Journal of Psychosomatic Research 58 (2005) 61 71
Comparative sensitivity of a simulated driving task to self-report,
physiological, and other performance measures during
prolonged wakefulness
J. Todd Arnedt*, M. Ainsley C. Geddes, Alistair W. MacLean
Department of Psychology, Queens University, Kingston, ON, Canada
Received 28 April 2003; accepted 12 May 2004
Abstract
Objectives: The objectives of this study were to compare (1) the
sensitivity of simulated driving to self-report measures, nocturnal
sleep latency tests (SLTs), and an auditory vigilance task and (2)
urban and motorway driving. Methods: Healthy males 18 to 35
years maintained wakefulness for one night and were tested at 2400,
0230, 0500 and 0730 h. In Study 1 (n = 11), the SLTs were followed
by auditory vigilance and simulated driving tasks; in Study 2
(n = 18), the SLTs were preceded and followed by simulated driving
Keywords: Prolonged wakefulness; Sleepiness; Driving; Simulated driving; Sensitivity
Introduction
The contribution of sleepiness to large-scale disasters,
errors in the workplace, and motor vehicle accidents has
been increasingly recognized and studied [13]. Conditions
that have been associated with an increased risk for sleep-
related accidents include shift work, the use of alcohol or
other drugs, and sleep disorders, in particular, obstructive
sleep apnea, narcolepsy, and chronic insomnia [4]. Esti-
mates of the number of automobile crashes attributable to
sleepiness range widely, but there is converging evidence
that from 16% to 23% are sleep-related [2,5]. Among the
demographic and driving behaviour characteristics associa-
ted with increased risk for fall-asleep crashes are younger
* Corresponding author. Sleep and Chronophysiology Laboratory,
Department of Psychiatry, University of Michigan, 2101 Commonwealth
Blvd., Suite D, Ann Arbor, MI 48105. Tel.: 734 764 1234; fax: 734 764
1229.
E-mail address: tarnedt@med.umich.edu (J.T. Arnedt).
0022-3999/04/$ - see front matter D 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.jpsychores.2004.05.002
on motorway and urban routes. Results: In Study 1, speed vari-
ability, tracking variability, and driving off the road on the driving
simulator had comparable sensitivity to d V on the auditory vigilance
task. In Study 2, driving performance was consistently worse on the
motorway route. Conclusion: The driving simulator was equally
sensitive to another performance measure during prolonged
wakefulness and impairments were greater with motorway driving.
D 2005 Elsevier Inc. All rights reserved.
age, male gender, higher education level and annual mileage
driven, and more time spent on motorway roads [68].
Clearly, there exists a need to develop measures adequate
to detect the presence of sleepiness. Currently, introspective
estimates and electrophysiological measures, such as the
Multiple Sleep Latency Test [9], are the primary instruments
used for this purpose. However, Dinges and Kribbs [10]
have argued that measures of functional capacity might be
more suitable for detecting the consequences of increased
sleepiness. A number of tests are well known to be sensitive
to conditions provoking sleepiness, such as sleep loss, and
the characteristics of such tests have been widely described.
These characteristics include long duration [11,12], no
knowledge of results [13], complex yet uninteresting [14],
monotonous [15], familiar [13], and experimenter paced
[14,16]. These features have been used for the development
of a number of performance tests, including vigilance [11],
simple [17] and choice [18] reaction time, logical reasoning
[19], and memory [20].
Broadbent [21] has suggested that a battery of tests
assessing a variety of task-relevant characteristics should be
62 J.T. Arnedt et al. / Journal of Psychosomatic Research 58 (2005) 6171
the approach of choice when assessing complex perform-
ance. Such tasks are generally well tolerated by student
subjects; however, patients frequently indicate that they lack
face validity and may therefore be less inclined to give their
best performance or may refuse to participate altogether. In
addition, these tests frequently have large practice effects,
making them impractical in clinical settings [22,23], and
their relative sensitivity to conditions causing performance
decrements in comparison with other measures of sleepiness
is unknown. The ideal performance measure would have
high face validity, involve a highly practised skill, and
assess a number of relevant aspects of functional capacity.
For these reasons, a driving simulator provides a possible
alternative to the use of a test battery. Driving simulators of
various levels of sophistication have been used to examine
the consequences of a variety of conditions, including sleep
deprivation [24,25], alcohol ingestion [24,26,27], and sleep
disorders [28,29].
The York Driving Simulator (York Computer Technolo-
gies, Kingston, ON, Canada) is a low-cost, portable driving
simulator that offers a flexible data output, allowing for the
measurement of a number of driving-related constituent
skills, such as vehicle control (e.g., steering and pedal
control) and on-the-road measures [e.g., lane position
(tracking) and speed deviation]. Such a device could prove
valuable in a clinical setting to provide objective information
regarding a patients ability to operate a motor vehicle safely.
Currently, most clinicians rely on self-report and physiolog-
ical measures, such as the Multiple Sleep Latency Test, to
guide such decisions. There seems considerable potential,
however, for the use of performance measures to provide a
more objective and relevant assessment. Driving perform-
ance on the York Driving Simulator has previously been
found to deteriorate following partial sleep deprivation (4.5-h
time in bed) and to ameliorate with slow-release caffeine
[30]. In addition, minimal practice effects on this simulator
were found in a sample of young participants with little
driving experience [31]. We sought to determine the degree to
which the York Driving Simulator is sensitive to the effects of
prolonged wakefulness in comparison with other self-report,
physiological, and behavioural measures of sleepiness.
This paper describes two studies. The purpose of the first
study was to compare the sensitivity of the York Driving
Simulator to other measures known to be sensitive to
sleepiness and to decrements in performance. The aim of the
second study was to compare the decrements occurring
while driving under urban and motorway conditions to
evaluate more closely the degree to which task character-
istics affect simulator performance.
Study 1
The aim of this study was to compare the sensitivity of
the York Driving Simulator to self-report measures, repeated
nocturnal sleep latency tests (SLTs), and a 30-min auditory
vigilance task under conditions likely to produce perform-
ance decrements.
