Accident Analysis and Prevention:Effects of major-road vehicle speed and driver age and gender on left-turn gap acceptance

a Center for Advanced Transportation Systems Simulation, Department of Civil & Environmental Engineering,
University of Central Florida, Orlando, FL 32816-2450, United States
b Computer Science Program, College of Business, Florida Gulf Coast University, 10501 FGCU Boulevard South,
Fort Myers, FL 33965-6565, United States
Received 28 August 2006; received in revised form 1 November 2006; accepted 12 December 2006
Abstract
Because the driver’s gap-acceptance maneuver is a complex and risky driving behavior, it is a highly concerned topic for traffic safety andoperation. Previous studies have mainly focused on the driver’s gap acceptance decision itself but did not pay attention to the maneuver processand driving behaviors. Using a driving simulator experiment for left-turn gap acceptance at a stop-controlled intersection, this study evaluatedthe effects of major traffic speed and driver age and gender on gap acceptance behaviors. The experiment results illustrate relationships amongdrivers’ left-turn gap decision, driver’s acceleration rate, steering action, and the influence of the gap-acceptance maneuver on the vehicles in themajor traffic stream. The experiment results identified an association between high crash risk 留學生dissertation網and high traffic speed at stop-controlled intersections.The older drivers, especially older female drivers, displayed a conservative driving attitude as a compensation for reduced driving ability, but also
showed to be the most vulnerable group for the relatively complex driving maneuvers.

Keywords: Driving simulator; Gap acceptance; Driving behavior; Stop-controlled intersection; Driver age difference; Driver gender difference

1. Introduction
Two-way stop-controlled (TWSC) intersections are the mostprevalent intersection type in the United States (Gattis and Low,1999). At stop-controlled intersections, drivers on the minor
road need to make use of proper gaps among the traffic tocross or merge into the major road. Because the driver’s gapacceptancemaneuver is a complex and risky driving behavior,
it is a highly concerned topic for traffic safety and operation.Retting et al. (2003) reported that nearly 700,000 motor vehiclecrashes occurred annually at stop signs in US and approximately
one-third of these crashes involved injuries in US. InMinnesota, there were 34,175 reported crashes on rural twolaneroads between 2000 and 2002. Over 32% (11,069) of these
crashes were intersection related, with 22% of the fatal ruralaccidents occurring at stop-controlled intersections (Labergeet al., 2006). Previous research has identified gap acceptanceproblems as a significant contributor to the stop-controlled intersectionaccidents and indicated that incorrect gap acceptancemight cause around 30% of left-turn accidents (Chovan et al.,1994). As an essential driving performance at stop-controlledintersections, the gap acceptance decision has been used as animportant measurement to analyze and predict traffic conflictsand accident rates at intersections (Spek et al., 2006; Alexanderet al., 2002).Furthermore, in the current AASHTO Manual (2001), thegap acceptance methodology is applied to determine intersection#p#分頁標題#e#
sight distances, which is based on previous researchworks (Harwood et al., 2000). In the Highway Capacity Manual(HCM, 2000), the critical gap accepted by drivers is akey parameter to calculate the minor traffic capacity. However,both important highway design manuals ignored theeffects of important factors such as driver age and gender andmajor-road vehicle speed on the gap acceptance, which hasgenerally been paid attention to by researchers in traffic safetyaspects.
1.1. Effect of traffic speed on gap acceptances
In the AASHTO and HCM manuals, it is assumed that thecritical gap accepted by drivers does not vary with major-roadvehicle speed, which is based on the previous analysis by Kyteet al. (1996). However, in the simulator experiment for turningleft from the major road into the minor road (Alexander etal., 2002), the velocity of the on-coming traffic was the variablethat had the greatest effect on the median accepted gapsize. This result corresponds to those of other previous studiesinvolving gap acceptance, which had shown that drivers accept asmaller gap at higher approach velocities (Darzentas et al., 1980;Staplin, 1995). Another field study indicated that for a giventime gap, the probability that a driver accepts the gap increasesas the speed of the opposing vehicle increases, which implies
that the distance is the primary determinant of gap acceptance(Davis and Swenson, 2004). Based on the speed dependencyof gap acceptance decision, Spek et al. (2006) developed a theoreticalcrash prediction model and found that the probabilitythat a crossing vehicle collides with the major stream vehiclecan be expected to increase when the major traffic speedincreases.
1.2. Age and gender differences in gap acceptances
It was reported that the process that driving performancebecomes progressively poorer with age accelerates somewherein the fifth decade of life (National Highway Traffic Safety
Administration, 1993). Older drivers have problems to adequatelydetect, perceive and accurately judge the safety of a gap(Laberge et al., 2006). Therefore, older drivers may experiencegreater difficulties at nonsignalized intersections as the result ofdiminished visual capabilities, such as depth and motion perception.Prior results indicated that judgments about whether
a potential collision would occur were less accurate for olderdrivers (40–64 years) compared with younger drivers (18–29years) (DeLucia et al., 2003). Lyles and Staplin (1991) examinedpolice-reported accidents in Michigan and Pennsylvaniaand found that when these were ordered according to olderdriverinvolvement rate, turning left across on-coming trafficand crossing or turning into a traffic stream were found to bethe most dangerous maneuvers. Moreover, it was indicated thatolder drivers are over-represented in severe injury crashes atintersections due in part to increases in frailty and functionaldisabilities that occur with age (Oxley et al., 2006).Lerner et al. (1995) concluded that older drivers are particularlyoverrepresented in multiple-vehicle, intersection-relatedaccidents, which could #p#分頁標題#e#留學生dissertation網be interpreted to reflect slow detectionof and reaction to other vehicles, or slow decision time forintersection-related maneuvers. Lerner et al. (1995) collectedgap acceptance data as a function of driver age for left turn,right turn, and through movements at stop-controlled intersections.
