at the end of the experiment.
traits on vigilance decrement
4.1. Data analysis
Driver vigilance was assessed through analysis of data
ent positions on the scalp (O1, O2, T5, T6, P3, P4 and F3)
following theInternational 10-20 Electrode Placement Sys-
tem at 80 Hz and are divided into 1 second epochs. Epochs
with too high/low values (threshold ±75µV), linear trends,
improbable data and/or abnormally distributed data were
rejected. A 4-term Blackman-Harris window and a 0.5 Hz
cut-off high-pass filter were also used to reduce low fre-
quency artefacts. Then Fast Fourier Transform (FFT) was
performed. This provided α, β, θ and δ band activities.α+θ
was used as an energy ratio. When increasing, this ratio
underlines an alertness decrease. These values were then
averaged over the different locations in 10 second time win-
dows. A mean value over the 5 first minutes was computed:
• the mean and standard deviation ofα+θ
above 2 standard deviations were then categorised as
high and correspond to decreased alertness.
β. Energy values
• a threshold (1.5 times the mean) for α and θ bursts.
Three consecutive epochs with values above the
threshold were categorised as microsleeps.
Figure 1: Screenshots of the four scenarios
The 5 first minutes were used as a reference for comparison
with the performance impairment throughout the driving.
Finally, high energy values and microsleeps were counted
over 120 second windows.
Vigilance was categorised into four levels: (i) level 1 simi-
lar to the 5 first minutes of driving, (ii) level 2 characterised
by alertness decrement alone, (iii) level 3 defined by mi-
crosleep alone and (iv) level 4 with microsleep during a
period of decreased alertness. Then a regression analysis
was performed to link the effects of different parameters
to the occurrence of alertness decrements and microsleeps.
This analysis required the use of Generalised Linear Mixed
Models (GLMMs) to take into account the correlation be-
tween repeated measures on the same participant (longitu-
dinal study). GLMMs with multinomial logit regression
were fitted to obtain the probabilities of the four different
vigilance levels. The different predictors used were:
• road monotony: road design and roadside monotony
• subjectsfactors: personalitytraits(SSSV,EPQR),test-
ing time, driving experience, age, amount of sleep the
previous night, usual sleeping times and caffeine con-
sumption before the experiment.
Participants’ personality traits were categorised into one of
the following classes: low (less than one standard deviation
(S.D.) in the available participants sample), normal (within
one S.D.) or high (greater than one S.D.) . Driving
experience was categorised into inexperienced (driver has
held licence for less than 5 years) and experienced (more
than 5 years). Age was categorised following age brack-
ets reported in government crash data: young (if less than
25) and middle aged (up to 49). Amount of sleep was cat-
egorised into less than 4 hours, between 4 and 6 hours, be-
tween 6 and 8 hours and more than 8 hours sleep. Caffeine
consumption prior to the experiment was categorised into
none, 1 to 2 hours ago, 2 to 4 hours ago and more than 4
hours ago. The participant’s ID and the week were consid-
ered as mixed effects in the model.
The multinomial logistic regression could not be obtained
directly using available statistical software. It was obtained
through the combination of binomial regressions (three in
this experiment). Therefore the linear combination of pre-
dictors could be factorised into three parts η1, η2and η3
ηi= α +?
jβj· Monotonyj+ γ · Time , i = 1...3
where Monotonyjis the level of road design/roadside vari-
ability and Profilejis one of the participant traits used as a
predictor. The resulting probabilities for each level were:
P[Level 1] = p1=
P[Level 2] = p2= (1 − p1) ·
P[Level 3] = p3= (1 − p1− p2) ·
P[Level 4] = p4= (1 − p1− p2) ·
Among the factors studied, the following were shown in
this experiment to have no statistically significant impact
(p-value p > 0.05) on the occurrences of alertness decre-
mentsormicrosleeps: theEPQRscale, gender, testingtime,
age , driving experience, amount of sleep (both usual and
the night preceding the experiment) and caffeine consump-
tion before the experiment. The other factors influencing
performance are given with their log-odds as well as their
p-value in Tables II to IV: time-on-task, road design vari-
ability, roadside variability and sensation seeking level. All
of these factors were statistically significant and had a no-
ticeable impact on vigilance impairment. The main fac-
tors were the time-on-task and the level of road design
monotony, these values depend on the sensation seeking
level of the participant.