Methods
Participants
Eleven male participants recruited from the Introductory
Psychology Subject Pool (mean age = 20.4 years, S.D. = 2.2,
range 1825; mean Epworth Sleepiness Scale [32] score 8.6,
S.D. = 4.7, range 218), who gave their informed consent,
experienced one night of prolonged wakefulness. Inclusion
criteria were a regular sleep schedule, neither extreme
evening nor morning type [33], good health, the absence of
medications that affect daytime sleepiness or the sleep/wake
cycle, and willingness to comply with the conditions of the
experiment. One participant had an Epworth Sleepiness
Scale score of 18, which exceeds the clinical cutoff for
excessive daytime sleepiness [32]. Participants were asked to
maintain a regular sleep schedule, with bedtime between
2300 and 0100 h and waking between 0700 and 0900 h, for
the week prior to the study and to refrain from the use of
caffeine for 24 h and alcohol and other drugs for 48 h prior to
the study. Compliance with the protocol was confirmed using
sleep logs and an activity/substance use diary. Consistent
with the policies of the Department of Psychology, partic-
ipants earned a 1% increase to their class mark in
Introductory Psychology for each 1 h of participation in
departmental research activities, up to a maximum of 5%.
Accordingly, for this study, they were awarded an extra 5% to
their class mark. Of 14 participants who volunteered for the
study, two did not meet the inclusion criteria and one became
ill on the test night for reasons unrelated to the experiment.
Dependent measures
Self-report
The Stanford Sleepiness Scale [34] and a modified form
of the Stanford Scale [35] were used to assess subjective
sleepiness. The latter test yields two factors provisionally
described as, respectively, ban energic or activating factorQ
(Factor 1; range 013) and a factor brelated to conscious-
ness, sleepiness, and a loss of control over remaining awakeQ
(Factor 2; range 011; Ref. [35], p. 38). Higher scores on
both factors indicate higher levels of subjective sleepiness.
Physiological
Repeated nocturnal SLTs were conducted consistent with
procedures outlined for the Multiple Sleep Latency Test [9].
Continuous monitoring of electroencephalogram (EEG),
electrooculogram (EOG), and electromyogram (EMG) was
conducted for each test. The test was terminated after 2 min
of Stage 2 sleep, the first epoch of any other stage of sleep, or
after 20 min without sleep onset. The resulting records were
scored independently by two scorers, and any discrepancies
 J.T. Arnedt et al. / Journal of Psychosomatic Research 58 (2005) 6171
not resolved by consultation were decided by a third scorer.
Electrophysiological activity was also recorded while
participants performed a 30-min auditory vigilance task,
and records were scored in 30-s epochs using standard
procedures [36]. A score of 30 min was given to participants
who showed no signs of sleep onset on this task. Consistent
with guidelines for the Multiple Sleep Latency Test, the
dependent variable for both physiological tests is the latency
from lights out to the first epoch scored as Stage 1 sleep.
Performance
The auditory vigilance task [37] required participants to
listen to a series of blocks of 10 tones of constant pitch
(1000 Hz) with a fixed intertone interval of 5 s. Noise tones
had a 1.0-s duration, while signal tones were 1.2-s long.
Each had an 82-dB SPL intensity and was played against a
background noise of 32 dB SPL. The signal tones occurred
once in each block of 10 tones and were randomly
interspersed among the nine noise tones. Participants were
instructed to sit comfortably and press a microswitch with
the preferred hand when they thought they heard a signal
tone. The dependent variables reported here are the widely
used signal detection parameters d prime (d V) and Beta ( b).
The former is typically interpreted as an individuals
intrinsic ability to discriminate signal from noise tones, with
higher scores representing greater sensitivity to signal tones.
On the other hand, b is an index of an individuals decision
criterion, or degree of caution, used to respond to a signal.
Higher b scores indicate a more cautious response style.
The York Driving Simulator runs on a personal computer
system, and an electronic interface provides driver inputs to
the software (steering, brakes, and accelerator). Simple line
graphics are used to portray objects and to induce a sense of
apparent motion by using real-time perspective generation.
The simulator presents a forward view from the drivers seat
of either an urban or a motorway road scene (Fig. 1), with
standard lane markings and signs and signals appropriate
for the specified road environment. The participant navi-
gates within the computer driving environment interactively
just as one would navigate a car in the real world, while
being told to obey all road signs and traffic signals and to
maintain the car in the appropriate lane on the road. The
participants drove in a simulated night environment and
were instructed to drive as far as they could in 30 min while
keeping the cars position in the middle of the right hand
lane and obeying traffic lights, speed restrictions, stop signs,
and other road signs. Each 30-min drive consisted of an
area of urban driving followed immediately by motorway
driving, each section taking approximately 15 min.
Vehicle and road dimensions in the York Driving
Simulator are measured in simulator units, where 1 ft
(0.3048 m) = 100 simulator units (su). The simulated car
was 600 su in width and 1000 su in length. The entire
simulated road is 3000 su wide; each lane on the two-lane
road is 1100 su wide, and 400 su on either side is devoted to
the road shoulder.
63
Urban Route
Motorway Route
Fig. 1. The urban and motorway routes of the York Driving Simulator.
The dependent variables on the simulated driving task
included tracking, measured as the deviation (in m) of the
centre of the vehicle from the centre of the right hand
lane; tracking variability, the standard deviation of track-
ing; speed deviation, calculated as the difference in speed
(in km/h) of the vehicle from the speed limit for that
stretch of road; speed variability, the standard deviation of
speed deviation; and off-road events, measured as the
number of occasions on which the vehicle left the
simulated road.
Procedure
Prior to the experiment, participants attended two train-
ing sessions. At each training session, they completed the
self-report assessments and were given a 15-min practice
trial on the auditory vigilance task and a 20-min trial on the
driving simulator. At the first training session, the partic-
ipants also had electrodes attached as they would for the
SLTs. Prior to the first test session on the experimental
night, participants were given a further 30-min practice trial
on the driving simulator and a further demonstration of the
auditory vigilance task.