The findings indicated that younger drivers (20–40 years)accept shorter gaps than older drivers (over 65 years). Alexanderet al. (2002) found that females require a larger median gapthan males, irrespective of age, and older subjects (65–79 years)require a larger gap than younger subjects (under 60 years) of thesame sex. Through both simulator and field measures, Staplin(1995) indicated that older drivers show relative insensitivityto vehicle approach speed in left-turn maneuvers across themajor road traffic when compared with younger drivers. Thismay increase the risk of accidents if there is a lone speeder inthe traffic scheme. Abdel-Aty et al. (1999) pointed out that atintersections, elderly drivers are over-represented in right andleft turns as well as angle collisions. Scialfa et al. (1991) concludedthat older drivers generally over-estimate the speed ofvehicles traveling at low speeds, while under-estimating thespeed of those traveling much faster, which could explain theover-involvement of elderly drivers in accidents at junctions.
1.3. Objective of the study
Although numerous previous studies have been done forthe gap-acceptance research, all of them only focused on thegap acceptance decision itself but did not pay attention to
the maneuver process and related driving behavior. Using adriving simulator experiment for left-turn gap acceptance, thisstudy evaluated the effects of major traffic speed and driver
age and gender on gap acceptance behaviors, such as minimumaccepted gap, driver’s acceleration rate, steering action, andinfluence of the gap-acceptance maneuver on the vehicles inthe major traffic stream. This research aimed at analyzing theinteractive effect between gap-acceptance drivers and vehicleson the major road, and identifying the driving behavior differencesassociated with major traffic speed and driver age andgender.
2. Methodology
2.1. Participants
The experiment was a 3 (age)×2 (gender)×2 (major trafficspeed) within-subject repeated measures design. Sixty-threepaid participants in three age groups, ranging in age from 20 to 83years, were recruited for this research. Table 1 lists the descriptivestatistics of subjects by age and gender. The young groupranged from 20 to 30 years old, the middle-age group rangedfrom 31 to 55 years old, and the old group ranged from 56 to83 years old. Every participant held a valid Florida’s driver’slicense with at least 2 years of driving experience. The experi-
Table 1
Descriptive statistics of subjects by age and gender
Age Gender Mean Standard
deviation
N
Young (20–30) Female 24.00 2.42 14#p#分頁標題#e#
Male 23.57 2.85 14
Middle (31–55) Female 39.10 9.27 10
Male 33.64 3.29 11
Old (56–83) Female 69.75 8.66 4
Male 73.30 8.86 10
Total Female 35.93 16.93 28
Male 40.94 21.83 35X. Yan et al. / Accident Analysis and Prevention 39 (2007) 843–852 845
ment lasted for about 20 min in total and the participants were
compensated 20 dollars for their participation.
2.2. Apparatus/equipment
This study used the UCF (University of Central Florida) driving
simulator as a tool for data collection. The driving simulator
is an I-Sim Mark-II system with a motion base capable of operation
with 6 degrees of freedom. It includes five channels (one
forward, two side views and two rear view mirrors) of image
generation, an audio and vibration system, and steering wheel
feedback. The simulated environment is projected at 180? of
field view and at a resolution of 1280×1024 pixels. The driving
simulation system is composed of the following components:
• Simulator Cab: Saturn Sedan, automatic transmission, air
conditioning, the left back mirror and the back mirror inside
the cab.
• Simview: The software provides the graphical display based
on the computation.
• Motion base: It provides motion when driving.
• Scenario Editor: The software helps researchers to edit a
tested traffic scenario.
• APIs for reading real-time data: APIs (Application Programmer
Interface) can read the real-time data from Simview. The
sampling frequency is 60 Hz.
2.3. Description of traffic scenarios
In the gap acceptance experiments, a long straight undivided
two-lane highway was used as the major road. Its length is
around 3000m with a lane width of 3.66 m. A driving scenario
consists of three stages as shown in Fig. 1. During the course
of the experiment, the driver first drives the simulator for 400m
along the minor road (stage 1). When the simulator approaches
the intersection, all vehicles on the major road begin to enter
the scene. Then, according to the voice reminder and indication
sign (stop sign and left-turn sign), the driver stops at the
intersection and waits for the appropriate gap in the traffic on
the major road to make his/her left turn maneuver (stage 2).
When the simulated vehicle enters the major road and its speed
is caught up with the major road traffic speed (stage 3), the
experiment is completed. All subjects were tested in two scenarios,
40.2 km/h (25 mph) major road speed (scenario A) and
88.5 km/h (55 mph) major road speed (scenario B), which is
shown in Fig. 1.
2.4. Traffic design on the major road
A challenge in designing this experiment was how to make
the drivers perform their minimum left-turn gap-acceptance
maneuver in the same way accomplished in the real world. For#p#分頁標題#e#
obtaining that goal, the oncoming traffic on the major road from
the right was composed of two classes of intermingled gaps to
make the traffic appear random. One gap classification had very
small gaps (less than 3 s) that were unlikely to be accepted by
the participants. The other class consisted of increasing gaps
in which the subsequent gap was 1 s larger than the previous
one. This design assured that the selected gap would be close to
the driver’s minimum acceptable gap and the major road traffic
looked realistic. This concept is illustrated in Fig. 2. The
uniformly increasing gaps ranged in duration from 1 to 16 s,
a large enough variation to accommodate all drivers. A previous
gap acceptance study suggested that there is no meaningful
information regarding driver gap acceptance behavior when the
accepted gap size is above 12 s (Gattis and Low, 1999).