Table 2: Linear regression estimates for η1
−2.08 · 10−2
−3.63 · 10−1
3.16 · 10−1
4.55 · 10−2
2.49 · 10−3
8.44 · 10−2
1.60 · 10−1
Table 3: Linear regression estimates for η2
−9.38 · 10−1
3.81 · 10−2
Table 4: Linear regression estimates for η3
6.69 · 10−2
−3.78 · 10−2
−3.86 · 10−1
6.99 · 10−3
9.49 · 10−3
1.52 · 10−1
Low road design/time
The evolution of the proportion of epochs characterised
by the decreased alertness state (p2 + p4) is shown in Fig-
ure 2 for the four different scenarios designed in this exper-
iment. Different curves are presented for the three different
levels of sensation seeking. It can be observed that the pro-
portion of epochs with high energy ratio increases through-
out the experiment for scenarios 1 and 2 (from around 17%
to 24%), while it remains fairly constant for scenarios 3
and 4 (at 17%). An increment in this proportion reveals
decrement in alertness, and this is observed for scenarios
characterised by a low road design variability. On the other
hand the effect of roadside variability is very small. High
and average sensation seekers performed approximately the
same in this experiment. Low sensation seekers have a dif-
ferent trend for driving scenarios with high roadside vari-
ability. They have an offset of 5% at the start of the experi-
ment. When the road is mainly straight, their alertness does
not improve throughout the experiment, while their alert-
ness increases on curvy roads to reach the level of the other
groups of sensation seekers at the end of the experiment.
The evolution with time of the proportion of occurrences
of microsleeps (p3+ p4) is shown in Figure 3 for the differ-
ent driving scenarios.
Microsleeps doubled throughout the experiment, on av-
erage from 10 to 25%. Road design variability had no effect
on microsleep occurrences. Roadside variability had an im-
pact only for low sensation seekers, with an offset of 5% in
the case of high roadside variability. Average and high sen-
sation seekers had the same pattern for microsleeps. This
suggests that during a low demanding driving task, mi-
crosleep occurrences tend to increase rapidly and almost
independently of the road design or roadside variability.
Monotonous road design is shown to have the most im-
pact compared to monotonous road scenery. Also, during
this low demanding driving task, microsleep occurrences
tend to increase rapidly and almost independently of the
road design or roadside variability. This monotony induced
impairment driver should not be mistaken for driver fatigue
due to sleep deprivation or circadian rhythms. Vigilance
decrement on straight roads emerges quickly and increases
as long as the task remains monotonous. Vigilance evolu-
tion over time during a driving task has not previously been
studied and so these results are therefore novel.
No difference is observed in this experiment between
medium and high sensation seekers. This research sug-
gests that monotonous roads are of concern not only for
high sensation seekers but also for average ones. This sup-
ulation. Particularly, assessing the level of sensation seek-
ing for professional drivers would not result in a reduction
of hypovigilance occurrences while driving. In this study
low sensation seekers do not perform better than the other
groups and perform even worse when the road design is not
varied. Such results are unexpected and might be the result
of the small number of low sensation seekers in our sample
(3). Further investigation would be required to conclude on
the effects of low sensation seeking on monotonous roads.
5. Analysis B: Correlation between vigilance level and
5.1. Data analysis
Episodes of reduced alertness and microsleeps - as as-
sessed through the states presented in section 4.1 - can be
correlated to different measures (surrogate measures) ob-
tained from sensors which can be used in real cars (see de-
tails in section 2). All epochs of each participants are cat-
egorised into the vigilance levels described in the previous
section. Values of surrogate measures are investigated by
vigilance level in order to detect any correlation between
the driving performance and the vigilance level.