Test sessions took place at 2400, 0230, 0500, and 0730 h;
each session took about 90 min to complete. The
participants came to the laboratory about 2 h prior to the
first session to be prepared for continuous recording of EEG
(C3/A2, O2/A1), referential EOG, and submental EMG
activity. They remained awake under supervision in the
laboratory, indulging in quiet activities between test
sessions. At each test session, the SLT was given first
followed by the two performance tests (auditory vigilance
and simulated driving tests); the order of the latter two tests
was counterbalanced across test sessions. Self-report sleepi-
64 J.T. Arnedt et al. / Journal of Psychosomatic Research 58 (2005) 6171

ness ratings were obtained prior to each SLT and each Table 1
Mean and standard error of the mean (S.E.M.) on the self-report,
performance test.
physiological, auditory vigilance, and simulated driving measures across
the four test times in Study 1
Test time 2400 h 0230 h 0500 h 0730 h
Results
Self-report measures
Stanford Mean 3.5 4.2 5.0 5.4
In general, the data were analysed using repeated Sleepiness S.E.M. 0.4 0.4 0.3 0.3
measures analysis of variance, in which time of testing Scale
(2400, 0230, 0500, and 0730 h) was extracted as a within- Modified Stanford Mean 2.0 3.2 5.1 6.3
subject factor. For the Stanford and Modified Stanford Scale: Factor 1 S.E.M. 0.9 0.9 0.9 0.8
Modified Stanford Mean 9.0 10.7 12.1 12.5
Sleepiness Scales, a second within-subject factor was also
Scale: Factor 2 S.E.M. 1.0 0.8 0.2 0.2
extracted corresponding to the time of administration (prior Physiological measures
to the SLT and the two performance tests). In the case of Sleep latency on Mean 13.8 6.4 4.4 2.4
the auditory vigilance and the simulated driving tasks, the SLT (min) S.E.M. 1.9 1.6 1.6 0.5
time-on-task was extracted by analysing the data in 5-min Sleep latency Mean 28.0 23.8 22.1 14.4
during the S.E.M. 1.4 3.2 3.2 3.4
blocks. Where appropriate, HuynhFeldt [38] corrections
Auditory
for lack of sphericity were applied, and the data were Vigilance
further analysed using orthogonal polynomials or the Test (min)
Tukey Honestly Significant Difference Test. The level of Auditory vigilance task
significance for all analyses was set at .05. dV Mean 2.63 2.02 1.65 1.17
S.E.M. 0.28 0.27 0.28 0.24
b Mean 3.36 3.84 3.54 3.28
Sensitivity of tests to experimental conditions S.E.M. 0.53 0.46 0.53 0.52
Simulated driving task
The summary statistics for all of the measures at the Tracking Mean 0.30 0.43 0.46 0.52
four test sessions are reported in Table 1. deviation from S.E.M. 0.09 0.06 0.09 0.09
lane centre (m)
Tracking Mean 0.46 0.55 0.58 0.64
Self-report variability (m) S.E.M. 0.03 0.03 0.06 0.06
With increasing time awake, participants reported Deviation from Mean 1.1 0.0 1.2 0.3
themselves as increasingly sleepy on the Stanford Sleepi- posted speed S.E.M. 1.1 1.6 0.6 1.2
ness Scale [ F(3,30) = 16.23, P = .0001] and on both Factors (km/h)
Variability of Mean 5.6 7.6 7.8 8.9
1 [ F(3,30) = 15.07, P b.0005] and 2 of the Modified
speed deviation S.E.M. 0.8 1.2 1.3 1.1
Stanford Scale [ F(3,30) = 8.58, P = .003]. There was also (km/h)
a small, but statistically significant, difference between Off-road incidents Mean 0.2 0.9 1.0 1.5
reports of subjective sleepiness [Stanford Sleepiness Scale: (number/5 min) S.E.M. 0.1 0.3 0.3 0.4
F(2,20) = 26.33, P = .0001; Modified Stanford Scale:
Factor 1: F(2,20) = 20.00, P b.0005; Factor 2: F(2,20) =
12.26, P b .0005] before the SLT and the performance tests, With increasing time awake on the simulated driving
but no difference in sleepiness ratings between the two task, the drivers drove increasingly to the left of the centre
performance tests. For example, for the Stanford Sleepiness of their lane [ F(3,30) = 6.81, P b.005], the variability of
Scale, the mean values before the SLT and the auditory their tracking increased [ F(3,30) = 9.54, P b.005], the
vigilance and the simulated driving tasks were 4.1, 4.8, and speed variability increased [ F(3,30) = 4.96, P b.009], and
4.7, respectively. the number of times that they drove off the road increased
linearly [ F(1,9) = 7.70, P = .022]. With increasing time-on-
Physiological measures task during each session, drivers again tracked to the left
On the SLT, latency to Stage 1 sleep decreased with of the lane centre [ F(5,50) = 7.78, P = .01], and tracking
increasing time awake [ F(3,30) = 15.80, P b.0005], as was variability increased [ F(5,50) = 16.72, P b.005]. At the
true also for the auditory vigilance task [ F(3,30) = 6.31, beginning of the task, drivers tended to drive below the
P = .002]. speed limit, but, as task duration increased, they increased
their speed until they were driving above the posted limit
Performance [ F(5,50) = 5.04, P b.03].
On the auditory vigilance task, there was a decline in d V
with increasing time awake [ F(3,30) = 15.70, P = .0001] Comparative sensitivity
and time-on-task [ F(5,50) = 9.05, P = .0001]. Beta did not
change significantly with either increasing time awake To compare the various dependent measures on a
[ F(3,30) = 0.49, P N.05] or time-on-task [ F(5,30) = 0.94, common scale, Cohens standardised difference measure
P N.05]. [39] was used. This was defined as (mean at other
 J.T. Arnedt et al. / Journal of Psychosomatic Research 58 (2005) 6171
times mean at 2400 h)/pooled standard deviation. As
suggested by Cohen [39], a standardised difference of 0.2
was considered small, 0.5 was considered medium, and
0.8 was considered a large difference. The results are
shown in Fig. 2.
Fig. 2. Cohens standardised differences (Cohens d) at each test session
(2400, 0230, 0500, and 0730 h) for self-report (SSS, MSS Factor 1, and
MSS Factor 2), physiological (Stage 1 latency on the SLT and auditory
vigilance task), and performance (d V on the auditory vigilance task, mean
tracking, tracking variability, speed variability, and off-road occurrences on
the simulated driving task) measures. Cohens d was calculated at each test
session using the following formula: (mean at other times mean at 2400
h)/pooled standard deviation.
65
Discussion
While the driving simulator was not the most sensitive
test, it did appear to reach a level of sensitivity sufficient
to warrant further investigation. These findings are
considered in more detail in the General Discussion below.