Fig. 1. Traffic scenario design for left-turn gap acceptance.
Fig. 2. Traffic design on the major road.846 X. Yan et al. / Accident Analysis and Prevention 39 (2007) 843–852
Whenever a car on the major road approached the intersection,
it was automatically changed from “record vehicle”
to the “normal vehicle”. The “record vehicles” move along a
fixed path with the unchangeable speed from the start point
to the end point, whereas the “normal vehicles” intelligently
move along the given route. While the normal vehicles follow
a designed speed, they can decelerate, accelerate, and pass
the slow impeding car according to the traffic situation. Therefore,
if the simulator is turning left into the major road very
slowly, the coming car on the major road will decrease its speed
to avoid a crash. Unless some participants drive the simulator
too aggressively or disobey the rationale, there will be no
crashes.
2.5. Experiment procedure
Upon arrival, the subjects were asked to fill out and sign
an informed consent form (per IRB). The subjects were then
advised to drive and behave as they normallywould and to adhere
to traffic laws as in real life situations. The subjects were also
notified that they could quit the experiment at any time in case
of motion sickness or any kind of discomfort. Prior to the formal
experiment, drivers were trained for at least 5 min to familiarize
with the driving simulator operation. During the course of
the practice, subjects exercised selected maneuvers including
straight driving, acceleration, deceleration, left/right turn, and
other basic driving behaviors.
After completing the familiarity course, the formal experiments
began during which all subjects faced the same set of two
driving scenarios, denoted A for the lower major traffic speed
(40.2 km/h) and B for the higher major traffic speed (88.5 km/h).#p#分頁標題#e#
The orders of scenarios were presented asA–BorB–Arandomly
for subject so as to eliminate the time order effect. For security
and liability reasons, each subject was escorted to the simulator
cabin to commence the experiment and he/she was allowed at
least 2 min to rest before running the next scenario.
2.6. Dependent measures
In the simulator experiment, the measured parameters related
to gap acceptance maneuver include minimum time gap
accepted by left turners (GAP), average steering angle velocity
during the left turn (SAV), average acceleration of the simulator
during the left-turn maneuver (ACC), minimum separation
between the minor-road vehicle and the following major-road
vehicle (MCD), speed reduction rate of the following vehicle on
the major road (SRR), and the deceleration rate of the following
vehicle (DEC). These dependent measures are explained as
follows:
• GAP (in s): the time headway between two vehicles on the
major road into which a left-turn driver may choose to turn.
• SAV (in rad/s): total rotation angles are divided by the total
time during which the simulator steer turn left and turn back
when subjects complete the left turn maneuver.
• ACC (in m/s2): average acceleration of simulator vehicle during
the period of making left turn maneuver.
• MCD (in m): minimum clearance distance between the simulator
and the following major-road vehicle before the left-turn
maneuver is completed.
• SRR: the speed reduction rate of the following vehicle on the
major road to accommodate the turning vehicle.
Table 2
Descriptive statistical results for dependent variables
Factors Parameter GAP (s) SAV (rad/s) ACC (m/s2) MCD (m) SRR DEC (m/s2)
Speed
40.2 km/h (N= 63) Mean 7.44 3.06 1.82 54.16 0.02 0.11
S.D. 2.08 0.79 0.61 17.75 0.05 0.21
88.5 km/h (N= 63) Mean 5.82 2.99 1.89 24.53 0.24 1.20
S.D. 1.46 0.82 0.68 13.38 0.13 1.25
Age
Young (N= 56) Mean 6.29 3.08 2.04 37.82 0.12 0.58
S.D. 1.79 0.69 0.60 18.68 0.14 0.81
Middle (N= 42) Mean 6.20 3.16 1.86 34.71 0.12 0.85
S.D. 1.50 0.82 0.61 18.58 0.14 1.51
Old (N= 28) Mean 7.94 2.71 1.47 49.35 0.16 0.52
S.D. 2.38 0.92 0.63 27.96 0.17 0.47
Gender
Female (N= 56) Mean 6.93 3.03 1.81 40.88 0.14 0.65
S.D. 2.02 0.80 0.66 22.90 0.17 0.86
Male (N= 70) Mean 6.38 3.02 1.89 38.12 0.12 0.66
S.D. 1.90 0.81 0.64 20.58 0.12 1.18
Total
(N= 126) Mean 6.63 3.03 1.85 39.34 0.13 0.65
S.D. 1.97 0.80 0.65 21.60 0.14 1.05X. Yan et al. / Accident Analysis and Prevention 39 (2007) 843–852 847
• DEC (in m/s2): the deceleration rate used by the following
vehicle on the major road to accommodate the turning vehicle.
3. Experiment results
The basic statistical descriptions of the experiment results#p#分頁標題#e#
are listed in Table 2. In the subsequent statistical analyses, an
ANOVAwas used to investigate differences between factors (see
Table 3). The hypothesis testing in the following analysis was
based on a 0.05 significance level.
3.1. Correlation analyses for dependent variables
Table 4 shows the correlation analysis among the dependent
variables. Clearly, theGAPis correlated with all the other parameters.