ECG data were recorded at 80 Hz and were used to au-
tomatically extract the heart rate (HR), inter-beat-interval
(IBI) and T-wave amplitude. Thresholds to detect peaks
were manually adapted for each session of each participant.
Unrealistic values obtained for IBIs were filtered using 500
and 1300 ms as lower and upper limits respectively.
5.1.2. Eye activity
Eye activity data were collected at 60 Hz. Blink fre-
quency, blink duration, eye closure and PERCLOS were ex-
tracted by Facelab. PERCLOS was computed using a 75%
threshold over a 3 minutes time window.
5.1.3. Skin conductance
Skin conductance was collected at 1 Hz. Skin conduc-
tance level (SCL) and non-specific fluctuations (NSF) were
extracted. A threshold of 0.02 µS was used to find non-
specific responses. NSFs were categorised by their rates,
amplitudes, rise times and NSF half-recovery time.
Figure 2: Evolution of the proportion (in percentage) of epochs with high energy ratio for the four different scenarios (taking into account the influence of
the sensation seeking level SS of the participant)
5.1.4. Simulator data
Car and environment variables were obtained from the
simulator. Simulator data were sampled at 20 Hz. Lane
lateral shift, speed, steering wheel movement and Time
to Lane Crossing (TLC) were used in this analysis. Only
straight parts of the road were used to compute these met-
Each variable was normalised for each participant us-
ing the 5 first minutes of driving of each of the 4 sessions.
This provided a reference for the assessment of impairment
for each individual. For instance, the normalised heart rate
HRnorm(t) at time t was obtained from the heart rate HR(t)
at time t as follows:
HRnorm(t) =HR(t) − µHR
where µHRand σ2
rate during the five first minutes of the experiment respec-
To assess whether the driving performance was impaired in
the different vigilance states a linear model was fitted on
all these different metrics with the two factors presence of
alertness decrement and presence of microsleeps.
HRare the mean and variance of the heart
Factors associated with p-value p > 0.05 were consid-
ered to show no statistically significant correlation between
vigilance state and an evolution of such factors. Statisti-
cally significant correlations are presented below and sum-
marised in Table V. In this Table, the normalised value of
each variable is presented for:
• episodes of good vigilance
• episodes of reduced alertness (hypovigilance)
• episodes of microsleeps.
The standard deviation of such estimate is reported in
brackets. Episodes of good vigilance are used as a refer-
ence and values for reduced alertness and microsleeps are
reported only when statistically different from this refer-
ence (p < 0.05).
This experiment showed that heart rate decreased from
0.101to 0.00 in the case of alertness decrement.
No difference with the reference was observed in the case
of microsleeps for heart rate and inter-beat-intervals.
Heart rate variability (SDNN) increased only when mi-
crosleeps occurred, from 0.07 to 0.16, while no trend was
observed for the T-wave amplitude.
5.2.2. Eye activity
The analysis of eye activity data revealed that the blink
frequency increased when the driver was in a low alertness
state from 0.12 to 0.39. This trend was also observed during
microsleeps, though with a smaller amplitude (increment to
Eye closure followed the same trend as blink frequency,
with similar impairment for reduced alertness and mi-
crosleeps with an increase to 0.16 from 0.05. On the other
hand blink durations and PERCLOS were not impacted by
the vigilance level during this experiment.
1This means that the average value observed in the reference level (no
lapses in alertness as assessed with an EEG) is 0.10 standard deviations
above the reference obtained during the first five minutes of driving.
Figure 3: Evolution of the proportion (in percentage) of epochs indicating microsleep for the four different scenarios (taking into account the influence of
the sensation seeking level SS of the participant)
5.2.3. Skin conductance
Most skin conductance metrics were shown to be corre-
lated to the vigilance state in this experiment. Only the NSF
amplitude did not change with the vigilance level. SCL di-
minished in the case of alertness decrement from 0.30 to
The NSF rate seemed to be smaller during alertness
decrement - −0.26 compared with −0.09 - whereas a small
increase was observed during microsleeps.
during alertness decrement. The NSF rise time increased
from 0.35 to 0.45, while the NSF half-recovery time in-
creased from 0.33 to 0.40.