However, it was first necessary to resolve an important
confound in the present study. An examination of the
driving simulator data for time-on-task effects indicates that
the greatest change in performance occurred during the
second half of the 30-min test. This happened to be the time
when participants were driving on the motorway route.
There is thus a confound between driving conditions and
time-on-task. To determine which of these factors is
responsible for the decline in performance, a second study
was carried out as described below.
Study 2
The aim of the second study was to eliminate the
confound of route and task length by counterbalancing the
order of presentation of urban and motorway routes.
Methods
Participants
Participants were recruited from the Introductory Psy-
chology Subject Pool. Twenty healthy males, who agreed to
comply with the requirements of the study, gave their
informed consent to participate. Inclusion criteria were a
valid drivers license, a regular sleep schedule, a moderate
to good sleeper [40], neither extreme morning nor evening
type ([33]; mean 48.9, S.D. = 8.7, range 3262), no evidence
of excessive daytime sleepiness ([32]; mean 6.8, S.D. = 2.3,
range 412), and no medications with sleep-altering
qualities. One participant included in the sample had an
Epworth score above the clinical cutoff for excessive
daytime sleepiness [32]. Of the original 20 participants,
two participants were excluded from the analyses because of
failure to comply with the instructions, resulting in a final
sample of 18 participants (mean age = 21.9 years, S.D. = 3.5,
range 1932 years).
Participants were instructed to refrain from using alcohol
and other drugs for 48 h and caffeine for 24 h prior to the
experimental night. They were also asked to maintain a
regular sleep schedule, retiring between 2300 and 0100 h
and rising between 0700 and 0900 h, beginning 48 h prior
to testing. Compliance with the conditions was assessed
using sleep logs and an activity/substance use diary.
Participants were either paid CAD$30 or awarded a 5%
addition to their overall mark in the Introductory Psychol-
ogy course for participation.
66 J.T. Arnedt et al. / Journal of Psychosomatic Research 58 (2005) 6171
Dependent measures
Self-report
The self-report variables were the same as those used in
Study 1.
Physiological
The SLT procedure was as described in Study 1.
Electrophysiological recordings were also made during the
simulated driving task; however, these data will not be
reported here. As in Study 1, the dependent variable was the
latency to Stage 1 sleep.
Performance
The York Driving Simulator was used to assess driving
performance. The urban and motorway routes were similar
to those used previously. As in Study 1, the dependent
variables were mean tracking, tracking variability, speed
deviation, speed variability, and the number of off-road
occurrences.
Procedure
Forty-eight hours prior to testing, the participants came
to the Sleep Laboratory for a 30-min training session on the
driving simulator, 15 min each on the urban and motorway
routes. A second similar training session took place at the
beginning of the experimental night.
On the experimental night, participants arrived at the
Sleep Laboratory between 2100 and 2130 h, where
electrodes were applied for continuous monitoring of
EEG (C3/A2, O2/A1), EOG, and EMG activity. Following
training, four 90-min test sessions took place at 2400,
0230, 0500, and 0730 h. For each test session, the
participants drove for 30 min on either the urban or
motorway route of the driving simulator, underwent the
SLT, and then drove the other route for another 30 min.
The order of the urban and motorway routes was counter-
balanced across participants. Participants completed the
Stanford Sleepiness and the Modified Stanford Scales
before and after each simulated drive, and before each
SLT. During breaks, the participants engaged in quiet
activities in the laboratory under supervision. At the
conclusion of testing, the participants were debriefed and
remunerated for their participation.
As part of a protocol separate from this study, 10
randomly selected participants wore a behavioural device,
the Magellan Monitor, on their nondominant wrist
throughout the experimental night, except during the
SLT. The monitor emitted vibratory stimuli, on average,
every 45 s, with a minimum and maximum interstimulus
interval of 30 and 60 s, respectively. The participants were
required to respond to the stimuli as quickly as possible by
flexing their index finger to activate a piezoelectric
response sensor.
Statistical analyses
In general, the data from the four test sessions were
analysed using split plot analyses of variance, with the
wearing of the Magellan Monitor (Magellan, no Magellan)
extracted as a between-subjects source of variance, and
test session (2400, 0230, 0500, and 0730 h) extracted as
a within-subjects source of variance. For the self-report
measures, two analyses were conducted. The first ex-
tracted the time of questionnaire administration (before
and after the driving sessions and before the SLT) as an
additional within-subjects factor to examine self-reported
changes in sleepiness with prolonged wakefulness. The
second analysis extracted route (motorway and urban) and
time of administration (before and after the driving
sessions) as additional within-subjects factors. For the
simulated driving task, route (motorway and urban) and
the six 5-min blocks of the task were extracted as within-
subjects effects.
Univariate F ratios were evaluated using the Huynh
Feldt correction factor for sphericity [38]. Orthogonal
polynomials and Bonferonni-adjusted t tests were used
where appropriate. The level of significance for all analyses
was set at .05.
Results
Effects of prolonged wakefulness
Self-report
With increasing time awake, participants rated them-
selves as more sleepy on all three self-report sleepiness
measures: the Stanford Sleepiness Scale [ F(3,48) = 86.35,
P b.001] and Factors 1 [ F(3,48) = 40.69, P b.001] and 2
[ F(3,48) = 36.68, P b.001] of the Modified Stanford Scale.
Post hoc analyses indicated that self-reported sleepiness
ratings on all three scales increased significantly from
2400 to 0730 h, with significant differences between
ratings at each successive test time, except between 0500
and 0730 h ( P b.001). There were also significant time
effects on the Stanford Sleepiness Scale [ F(2,32) = 21.71,
P b.001] and on Factors 1 [ F(2,32) = 12.78, P b.001] and
2 [ F(2,32) = 47.27, P b.001] of the Modified Stanford
Scale. In all cases, ratings were significantly higher before
the SLT and after the driving task than before the driving
task ( P b.01). Sleepiness ratings before the SLT for Factor
2 of the Modified Stanford Scale were also significantly
higher than after the simulated driving task ( P = .013).