Intuitively, if drivers accept smaller gaps, they tend to
make left-turn maneuvers with a faster steering velocity and
higher acceleration rate, and result in shorter minimum clearance
distances to the following vehicles, and larger percentage
of speed reduction and higher deceleration rates of the following
vehicles.
Furthermore, it was found that MCD, SRR, and DEC are
highly correlated with each other. When the left-turn vehicle
enters onto the major road at a slow speed, the following car
on the major road decreases its speed to avoid a crash. As the
following car on the major road is approaching the turning-into
vehicle whose speed keeps increasing, the distance between the
turning-into vehicle and the following major-road vehicle gets
smaller, until the minimum clearance distance between the two
vehicles occurs. At that same time, the following car’s speed
Table 5
Mean values of GAP for speed×gender×age
Category Young Middle Old
40.2 km/h
Female 7.56 6.97 10.99
Male 6.35 6.60 8.76
88.5 km/h
Female 6.00 5.63 7.11
Male 5.26 5.61 6.23
is lowest and its speed reduction is greatest. An example for
this analysis is illustrated in Fig. 3. When time unit is equal to
114.45, the clearance distance is minimum and equal to 19.53 m,
and the speed of the following car is 63.46 km/h, which is close
to that (63.14 km/h) of the simulator vehicle. It can be concluded
that the minimum clearance distance and the maximum speed
reduction always happen at the same experiment sampling time.
After minimum clearance distance occurs, the left turn maneuver
can be considered to be successfully completed by the subject.
The smaller MCD and the larger SRR and DEC indicate that the
left-turn vehicles have more effects on the major road traffic.
3.2. Minimum gap acceptance
Table 5 lists mean values of GAP for each speed and for
each combination of gender×age. TheANOVAresults showthe
significant effects of age (p < 0.001), gender (p = 0.004), speed
(p < 0.001), and two-way interaction between age and speed
(p = 0.035) on driver’s gap acceptance (see Table 3).
Table 3
Analysis of variance table for dependent measures
Source d.f. F-ratio
GAP SAV ACC MCD SRR DEC
Gender 1 8.514** .112 2.601 3.660 6.690* .024
Age 2 14.798** 3.587* 9.587** 9.749** 4.447* 1.325#p#分頁標題#e#
Speed 1 37.044** .752 .159 111.678** 155.082** 36.781**
Gender×age 2 1.468 2.943 1.557 .337 4.053* .138
Gender×speed 1 1.314 1.129 1.003 .009 2.012 .019
Age×speed 2 3.462* .388 .018 1.824 .812 .545
Gender×age×speed 2 .204 .029 .171 .063 1.194 .302
Mean square error 114 2.560 .621 .382 219.173 .009 .837
** Significant at the 0.01 level.
* Significant at the 0.05 level.
Table 4
Correlation among dependent variables
GAP SAV ACC MCD SRR DEC
GAP 1 −.280** −.302** .591** −.218* −.298**
SAV −.280** 1 .425** −.074 −.224* −.081
ACC −.302** .425** 1 .324** −.455** −.234**
MCD .591** −.074 .324** 1 −.557** −.448**
SRR −.218* −.224* −.455** −.557** 1 .593**
DEC −.298** −.081 −.234** −.448** .593** 1
** Correlation is significant at the 0.01 level (2-tailed).
* Correlation is significant at the 0.05 level (2-tailed).848 X. Yan et al. / Accident Analysis and Prevention 39 (2007) 843–852
Fig. 3. Relationship between clearance distance and speeds of the left-turning
vehicle (simulator) and the following car.
Old drivers tend to accept larger gaps (M = 7.94 s,
S.D. = 2.38 s), compared to younger drivers (M = 6.29 s,
S.D. = 1.79 s) and middle-age drivers (M = 6.20 s, S.D. = 1.50 s).
There is no significant difference between young drivers and
middle-age drivers [t(96) = 0.283, p = 0.778]. For the gender factor,
it appears that male drivers accept smaller gaps (M = 6.38 s,
S.D. = 1.90 s) than female drivers(M = 6.93 s, S.D. = 2.02 s). The
findings suggest that older drivers and female drivers are more
conservative than the other groups. Although the interaction
effect between age and gender was not found to be significant
by ANOVA, a plot of the minimum gap versus driver’s
actual age shows different trends of gap acceptance between
males and females (see Fig. 4). The comparison of the 2ndorder
polynomial regression lines shows that the minimum gaps
Fig. 4. Minimum GAP vs. driver’s actual age by gender.
accepted by male drivers tend to gradually increase as the driver
age increases while those for female drivers show a U-shape
pattern, which indicates that middle age groups accept smaller
gaps and the younger and older groups have relatively larger
gaps. Furthermore, it shows that the gender difference in GAP
is not significant for drivers from 35 to 55 years; however, for
drivers younger than 35 years or older than 55 years, females
appear to accept larger gaps than males.
The results show that the oncoming vehicle speed on the
major road plays an important role in the gap size accepted
by drivers. Drivers would accept smaller gaps (M = 5.82 s,#p#分頁標題#e#
S.D. = 1.46 s) for the higher major-road traffic speed scenario
compared to those (M = 7.44 s, S.D. = 2.08 s) for the lower
major-road traffic speed scenario. It implies that drivers rely
on both space and time information to perform gap-acceptance
judgments, which is corresponding to the conclusion of prior
studies (Alexander et al., 2002; Davis and Swenson, 2004).