5.2.4. Driving simulator data
As for the data obtained from the car and the environ-
ment, it can be observed that the standard deviation of the
lane lateral shift decreased from −0.04 to −0.20 in case of
alertness decrement. The same trend was observed during
microsleep with −0.09.
For the speed, an increase from 0.22 to 0.27 was obtained
during microsleep. The standard deviation of the speed
tended to decrease during alertness decrement: −0.28 com-
pared to −0.22. But it led to a small increase in the case of
microsleeps to −0.19.
In terms of the standard deviation of the steering wheel
movement, only the alertness decrement had an effect, with
a decrease from −0.11 to −0.28.
Finally the Time to Lane Crossing tends to decrease both
during alertness decrement and microsleep with 0.12 and
0.18 compared with 0.29.
This experiment shows that the vigilance classification
presented in section 4 is correlated to the driving perfor-
mance. Indeed the driving performance and some physio-
logical markers of alertness are impaired when the alertness
During an alertness decrement, the heart rate tended to de-
crease and as a consequence IBIs tended to increase. This
suggests a reduced workload with no effortful mental pro-
cessing. This was expected during monotonous driving
since there is a low level of important information to pro-
Eye activity was also importantly impaired. Blink fre-
quency and eye closure were increasing while blink dura-
tion and PERCLOS remained the same. Eye closure in-
creases only due to the higher rate of blinks since the PER-
CLOS remains constant (blinks are removed for the PER-
CLOS computation). These results are consistent with a
hypovigilant driver as opposed to a drowsy or falling asleep
driver using Lal and Craig ’s categorisations. Indeed
drowsy and falling asleep drivers would also have longer
blink durations and eye closure would also increase be-
tween eye blinks. Furthermore the effects of monotony on
PERCLOS appear to be different from ones observed with
fatigue. The use of PERCLOS to identify and warn drivers
of the potential for a crash should be limited to fatigue and
this study suggests that PERCLOS would not be efficient in
the event of boredom.
Non specific electrodermal activity was reduced. This
shows a reduced readiness for action. This suggests that
driver’s attention is no longer focused toward the task to
perform (here a lane keeping task).
In terms of driving performance, the standard deviation
of the lateral position of the car decreased. In the litera-
ture, this is expected to increase as the driver becomes less
vigilant . Indeed, the car tends to go closer to the edge
of the road, and the driver takes more time to realise it and
Table 5: Effects of vigilance level on normalised surrogate measurements
Vigilance level (from EEG)
Heart rate variability
NSF rise time
NSF half-recovery time
SD lane lateral shift
SD steering wheel
−0.22 (< 0.01)
0.00 (0.03) ?
0.05 (0.03) ⇈
0.39 (0.07) ⇈
0.16 (0.03) ⇈
0.10 (0.06) ?
−0.26 (0.03) ?
0.45 (0.03) ⇈
0.40 (0.03) ↑
−0.20 (0.03) ?
−0.28 (0.02) ↓
−0.28 (0.03) ?
0.12 (0.06) ?
0.16 (0.03) ↑
0.26 (0.04) ⇈
0.16 (0.02) ⇈
−0.01 (0.03) ↑
−0.09 (0.02) ↓
0.27 (0.02) ↑
−0.19 (0.01) ↑
0.18 (0.04) ?