There were no significant TimeMagellan interactions or
three-way interactions involving test, time, and Magellan,
but a significant TestMagellan interaction was found for
Factor 2 of the Modified Stanford Scale [ F(3,49) = 3.41,
P b.05], and TestTime interactions were evident for both
the Stanford Sleepiness Scale [ F(6,96) = 2.92, P b.05] and
 J.T. Arnedt et al. / Journal of Psychosomatic Research 58 (2005) 6171 67

Factor 2 of the Modified Stanford [ F(6,96) = 3.65, Table 2

Mean and standard error of the mean (S.E.M.) for self-report and driving
P b.01].
measures on the urban and motorway routes
Route type Urban Motorway Cohens d
Physiological measures
On the SLT, sleep onset latency to Stage 1 decreased Self-report measures
Stanford Mean 3.9 4.0 0.10
with increasing time awake [ F(3,48) = 26.97, P b.001].
Sleepiness Scale S.E.M. 0.2 0.2
There were no significant main effects of Magellan or Modified Stanford Mean 9.1 9.4 0.15
Test  Magellan interactions. Scale: Factor 1 S.E.M. 0.4 0.5
Modified Stanford Mean 3.1 3.3 0.11
Driving performance Scale: Factor 2 S.E.M. 0.4 0.4
Simulated driving task
There were no main effects of wearing the Magellan
Mean tracking (m) Mean 0.13 0.20 1.35
monitor on any of the dependent variables of the driving S.E.M. 0.01 0.01
task. Consistent with the findings from Study 1, with Tracking variability (m) Mean 0.48 0.61 0.50
increasing time awake, the participants drifted to the left of S.E.M. 0.01 0.01
the centre of the lane [ F(3,48) = 10.14, P b.001], tracking Speed deviation (km/h) Mean 1.43 5.97 1.15
S.E.M. 0.16 0.41
became more variable [ F(3,48) = 16.97, P b.001], speed
Variability of speed Mean 6.27 12.25 1.66
variability increased [ F(3,48) = 21.80, P b.001], and off- deviation (km/h) S.E.M. 0.13 0.21
road incidents increased linearly [ F(1,16) = 4.85, P b.05]. Off-road incidents Mean 0.20 0.46 0.28
Post hoc analyses indicated that mean tracking, tracking (number/5 min) S.E.M. 0.04 0.05
variability, and speed variability were worse at the 0730-h
test time relative to both the 2400- and 0230-h test times
( P b.003). Additionally, mean tracking and speed variability Driving performance
were worse at 0730 h compared with the test at 0500 h The significant differences in driving parameters between
( P b.008). There was a significant TestMagellan interac- the urban and motorway routes are shown in Fig. 3.
tion only for speed variability [ F(3,48) = 3.78, P b.05], Performance was significantly worse on the motorway
which revealed worse performance at the 0730-h test time route for all five simulated driving variables: mean tracking
in the group wearing the device. Performance on the [ F(1,16) = 8.35, P = .01], tracking variability [ F(1,16) =
following dependent variables also deteriorated with increas- 26.23, P b.001], mean speed deviation from the posted
ing time-on-task: mean tracking [ F(5,80) = 5.17, P = .001], speed signs [ F(1,16) = 70.38, P b.001], speed variability
tracking variability [ F(5,80) = 10.95, P b.001], speed varia- [ F(1,16) = 148.33, P b.001], and the number of off-road
bility [ F(5,80) = 3.55, P b.01], and off-road occurrences incidents [ F(1,16) = 15.47, P = .001]. Across the six 5-min
[ F(5,80) = 3.50, Pb.05]. Mean tracking [ F(15,240) = 2.73, blocks of the driving task, off-road events [ F(5,80) = 2.53,
P b.01], tracking variability [ F(15,240) = 4.31, P b.001], P b.05] increased more rapidly on the motorway route. There
speed variability [ F(15,240) = 4.27, P b.001], and off-road were no Route  Test or Magellan  Route interactions, but a
events [ F(15,240) = 2.74, P b.05] deteriorated more rapidly significant three-way interaction involving Magellan, route,
during the 30-min driving task as the night progressed. There and block was found for off-road incidents [ F(5,80) = 2.66,
were no significant Magellan  Block interactions, but speed P b.05].
deviation [ F(15,240) = 2.69, P b.01] and speed variability
[ F(15,240) = 1.71, P = .05] deteriorated more rapidly across
the 30-min driving task at the later test times for participants General Discussion
wearing the Magellan monitor.
The primary objective of the first study was to evaluate
Effects of route type on measures of driving performance the sensitivity of a face valid measure of driving ability, the
York Driving Simulator, relative to measures that are widely
Descriptive data for the self-report measures and the used and known to be sensitive to the effects of sleep loss.
simulated driving task with effect sizes (Cohens d) are The second study was conducted to clarify an important
presented in Table 2. confound in the first study, as well as to examine more
closely the degree to which task characteristics (in
Self-report particular, monotony) affect simulated driving performance
Ratings on the three self-report sleepiness measures did in a manner consistent with other performance measures.
not differ by route type, but ratings were consistently higher
after the driving task than before: Stanford Sleepiness Scale Sensitivity of the York Driving Simulator
[ F(1,16) = 80.30, P b.001] and Factors 1 [ F(1,16) = 29.90,
P b.001] and 2 [ F(1,16) = 94.69, P b .001]. There were no By placing the various sleepiness measures on a common
significant two- or three-way interactions involving Magel- metric in Study 1, it was possible to compare their relative
lan, route, and time. sensitivity to the experimental manipulation. The findings
68 J.T. Arnedt et al. / Journal of Psychosomatic Research 58 (2005) 6171

Fig. 3. Comparison of overall driving performance on urban and motorway routes of the York Driving Simulator. Results for the five dependent measures of the
driving taskmean tracking, tracking variability, speed deviation, speed variability, and off-road incidentsare plotted as meansFS.E.M.s.

were generally consistent with the model of sleepiness
outlined by Carskadon and Dement [41]. Sleep latency on
the SLT was the most sensitive measure, presumably due, in
large part, to the fact that the test is conducted in an
environment that is explicitly devoid of external stimula-
tionthe lights are dimmed, the participant is supine, noise
levels are low, room temperature is constantand the
participant is instructed to try to fall asleep. As such, the
SLT is the most direct measure of physiological or basal
sleepiness. The four parameters of the York Driving
Simulator that showed significant deterioration with increas-
ing time awake in both studiesmean tracking, tracking
variability, speed variability, and off-road eventswere
considerably less sensitive than the SLT and slightly less
sensitive than the other performance measure, d V on the
auditory vigilance task. These findings too would be
predicted by the model of Carskadon and Dement [41].