Furthermore, the interaction effect between speed and driver
age shows that for the lower major-road traffic speed scenario,
there is a strong gap acceptance difference in driver age [F(2,
60) = 34.756, p < 0.001], and the old drivers apparently accept
larger gaps (M = 9.40 s, S.D. = 2.24 s) compared to the younger
drivers (M = 6.96 s, S.D. = 1.82 s) and the middle-age drivers
(M = 6.77 s, S.D. = 1.49 s); however, for the higher major-road
traffic speed scenario, the minimum gaps accepted by old drivers
(M = 6.48 s, S.D. = 1.48 s) are not significantly larger than the
younger drivers (M = 5.63 s, S.D. = 1.51 s) and the middle-age
drivers (M = 5.62 s, S.D. = 1.31 s) [F(2, 60) = 1.926, p = 0.155].
The interaction effect between speed and driver age is illustrated
in Fig. 5.
3.3. Steering angle velocity during left turning
In the output file of driving simulator experiment, steering
control data display the angle of the selected simulator’s steering
wheel. According to theANOVAresults, only age (p = 0.031)
has a significant effect on SAV. During the process of turning
left, old drivers (M = 2.71 rad/s, S.D. = 0.92 rad/s) tend to turn
the simulator steering wheel more slowly than younger drivers
(M = 3.08 rad/s, S.D. = 0.69 rad/s) [t(82) = 2.116, p = 0.037]X. Yan et al. / Accident Analysis and Prevention 39 (2007) 843–852 847
• DEC (in m/s2): the deceleration rate used by the following
vehicle on the major road to accommodate the turning vehicle.
3. Experiment results
The basic statistical descriptions of the experiment results
are listed in Table 2. In the subsequent statistical analyses, an
ANOVAwas used to investigate differences between factors (see
Table 3). The hypothesis testing in the following analysis was
based on a 0.05 significance level.
3.1. Correlation analyses for dependent variables
Table 4 shows the correlation analysis among the dependent
variables. Clearly, theGAPis correlated with all the other parameters.
Intuitively, if drivers accept smaller gaps, they tend to
make left-turn maneuvers with a faster steering velocity and
higher acceleration rate, and result in shorter minimum clearance
distances to the following vehicles, and larger percentage
of speed reduction and higher deceleration rates of the following
vehicles.
Furthermore, it was found that MCD, SRR, and DEC are
highly correlated with each other. When the left-turn vehicle#p#分頁標題#e#
enters onto the major road at a slow speed, the following car
on the major road decreases its speed to avoid a crash. As the
following car on the major road is approaching the turning-into
vehicle whose speed keeps increasing, the distance between the
turning-into vehicle and the following major-road vehicle gets
smaller, until the minimum clearance distance between the two
vehicles occurs. At that same time, the following car’s speed
Table 5
Mean values of GAP for speed×gender×age
Category Young Middle Old
40.2 km/h
Female 7.56 6.97 10.99
Male 6.35 6.60 8.76
88.5 km/h
Female 6.00 5.63 7.11
Male 5.26 5.61 6.23
is lowest and its speed reduction is greatest. An example for
this analysis is illustrated in Fig. 3. When time unit is equal to
114.45, the clearance distance is minimum and equal to 19.53 m,
and the speed of the following car is 63.46 km/h, which is close
to that (63.14 km/h) of the simulator vehicle. It can be concluded
that the minimum clearance distance and the maximum speed
reduction always happen at the same experiment sampling time.
After minimum clearance distance occurs, the left turn maneuver
can be considered to be successfully completed by the subject.
The smaller MCD and the larger SRR and DEC indicate that the
left-turn vehicles have more effects on the major road traffic.
3.2. Minimum gap acceptance
Table 5 lists mean values of GAP for each speed and for
each combination of gender×age. TheANOVAresults showthe
significant effects of age (p < 0.001), gender (p = 0.004), speed
(p < 0.001), and two-way interaction between age and speed
(p = 0.035) on driver’s gap acceptance (see Table 3).
Table 3
Analysis of variance table for dependent measures
Source d.f. F-ratio
GAP SAV ACC MCD SRR DEC
Gender 1 8.514** .112 2.601 3.660 6.690* .024
Age 2 14.798** 3.587* 9.587** 9.749** 4.447* 1.325
Speed 1 37.044** .752 .159 111.678** 155.082** 36.781**
Gender×age 2 1.468 2.943 1.557 .337 4.053* .138
Gender×speed 1 1.314 1.129 1.003 .009 2.012 .019
Age×speed 2 3.462* .388 .018 1.824 .812 .545
Gender×age×speed 2 .204 .029 .171 .063 1.194 .302
Mean square error 114 2.560 .621 .382 219.173 .009 .837
** Significant at the 0.01 level.
* Significant at the 0.05 level.
Table 4
Correlation among dependent variables
GAP SAV ACC MCD SRR DEC
GAP 1 −.280** −.302** .591** −.218* −.298**
SAV −.280** 1 .425** −.074 −.224* −.081
ACC −.302** .425** 1 .324** −.455** −.234**
MCD .591** −.074 .324** 1 −.557** −.448**
SRR −.218* −.224* −.455** −.557** 1 .593**
DEC −.298** −.081 −.234** −.448** .593** 1#p#分頁標題#e#
** Correlation is significant at the 0.01 level (2-tailed).
* Correlation is significant at the 0.05 level (2-tailed).848 X. Yan et al. / Accident Analysis and Prevention 39 (2007) 843–852
Fig. 3. Relationship between clearance distance and speeds of the left-turning
vehicle (simulator) and the following car.