Values in the Table are: Normalised value (Standard Deviation)
-no statistical difference with reference
↑ increment < 0.1
⇈ increment > 0.1
↓ decrement < 0.1
? decrement > 0.1
correct the car trajectory. On monotonous roads (in terms
of road design), this standard deviation increases with time
as it does for driver hypovigilance. An explanation of our
result is that during the alertness decrement, the car moved
toward the edge slowly, therefore leading to a small stan-
dard deviation. This is consistent with the observation that
the standard deviation of the steering wheel movement de-
creased during that period and that the time to lane crossing
diminished. These other measures suggest that the driver
was inattentive. When the driver realised that the trajectory
had to be corrected, they moved rapidly toward the centre
of the road, which increased the standard deviation. The
latter part is in fact included in the vigilant level (since it
would be captured by the EEG). This explanation is sup-
ported by the fact that independently of the vigilance level,
the standard deviations tended to increase with time, as was
observed for the alertness decrement probability. Another
possible explanation of these results is the fact that the stan-
dard deviation of the lane positioning increases in the lit-
erature when random lane deviations are generated by the
driving simulator , which was not the case in this ex-
During this alertness decrement, it was also observed that
the speed was maintained more closely, which was the only
result not consistent with a hypovigilant driver.
The state characterised by microsleeps was also consistent
with driving impairment. During microsleeps, the heart rate
variability increased. Heart rate variability is expected to
decrease during effortful mental processing. This shows
that mental processing is reduced during short lapses in vig-
ilance, which is of great concern while driving.
These results were consistent with the reduced eye activity,
similar to the one observed during decreased alertness. In
terms of non-specific electrodermal responses, the only dif-
ference was a small decrease in NSF rate as compared with
the decrease observed during alertness decrement.
The same trend was observed for the standard deviation of
the lane position and for the time to lane crossing, but no
trend was observed in terms of steering wheel variability.
Speed tended to increase and to be more variable during
The vigilance states obtained from the analysis of EEG data
were shown to be highly correlated to many measures sug-
gesting a driver in a low vigilance state, with reduced abil-
ity to analyse the road surrounding them and with impaired
driving performance. Therefore driver’s inattention results
in driving performance impairment which can be detected
through measurements from the driver, the car and the en-
The research reported in this article features a number of
limitations which should be acknowledged.
A relatively small number of participants is used in this
experiment. However the sample size is statistically suf-
ficient to study effects of monotony on a range of surro-
gate measures. Females and university students are over-
represented in the samples used in this study. While the
likely impact of such bias is unclear, it still represents a po-
tential limitation. Nevertheless no statistical difference was
observed between genders.
University student received course credit while other par-
ticipants received money. Such methodology might have
dents. Nevertheless participants were not told that the aim
of the experiment what to monitor their alertness level dur-
ing monotonous driving. It is unlikely that paying student
would have increased their performance, since (i) the exper-
iment was very repetitive and they had to do it four times
(coming four times), (ii) the amount of money paid was not
dependant on the performance and (iii) participants did not
know what was investigated. Furthermore no statistical dif-
ference was observed for the group of participants paid and
the group of participants receiving course credit.
Another limitation of this study is that participants were
not asked to avoid drinking caffeine beverages before the
study. Nevertheless, only four participants had a coffee less
than two hours before the study (and more than one hour
before the study). All other participants did not have any
caffeinated drinks or more than two hours before the exper-
iment. Moreover no statistical difference has been observed
related to caffeine, which should insure the independence of
the result to effects of caffeine.
The findings in this study supports that monotony con-
tributes to changes in alertness in driving. Circadian rhyth-
micity and sleep restriction might be confounding variables
since participants were not driving at the same time and
their sleep before the experiment was not monitored (con-
sociated with time-of-day or self-reported amount of sleep
the previous night (all participants reported at least 6 hours
of sleep) was observed.
The impact of monotony on driver vigilance has not been
thoroughly studied, although it is an important factor con-
tributing to crashes. This experiment shows that during
a monotonous driving task two dimensions of monotony,
namely the road design monotony and the roadside vari-
ability, can lead to a rapid decrement of the alertness of
the driver. Such impairment is assessed with EEG analy-
sis and is associated with increasing probabilities of alert-
ness decrement and microsleeps as time increases. This
is emphasised when the road design is mainly straight (as
in the case of highways). Such inattention results in driv-
ing performance impairment which can be detected through
measurements from the driver, the car and the environment.