The auditory vigilance task was administered while the
participant sat upright in a chair and had to make a simple
motor response; in contrast, the driving task required
sustained attention to a more complex stimulus and the
manipulation of steering wheel and pedals.
Self-report and performance measures
As expected, self-reported sleepiness increased monot-
onically across the night of prolonged wakefulness in both
studies. These findings are consistent with several studies
 J.T. Arnedt et al. / Journal of Psychosomatic Research 58 (2005) 6171
showing increased subjective sleepiness during the night
assessed in experimentally induced sleep deprivation studies
(e.g., Ref. [42]), constant routine protocols (e.g., Ref. [43]),
and studies of shift workers in a variety of industries
[44,45]. In addition, it has been shown repeatedly that
alertness follows a self-sustained circadian pattern reaching
its nadir in the early morning around 0500 h [46,47]. In-
terestingly, although objective performance was consistently
worse on the motorway route of the driving task, there were
no significant differences in self-reported sleepiness
between the motorway and urban routes. This finding
suggests that the degree of impairment associated with
sleepiness may be underrecognized, which has implications
for decisions about ones capacity to operate safely a motor
vehicle. The inability to appreciate fully the dangers
associated with drowsy driving has also been highlighted
by other researchers [48].
Consistent with previous studies [49,50], latency to Stage
1 sleep onset declined linearly across the night on the SLT.
The decline in dV on the auditory vigilance task with
increasing time awake is consistent with previous findings
and has typically been interpreted as indicating a reduction
in the intrinsic capacity to discriminate a signal [5153].
Naitoh [54] has cautioned against a simple interpretation of
the signal detection measures dV and b and has questioned
the appropriateness of applying signal detection theory to
sleep loss studies. He argues that independent evidence,
such as physiological measures, is necessary to ensure bthat
the subject observe and attend to the stimuliQ (Ref. [54],
p. 360). Because no such evidence was gathered in this
study, the findings may indicate either a true reduction of
intrinsic capacity or a loss of attentional capacity.
Simulated driving performance during
prolonged wakefulness
Simulated driving performance deteriorated progres-
sively across the night in both studies. More specifically,
with increasing time awake, participants drove closer to the
middle of the road, tracking and speed became more
variable, and the number of off-road incidents increased.
This overall decline in driving ability during periods of
increased sleepiness is consistent with the findings from
both simulated driving [55] and nondriving performance
studies conducted during the circadian trough of alertness
[56,57]. The findings from the present studies are also
consistent with the well-identified nighttime peak for
sleepiness-related motor vehicle accidents [8,58,59] and
with the occurrence of several large-scale catastrophic
events during the early morning hours [3]. One significant
shortcoming with the current study is that the experimental
design did not permit a systematic analysis of the relative
contribution of circadian and homeostatic influences to the
observed decrements in performance across the night. This
is an important issue, and one that could be addressed in
future studies experimentally by, e.g., testing participants at
69
the same circadian phase, but altering the length of
wakefulness prior to testing.
In addition to the deficits observed across the night, there
were characteristic changes in driving performance within
the 30-min driving task. With increasing time-on-task,
participants again drifted towards the middle of the road,
and their tracking became more variable. These findings are
consistent with previous studies showing deviations in
lateral position as a result of prolonged driving [60,61].
Although the driving task used in the present studies was
considerably shorter (30 min) in duration than those in the
latter two studies (2 h), it is likely that time-on-task fatigue
was a significant contributor to the observed deterioration in
performance. Others have noted that fatigue is manifested
more rapidly in driving simulators than in actual driving
conditions (e.g., Ref. [55]), and participants in the present
study rated themselves as significantly sleepier after the
driving task than before. From a practical perspective, a test
that is sensitive to detecting the decrements related to
sleepiness and of a relatively short duration is more likely to
be preferred by both clinicians and patients.
Contrary to the present findings, Lenne et al. [25,55]
have found that participants drifted towards the outer edge
rather than towards the midline of the road with increasing
task duration. Their simulated driving task may have been
more alerting, however, because participants were required
to perform a secondary reaction time task while driving, and
there were other simulated cars on the road. In a more
monotonous driving task without other simulated cars, the
sleepy driver may approach the road midline in an effort to
avoid an off-road incident. Stewart and MacLean [62] have
also observed that when other cars are introduced, partici-
pants drive towards the outer edge of the road, perhaps, to
reduce the likelihood of a collision with another vehicle.
Effects of route type on driving performance
The results from Study 2 demonstrated that driving
performance was consistently worse on the motorway route.
More specifically, participants ability to maintain a
consistent road position and a constant speed was more
variable, their speed deviated more from the posted speed
limit, and they drove off the motorway road more
frequently. These findings may have been due to the greater
monotony of the motorway route, which has few turns and
no road signs or stop lights. In addition to being consistent
with predictions, the deterioration in driving performance is
supported by accident data demonstrating a 2025% higher
incidence of sleep-related motor vehicle accidents on
motorway than on urban routes [2,48]. Studies of the effects
of sleep loss on performance have consistently shown that
monotonous tasks are more sensitive to the effects of
sleepiness than are intrinsically more interesting tasks
[11,13,15,63]. Moreover, the greater decline in performance
on the motorway route is consistent with the model of
Carskadon and Dement [41] of the constructs underlying
70 J.T. Arnedt et al. / Journal of Psychosomatic Research 58 (2005) 6171
sleepiness because physiological sleepiness is theoretically
unmasked in the absence of alerting factors [64]. The results
of the second study, in particular, suggest that simulated
motorway routes may be more sensitive to the decrements in
performance observed with sleepiness than urban routes are
and may therefore be more appropriate for assessing
driving-related impairments associated with sleepiness.
Finally, half of the participants in Study 2 were required
to respond to the Magellan Monitor during the driving task,
which could have resulted in better performance in this
group due to increased arousal. There were, however, no
main effects of the Magellan on driving performance, and
the only significant two-way interaction found worse
performance in the group wearing the device. Thus, the
Magellan seemed to have little direct effect on arousal level
while driving. Comparing the concordance of the Magellan
data with the other measures of sleepiness will reveal
whether this device might serve as an on-board real-time
indicator of a drivers level of alertness.