Old drivers tend to accept larger gaps (M = 7.94 s,
S.D. = 2.38 s), compared to younger drivers (M = 6.29 s,
S.D. = 1.79 s) and middle-age drivers (M = 6.20 s, S.D. = 1.50 s).
There is no significant difference between young drivers and
middle-age drivers [t(96) = 0.283, p = 0.778]. For the gender factor,
it appears that male drivers accept smaller gaps (M = 6.38 s,
S.D. = 1.90 s) than female drivers(M = 6.93 s, S.D. = 2.02 s). The
findings suggest that older drivers and female drivers are more
conservative than the other groups. Although the interaction
effect between age and gender was not found to be significant
by ANOVA, a plot of the minimum gap versus driver’s
actual age shows different trends of gap acceptance between
males and females (see Fig. 4). The comparison of the 2ndorder
polynomial regression lines shows that the minimum gaps
Fig. 4. Minimum GAP vs. driver’s actual age by gender.
accepted by male drivers tend to gradually increase as the driver
age increases while those for female drivers show a U-shape
pattern, which indicates that middle age groups accept smaller
gaps and the younger and older groups have relatively larger
gaps. Furthermore, it shows that the gender difference in GAP
is not significant for drivers from 35 to 55 years; however, for
drivers younger than 35 years or older than 55 years, females
appear to accept larger gaps than males.
The results show that the oncoming vehicle speed on the
major road plays an important role in the gap size accepted
by drivers. Drivers would accept smaller gaps (M = 5.82 s,
S.D. = 1.46 s) for the higher major-road traffic speed scenario
compared to those (M = 7.44 s, S.D. = 2.08 s) for the lower
major-road traffic speed scenario. It implies that drivers rely
on both space and time information to perform gap-acceptance
judgments, which is corresponding to the conclusion of prior
studies (Alexander et al., 2002; Davis and Swenson, 2004).
Furthermore, the interaction effect between speed and driver
age shows that for the lower major-road traffic speed scenario,
there is a strong gap acceptance difference in driver age [F(2,
60) = 34.756, p < 0.001], and the old drivers apparently accept
larger gaps (M = 9.40 s, S.D. = 2.24 s) compared to the younger
drivers (M = 6.96 s, S.D. = 1.82 s) and the middle-age drivers
(M = 6.77 s, S.D. = 1.49 s); however, for the higher major-road
traffic speed scenario, the minimum gaps accepted by old drivers#p#分頁標題#e#
(M = 6.48 s, S.D. = 1.48 s) are not significantly larger than the
younger drivers (M = 5.63 s, S.D. = 1.51 s) and the middle-age
drivers (M = 5.62 s, S.D. = 1.31 s) [F(2, 60) = 1.926, p = 0.155].
The interaction effect between speed and driver age is illustrated
in Fig. 5.
3.3. Steering angle velocity during left turning
In the output file of driving simulator experiment, steering
control data display the angle of the selected simulator’s steering
wheel. According to theANOVAresults, only age (p = 0.031)
has a significant effect on SAV. During the process of turning
left, old drivers (M = 2.71 rad/s, S.D. = 0.92 rad/s) tend to turn
the simulator steering wheel more slowly than younger drivers
(M = 3.08 rad/s, S.D. = 0.69 rad/s) [t(82) = 2.116, p = 0.037]X. Yan et al. / Accident Analysis and Prevention 39 (2007) 843–852 849
Fig. 5. Interaction effect between age and speed on gap acceptance.
and middle-age drivers (M = 3.16 rad/s, S.D. = 0.82 rad/s)
[t(68) = 2.152, p = 0.035]. Between young and middle-age
groups, there is no statistical difference in SAV [t(96) = 0.488,
p = 0.627]. This finding indicates the trend of decreasing driving
ability for older drivers. Experiment results show that the mean
of SAV is 3.06 and 2.99 rad/s for scenarios A and B, respectively,
and there is no statistical difference in the effect of the
major road design speed [F(1, 124) = 0.299, p = 0.585].
3.4. Left-turning acceleration
The ANOVA result shows that only driver age (p < 0.001)
has significant effect on driver’s acceleration rate during the
period of the left turn at the intersection. Older drivers are
more likely to use smaller acceleration rates (M = 1.47 m/s2,
S.D. = 0.63 m/s2) to achieve the major road traffic speed
compared to the younger drivers (M = 2.04, S.D. = 0.60) and
middle-age drivers (M = 1.86 m/s2, S.D. = 0.61 m/s2). Between
young and middle-age groups, there is no statistical difference
in ACC [t(96) = 1.473, p = 0.144]. Generally, the larger gaps that
drivers accepted, the smaller accelerations that drivers used to
turn into the major road. Therefore, the driver age effect on
acceleration rate is corresponding to the trend that older drivers
accept larger gaps, and reflect older drivers’ conservative driving
attitude.
3.5. Minimum clearance distance and speed reduction rate
and deceleration rate of the following vehicle on the major
road
Based on the ANOVA, it was found that both age (p < 0.001)
and speed (p < 0.001) are significantly associated with the
minimum clear distance (MCD) between the left-turning simulator
and the following vehicle. Older drivers have larger
Fig. 6. Speed effect on minimum clearance distance.