Such measurements include blinks, non-specific electroder-
mal responses rates, heart rate variability, variability of the
lateral lane positioning of the car and the time to lane cross-
ing. These measurements are consistent with symptoms of
a hypovigilant driver with decreased driving performance.
This experiment supports the vision to predict driver vigi-
lance in monotonous driving conditions through surrogate
measurements related to the driver, the car and the environ-
ment. Further research is required to model the occurrences
of hypovigilance to allow their prediction in real time.
The authors are indebted to Rebecca Michael, S´ ebastien
Demmel and Renata Meuter for their help with the design
and the collection of data used in this modelling study. The
the research program on monotony.
 Amditis, A., Andreone, L., Pagle, K., Markkula, G., Deregibus, E.,
Rue, M. R., Bellotti, F., Engelsberg, A., Brouwer, R., Peters, B.,
De Gloria, A., 2010. Towards the Automotive HMI of the Future:
Overview of the AIDE-Integrated Project Results. IEEE Transac-
tions on Intelligent Transportation Systems 11 (3), 567–578.
 Australian Transport Safety Bureau, 2008. Road crash casualties and
rates, Australia, 1925 to 2005. Tech. rep.
 Bastien, C. H., Ladouceur, C., Campbell, K. B., 2000. EEG char-
acteristics prior to and following the evoked K-Complex. Canadian
Journal of Experimental Psychology 54 (4), 255–265.
 Bekiaris, E., Amditis, A., Wevers, K., 2001. Advanced Driver Mon-
itoring: The AWAKE Project. In: 8th World Congress on ITS. Syd-
 Campagne, A., Pebayle, T., Muzet, A., 2005. Oculomotor changes
due to road events during prolonged monotonous simulated driving.
Biological Psychology 68 (3), 353–368.
 Cerezuela, G. P., Tejero, P., Choliz, M., Chisvert, M., Monteagudo,
M. J., 2004. Wertheim’s hypothesis on ‘highway hypnosis’: empiri-
cal evidence from a study on motorway and conventional road driv-
ing. Accident Analysis & Prevention 36 (6), 1045–1054.
 Critchley, H. D., Elliott, R., Mathias, C. J., Dolan, R. J., 2000. Neu-
ral activity relating to generation and representation of galvanic skin
conductance responses: A functional magnetic resonance imaging
study. Journal of Neuroscience 20 (8), 3033–3040.
 de Waard, D., 1996. The measurement of drivers’ mental workload.
PhD thesis, University of Groningen.
 Dinges, D. F., Mallis, M. M., Maislin, G., Powell, J. W., 1998. Eval-
uation of techniques for ocular measurement as an index of fatigue
and the basis for alertness management. National Highway Traffic
 Duta, M., Alford, C., Wilson, S., Tarassenko, L., 2004. Neural net-
work analysis of the mastoid EEG for the assessment of vigilance.
International Journal of Human-Computer Interaction 17 (2), 171–
 Eoh, H. J., Chung, M. K., Kim, S.-H., 2005. Electroencephalo-
graphic study of drowsiness in simulated driving with sleep depriva-
tion. International Journal of Industrial Ergonomics 35 (4), 307–320.
 Eysenck, H., 1967. The biological basis of personality. Thomas
 Eysenck, S., Eysenck, H., Barrett, P., 1985. A revised version of
the psychoticism scale. Personality and Individual Differences 6 (1),
 Fell, D. L., Black, B., 1997.Driver fatiguein thecity. Accident Anal-
ysis & Prevention 29 (4), 463–469.
 Fletcher, L., Petersson, L., Zelinsky, A., 2005. Road scene monotony
Intelligent Vehicles Symposium. pp. 484–489.
 Gillberg, M., Kecklund, G., Akerstedt, T., 1996. Sleepiness and per-
formance of professional simulator - comparisons between day and
night driving. Journal of Sleep Research 5, 12–15.