In summary, the findings from both investigations
suggest that the York Driving Simulator is sensitive to the
decrements in performance evident with increasing time
awake, and it appears to have at least comparable
sensitivity under these conditions to an auditory vigilance
task. Moreover, consistent with what has been demon-
strated with other performance measures, increasing the
monotony of the task produces an even more marked
deterioration in driving performance. The simulated
driving task offers advantages over self-report, physio-
logical, and other performance measures in the assessment
of driving performance, such as providing a more face
valid measure of driving and using a highly practised
skill. We have recently shown that performance impair-
ments on the York Driving Simulator following prolonged
wakefulness are equivalent to the decrements produced by
levels of alcohol intoxication that are illegal for operating
a motor vehicle [65,66]. These findings suggest that such
a driving task may have clinical utility in the assessment
of driving ability in patients who are at an increased risk
for sleepiness-related accidents.
Acknowledgments
The support of the Ontario Ministry of Transportation
(Grant No. 9505) and of Queens University is gratefully
acknowledged, as is the technical assistance of J. Brown, S.
Ferguson, and M. York.
References
[1] Dinges DF. An overview of sleepiness and accidents. J Sleep Res
1995;4:4 14.
[2] Horne JA, Reyner LA. Sleep related vehicle accidents. BMJ 1995;
310:565 7.
[3] Mitler MM, Carskadon MA, Czeisler CA, Dement WC, Dinges DF,
Graeber RC. Catastrophes, sleep, and public policy: consensus report.
Sleep 1988;11:100 9.
[4] Lyznicki JM, Doege TC, Davis RM, Williams MA. Sleepiness, driving,
and motor vehicle crashes Council on Scientific Affairs, American
Medical Association [see comments]. JAMA 1998;279:1908 13.
[5] Garbarino S, Nobili L, Beelke M, De Carli F, Ferrillo F. The
contributing role of sleepiness in highway vehicle accidents. Sleep
2001;24:203 6.
[6] Maycock G. Sleepiness and driving: the experience of UK car drivers.
J Sleep Res 1996;5:229 37.
[7] McCartt AT, Ribner SA, Pack AI, Hammer MC. The scope and nature
of the drowsy driving problem in New York state. Accid Anal Prev
1996;28:511 7.
[8] Pack AI, Pack AM, Rodgman E, Cucchiara A, Dinges DF, Schwab
CW. Characteristics of crashes attributed to the driver having fallen
asleep. Accid Anal Prev 1995;27:769 75.
[9] Carskadon MA, Dement WC, Mitler MM, Roth T, Westbrook PR,
Keenan S. Guidelines for the Multiple Sleep Latency Test (MSLT): a
standard measure of sleepiness. Sleep 1986;9:519 24.
[10] Dinges DF, Kribbs NB. Performing while sleepy: effects of exper-
imentally-induced sleepiness. In: Monk TH, editor. Sleep, sleepiness
and performance. Chicester7 Wiley, 1991. pp. 97 128.
[11] Wilkinson RT. The effect of lack of sleep on visual watch-keeping. Q J
Exp Psychol 1960;12:36 40.
[12] Wilkinson RT. Effects of up to 60 hours sleep deprivation on different
types of work. Ergonomics 1964;7:175 86.
[13] Wilkinson RT. Interaction of lack of sleep with knowledge of results,
repeated testing, and individual differences. J Exp Psychol 1961;62:
263 71.
[14] Williams HL, Lubin A, Goodenough JJ. Impaired performance with
acute sleep loss. Psychol Monogr 1959;73:1 26.
[15] Hockey GRJ. Changes in attention allocation in a multicomponent
task under loss of sleep. Br J Psychol 1970;61:473 80.
[16] Williams HL, Lubin A. Speeded addition and sleep loss. J Exp
Psychol 1967;73:313 7.
[17] Dinges DF, Powell JE. Microcomputer analyses of performance on a
portable, simple visual RT task during sustained operations. Behav
Res Methods Instrum Comput 1985;17:652 5.
[18] Wilkinson RT, Houghton D. Portable four-choice reaction time test
with magnetic memory tape. Behav Res Methods Instrum Comput
1975;7:441 6.
[19] Baddeley AD. A 3-min reasoning test based on grammatical
transformation. Psychon Sci 1968;10:341 2.
[20] Sternberg S. Memory scanning: mental processes revealed by reaction
time experiments. Am Sci 1969;57:421 57.
[21] Broadbent DE. Performance and its measurement. Br J Clin
Pharmacol 1984;18:5S 9S.
[22] Monk TH. Circadian rhythms in subjective activation, mood, and
performance efficiency. In: Kryger MH, Roth T, Dement WC,
editors. Principles and practice of sleep medicine, 2nd ed. Philadelphia
(PA)7 Saunders, 1994. pp. 321 30.
[23] Van Dongen HP, Dinges DF. Circadian rhythms in fatigue, alertness,
and performance. In: Kryger MH, Roth T, Dement WC, editors.
Principles and practice of sleep medicine, 3rd ed. Philadelphia (PA)7
Saunders, 2000. pp. 391 9.
[24] Fairclough SH, Graham R. Impairment of driving performance caused
by sleep deprivation or alcohol: a comparative study. Hum Factors
1999;41:118 28.
[25] Lenne MG, Triggs TJ, Redman JR. Interactive effects of sleep
deprivation, time of day, and driving experience on a driving task.
Sleep 1998;21:38 44.
[26] Gawron VJ, Ranney TA. The effects of alcohol dosing on driving
performance on a closed course and in a driving simulator. Ergonomics
1988;31:1219 44.
[27] Roehrs T, Beare D, Zorick F, Roth T. Sleepiness and ethanol effects on
simulated driving. Alcohol Clin Exp Res 1994;18:154 8.
 J.T. Arnedt et al. / Journal of Psychosomatic Research 58 (2005) 6171
[28] Findley L, Unverzagt M, Guchu R, Fabrizio M, Buckner J, Suratt PM.
Vigilance and automobile accidents in patients with sleep apnea or
narcolepsy. Chest 1995;108:619 24.
[29] George CFP, Boudreau AC, Smiley A. Comparison of simulated
driving performance in narcolepsy and sleep apnea patients. Sleep
1996;19:711 7.