MCD (M = 49.35 m, S.D. = 27.96 m), compared to those for#p#分頁標題#e#
young drivers (M = 37.82 m, S.D. = 18.68 m) and middle-age
drivers (M = 34.71 m, S.D. = 18.58 m). Another interesting finding
shows that drivers have much smaller MCD for the higher
major-road traffic speed (88.5 km/h) compared to that for the
lower major-road speed (40.2 km/h), as shown in Fig. 6.
Furthermore, theANOVAresults showthat speed (p < 0.001),
age (p = 0.014), gender (p = 0.011), and two-way interaction
between age and gender (p = 0.020) have significant effects on
the speed reductions of the following vehicles on the majorroad
(SRR) (see Table 3). When drivers turn into the higher
speed major-road traffic, they cause the following vehicle to
reduce speed by a higher rate (M = 0.24, S.D. = 0.13), compared
to turning into the lower speed traffic (M = 0.02, S.D. = 0.05).
As aforementioned before, the speed reduction rate of the
major-road vehicle is highly correlated to its deceleration rate.
The ANOVA shows that the deceleration rate (M = 1.20 m/s2,
S.D. = 1.25 m/s2) of the following vehicle in the higher majorroad
traffic speed scenario is also much higher than that
(M = 0.11 m/s2, S.D. = 0.21 m/s2) in the lower major-road traffic
speed scenario [F(1, 124) = 45.863, p < 0.001]. The results
imply a general trend that left-turn vehicles have more effects
on higher speed major-road traffic and drivers take more risks
for gap acceptance maneuvers at intersections with higher design
speeds.
It was also found that the left turns of older drivers contribute
to more SRR (M = 0.16, S.D. = 0.17), compared to
those for young drivers (M = 0.12, S.D. = 0.14) and middle-age
drivers (M = 0.12, S.D. = 0.14); and female drivers result in a
higher SRR (M = 0.14, S.D. = 0.17) than male drivers (M = 0.12,
S.D. = 0.12). Furthermore, the interaction effect between age and
gender illustrates that the old female drivers cause the highest
SRR (see Fig. 7) but other age-gender groups showsimilar SRR.
Therefore, older female drivers have a main contribution to gen850 X. Yan et al. / Accident Analysis and Prevention 39 (2007) 843–852
Fig. 7. Interaction effect between age and gender on speed reduction rate.
der and age difference in the following vehicle speed reduction
rate.
4. Discussions
The experiment results illustrated that the major road
traffic speed contributes to an interactive effect between gapacceptance
drivers and vehicles on the major road. First, drivers
are more likely to accept smaller gaps in the higher major-road
traffic speed scenario than the lower major-road traffic speed
scenario. It implied that drivers show relatively less sensitivity
to vehicle approach speed but more sensitivity to on-coming
vehicle’s distance or position. The perceptual task of drivers for
gap acceptance requires integrating speed and distance information#p#分頁標題#e#
of a longitudinally coming-through vehicle moving in depth
without significant change in visual direction. Olson (1996) suggests
that the speed of an approaching vehicle may not easily
be perceived until it is very near. This finding corresponds to
the previous study results of gap acceptances for left-crossing
maneuver (Cooper et al., 1977; Hancock et al., 1991; Alexander
et al., 2002).
On the other hand, due to accepting smaller gaps for the
higher traffic speed, drivers have shorter separation from the following
vehicle, result in more speed reduction and deceleration
rate of the following vehicles to accommodate the turning vehicle,
and therefore contribute to more effects on major road traffic
operation. These findings implied a connection between high
crash risk and high traffic speed at stop-controlled intersections.
A previous study reported that stop-controlled intersections
where high-speed, high-volume roads are intersected by lowerspeed,
lower-volume roads controlled by a stop sign are a major
problem for traffic safety, particularly in rural areas (Laberge et
al., 2006). Therefore, effective speed control countermeasures to
reduce speeding behaviors may lead to decrease in crashes happening
at stop-controlled intersections. Also, it is suggested that
a further analysis of the quantitative relationship between intersection
operation speed and gap-acceptance crash rate based on
crash databases be conducted.
The experiment results showed that there is an apparent age
difference in the left-turn driving behavior, as well as gender
difference. Compared to younger drivers and middle-age
drivers, the older subjects tend to select larger gaps to make
left turns, turn the steering wheel more slowly, and keep a further
clearance distance from the following car. The conservative
driving attitude of older drivers would compensate driving ability
decrease. In the previous research, there is also evidence that
elderly drivers perceive potential risk situations at intersections
and change their driving habits to avoid them. It was demonstrated
that drivers 55 years and older, particularly those who
had a previous history of crashes, avoided driving in rain, during
rush hour and making left turns across traffic (Ball et al., 1998).
However, when making left turns in the higher major-road traffic
speed scenario, the older subjects did not show age difference in
gap acceptance. The results confirmed the conclusion that older
drivers rely primarily or exclusively on perceived distance to
perform gap-acceptance judgments, reflecting a reduced ability
to integrate time based on speed and distance information with
increasing age (Staplin et al., 1993). Spek et al. (2006) indicated
that the effect that drivers tend to accept smaller time gaps as#p#分頁標題#e#
the approach speed increases appears to be stronger for older
drivers than for younger drivers, suggesting that the older driver
is more prone to collide with speeding vehicles.