 Jap, B. T., Lal, S., Fischer, P., Bekiaris, E., 2009. Using EEG spec-
tral components to assess algorithms for detecting fatigue. Expert
Systems with Applications 36 (2, Part 1), 2352–2359.
 Lal, S. K., Craig, A., 2005. Reproducibility of the spectral com-
ponents of the electroencephalogram during driver fatigue. Interna-
tional Journal of Psychophysiology 55 (2), 137–143.
 Lal, S. K. L., Craig, A., 2002. Driver fatigue: Electroencephalogra-
phy and psychological assessment. Psychophysiology 39 (3), 313–
 Lal, S. K. L., Craig, A., Boord, P., Kirkup, L., Nguyen, H., 2003.
Development of an algorithm for an EEG-based driver fatigue coun-
termeasure. Journal of Safety Research 34 (3), 321–328.
 Larue, G. S., Rakotonirainy, A., Pettitt, A. N., 2010. Real-time per-
formance modelling of a sustained attention to response task. Er-
gonomics 53 (10), 1205–1216.
 Oron-Gilad, T., Ronen, A., Shinar, D., 2008. Alertness maintaining
tasks (AMTs) while driving. Accident Analysis & Prevention 40 (3),
 Pollock, V. E., Schneider, L. S., Lyness, S. A., 1991. Reliability of
topographic quantitative EEG amplitude in healthy late-middle-aged
and elderly subjects. Electroencephalography and Clinical Neuro-
physiology 79 (1), 20–26.
 Queensland Transport, 2005. 2003 road traffic crashes in Queens-
land. Tech. rep.
 Santamaria, J., Chiappa, K., 1987. The EEG of drowsiness in normal
adults. J. Clin. Neurophysiol. 4 (327-382).
 Scerbo, M. W., 1998. What’s so boring about vigilance? In: Hoff-
man, R. R., Sherrick, M. F., Warm, J. S. (Eds.), Viewing psychology
as a whole: The integrative science of William N. Dember. American
Psychological Association, pp. 145–166.
 Schell, A. M., Dawson, M. E., Nuechterlein, K. H., Subotnik, K. L.,
Ventura, J., 2002. The temporal stability of electrodermal variables
over a one-year period in patients with recent-onset schizophrenia
and in normal subjects. Psychophysiology 39 (2), 124–132.
 Schmidt, S., Walach, H., 2000. Electrodermal activity (EDA): State-
of-the-art measurement and techniques for parapsychological pur-
poses. Journal of Parapsychology 64 (2), 139–163.
 Stanton, N., Hedge, A., Brookhuis, K.A., Salas, E., Hendrick, H.W., Download full-text
2004. Handbook of Human Factors and Ergonomics Methods. CRC
 Steele, T., Cutmore, T., James, D., Rakotonirainy, A., 2004.
An investigation into peripheral physiological markers that predict
monotony.2004RoadSafety Research, PolicingandEducationCon-
 Thiffault, P., Bergeron, J., 2003. Fatigue and individual differences
in monotonous simulated driving. Personality and Individual Differ-
ences 34 (1), 159–176.
 Thiffault, P., Bergeron, J., 2003. Monotony of road environment and
driver fatigue: a simulator study. Accident Analysis & Prevention
35 (3), 381–391.
 Tomarken, A. J., Davidson, R. J., Wheeler, R. E., Kinney, L., 1992.
Psychometric properties of resting anterior EEG asymmetry: Tem-
poral stability and internal consistency. Psychophysiology 29 (5),
 Yamakoshi, T., Rolfe, P., Yamakoshi, Y., Hirose, H., 2009. A novel
physiological index for driver’s activation state derived from simu-
lated monotonous driving studies. Transportation Research Part C:
Emerging Technologies 17 (1), 69–80.
 Zuckerman, M., 1972. What is the sensation seeker? Personality trait
and experience correlates of the Sensation-Seeking Scales. Journal
of Consulting and Clinical Psychology 39 (2), 308–321.
 Zuckerman, M., 1994. Behavioral expressions and biosocial bases of
sensation seeking. Cambridge University Press, New York.