[30] De Valck E, Cluydts R. Slow-release caffeine as a countermeasure to
driver sleepiness induced by partial sleep deprivation. J Sleep Res
2001;10:203 9.
[31] Arnedt JT, Acebo C, Seifer R, Carskadon MA. Assessment of a
simulated driving task for sleep research. Sleep 2001;24:A413.
[32] Johns MW. A new method for measuring daytime sleepiness: the
Epworth Sleepiness Scale. Sleep 1991;14:540 5.
[33] Horne JA, Ostberg O. A self-assessment questionnaire to determine
morningnesseveningness in human circadian rhythms. Int J Chrono-
biol 1976;4:97 110.
[34] Hoddes E, Dement W, Zarcone V. Quantification of sleepiness: a new
approach. Psychophysiology 1973;10:431 6.
[35] MacLean AW, Fekken CG, Saskin P, Knowles JB. Psychometric
evaluation of the Stanford Sleepiness Scale. J Sleep Res 1992;1:
35 9.
[36] Rechtschaffen A, Kales A. A manual of standardized terminology,
techniques, and scoring system for sleep stages of human subjects. Los
Angeles7 UCLA Brain Information Service/Brain Research Institute,
1968.
[37] Cairns J, Knowles JB, MacLean AW. Effect of varying the time of
sleep on sleep, vigilance, and self-rated activation. Psychophysiology
1982;19:623 8.
[38] Huyhn H, Feldt LS. Estimation of the box correction factor for degrees
of freedom from sample data in randomised block and split-plot
designs. J Educ Stat 1976;1:69 82.
[39] Cohen J. Statistical power for the behavioral sciences, 2nd ed. New
York7 Academic Press, 1988.
[40] Monroe LJ. Psychological and physiological differences between
good and poor sleepers. J Abnorm Psychol 1967;72:255 64.
[41] Carskadon MA, Dement WC. The Multiple Sleep Latency Test: what
does it measure? Sleep 1982;5:S6772.
[42] Froberg JE. Twenty-four-hour patterns in human performance, sub-
jective and physiological variables and differences between morning
and evening active subjects. Biol Psychol 1977;5:119 34.
[43] Dijk D-J, Duffy JF, Czeisler CA. Circadian and sleep/wake dependent
aspects of subjective alertness and cognitive performance. J Sleep Res
1992;1:112 7.
[44] 2kerstedt T. Shift work and disturbed sleep/wakefulness. Sleep Med
Rev 1998;2:117 28.
[45] Torsvall L, 2kerstedt T. Sleepiness on the job: continuously measured
EEG changes in train drivers. Electroencephalogr Clin Neurophysiol
1987;66:502 11.
[46] 2kerstedt T. Work hours, sleepiness and the underlying mechanisms.
J Sleep Res 1995;4:15 22.
71
[47] Monk TH. Subjective ratings of sleepinessthe underlying circadian
mechanisms. Sleep 1987;10:343 53.
[48] Horne JA, Reyner LA. Driver sleepiness. J Sleep Res 1995;4:23 9.
[49] Sugerman JL, Walsh JK. Physiological sleep tendency and ability to
maintain alertness at night. Sleep 1989;12:106 12.
[50] Walsh JK, Sugerman JL, Muehlbach MJ, Schweitzer PK. Physio-
logical sleep tendency on a simulated night shift: adaptation and
effects of triazolam. Sleep 1991;11:251 64.
[51] Deaton M, Tobias JS, Wilkinson RT. The effect of sleep deprivation on
signal detection parameters. Q J Exp Psychol 1971;23:449 52.
[52] Horne JA, Anderson NR, Wilkinson RT. Effects of sleep deprivation
on signal detection measures of vigilance: implications for sleep
function. Sleep 1983;6:347 58.
[53] Wilkinson RT. Sleep deprivation: performance tests for partial and
selective sleep deprivation. In: Abt LA, Riess BF, editors. Progress in
clinical psychology: dreams and dreaming, vol. 8. New York7 Grune
and Stratton, 1968. pp. 28 43.
[54] Naitoh P. Signal detection theory as applied to vigilance perform-
ance of sleep-deprived subjects. Sleep 1983;6:359 61.
[55] Lenne MG, Triggs TJ, Redman JR. Time of day variations in driving
performance. Accid Anal Prev 1997;29:431 7.
[56] Dawson D, Reid K. Fatigue, alcohol and performance impairment
[letter]. Nature 1997;388:235.
[57] Gillberg M, Kecklund G, Akerstedt T. Relations between performance
and subjective ratings of sleepiness during a night awake. Sleep 1994;
17:236 41.
[58] Summala H, Mikkola T. Fatal accidents among car and truck drivers:
effects of fatigue, age, and alcohol consumption. Hum Factors 1994;
36:315 26.
[59] Schwing RC. Exposure-controlled highway fatality rates: temporal
patterns compared to some explanatory variables. Alcohol Drugs Driv
1989/1990;5/6:275 85.
[60] DeWaard D, Brookhuis KA. Assessing driver status: a demonstration
experiment on the road. Accid Anal Prev 1991;23:297 307.
[61] Dureman EI, Boden C. Fatigue in simulated car driving. Ergonomics
1972;15:299 308.
[62] Stewart CP, MacLean AW. Effect of dynamic vehicles in simulated
driving during prolonged wakefulness. Sleep 1998;21:231.
[63] Thiffault P, Bergeron J. Monotony of road environment and driver
fatigue: a simulator study. Accid Anal Prev 2003;35:381 91.
[64] Dinges DF. The nature of sleepiness: causes, contexts, and consequences.
In: Stunkard AJ, Baum A, editors. Perspectives in behavioral medicine.
Hillsdale (NJ)7 Lawrence Erlbaum Associates, 1989. pp. 147 79.
[65] Arnedt JT, Wilde GJS, Munt PW, MacLean AW. Simulated driving
performance following prolonged wakefulness and alcohol consump-
tion: separate and combined contributions to impairment. J Sleep Res
2000;9:233 41.
[66] Arnedt JT, Wilde GJS, Munt PW, MacLean AW. How do prolonged
wakefulness and alcohol compare in the decrements they produce on a
simulated driving task? Accid Anal Prev 2001;33:337 44.