For the gender factor, the results showed that male drivers
accept smaller gaps than female drivers, which reflect a general
conservative driving attitude of female drivers. By assessing
both general perceptual-motor performance and safety concerns,
it was found that male drivers consistently overestimate their
perceptual-motor skills, whereas safety skills are more prominent
among female drivers (Lajunen et al., 1998; Lajunen and
Summala, 1995). However, the result also showed that the older
female drivers contribute to the most significant influence in the
major road vehicle speed. The finding implied that older females
could be the most vulnerable group for the relatively complex
driving maneuvers because of more reductions in their driving
ability. It is worthwhile to mention that the older female group
in this study only includes four observations, which led to larger
variations of the related experiment results. Nevertheless, the
observed trend in this study implies that a further safety concern
is needed for this high risk driver group.
Driving simulators provide a safe and controlled virtual environment
to test high risk driving behaviors. The convenient
traffic design in the driving simulator allowed researchers to
observe the minimum gap selected by subjects, while it is
very difficult in a field study since drivers’ gap acceptances
are influenced by the traffic volume distribution on the major
road. According to the experience of this simulator experiment,
the scenario design method can be used to analyze other gapacceptance
behaviors at priority intersections, such as right turn
from the minor road and crossing maneuver from the minor road
or major road. Furthermore, the driving simulator enabled the
traffic parameters, such as SAV, ACC, MCD, SRR, and DECX. Yan et al. / Accident Analysis and Prevention 39 (2007) 843–852 851
related to drivers’ gap decisions, to be efficiently collected.
These parameters were generally omitted by previous gap acceptance
studies. In this study, it was found that older drivers tended
to accept larger gaps to compensate the reducing driving ability,
but they still showed more influence on the major road traffic
than younger and middle-age drivers. Thus, the results implied
that the gap acceptance accident risk may not be accurately predicted
only based on the probability that drivers accept small
gaps.
However, it should be mentioned that for simulator experiments,
there are some general limitations and validation
considerations.Aproblem that occurs frequently with all driving
simulators is simulator sickness. Based on the authors’ experiences#p#分頁標題#e#
in previous driving simulator experiments, the simulator
sickness is particularly correlated with turning maneuvers and
braking behaviors. Without a real safety risk as field test, sickness
in a simulator experiment may significantly harm drivers’
gap acceptance decisions because drivers would try to complete
the experiment as soon as possible to reduce the discomfort level.
Therefore, this study attempted to avoid the sickness effect on the
experiment results through encouraging participants with sickness
to quit the experiment (they still obtained payment for their
involvements even if not completing the experiment). Another
purpose of this policy is to protect older participants since the
oldest subject is 83 years old. In this study, the older female subjects
were most sensitive to simulator sickness and only 4 of 10
participants fully completed the experiments without suffering
any sickness.
Previous driving simulator validity studies related driving
behaviors are measured using two types of validity: absolute
validity (when the numerical values between the two systems
are the same) and relative validity (when differences found
between experimental conditions are in the same direction, and
have a similar or identical magnitude on both systems) (Blaauw,
1982). For the absolute validity, the conclusions of the simulator
validity varied in different types of driving simulators
and test scenarios. However, from the perspective of the relative
validity, most previous studies concluded that driving simulators
can reflect a similar trend of the driving performance in driving
simulators to that in the real world. Lee (2002) compared driving
performance of older adult drivers (60–90 years) between a
PC-based simulator and subjects’ own cars. He found a covariance
(r2 = 0.66) between the two measures and concluded that
simulator usage was a safer and more economical method than
the on-road testing to assess the driving performance of older
adult drivers. For the speed and lane position control studies,
Godley et al. (2002) and T¨ornros’s (1998) reported that participants
generally drove faster in the instrumented car than
the simulator, whereas relative validity was achieved for both
speed and lateral position. More recently, a speed validation
study focused on the effectiveness of temporary traffic signs
on highways (Bella, 2005). The results showed that differences
between the speeds observed in the real situation and those measured
with the simulator were not statistically significant and
validated the driving simulator in absolute terms. In this study,
since the authors did not conduct a field validation for the simulator
data, caution should be taken in transferring these findings
directly to practice in absolute terms. However, the observed#p#分頁標題#e#
trends and patterns associated with the gap-acceptance behaviors
should be valid and contribute to the further research or
practical work.
5. Conclusions
In summary, this driving simulator experiment illustrated
relationships among drivers’ left-turn gap decision, corresponding
driving behaviors, and consequent influence on the vehicles
in the major traffic stream. The results indicated that major
road traffic speed and driver age and gender have significant
effects on the gap-acceptance maneuver. The findings identified
an association between high crash risk and high traffic speed
at stop-controlled intersections. The older drivers displayed a
http://www.400190.com/dissertation_writing/conservative driving attitude as a compensation for the reducing
driving ability. However, due to the aging effect on cognitive,
perceptual, and motor abilities, they may still be the most vulnerable
group for the relatively complex driving maneuvers,
especially for older female drivers.
Acknowledgments
The authors would like to acknowledge the Florida Department
of Transportation (FDOT) for its sponsorship of this
research project. Appreciation is also extended to CATSS staff
and students for providing assistance with the execution of the
simulation of the experiment. The recommendations of this
study are those of the authors and they do not represent views
of FDOT.
References
American Association of State Highway and Transportation Officials
(AASHTO), 2001. A Policy on Geometric Design of Highways and Streets
Washington, DC.
Abdel-Aty, M., Chen, C., Radwan, E., 1999. Using conditional probability to
find the driver age effect in crashes, J. Transp. Eng., 1999, ASCE, vol. 125,
No. 6.
Alexander, J., Barham, P., Black, I., 2002. Factors influencing the probability
of an incident at a junction: results from an interactive driving simulator.
Accid. Anal. Prev. 34, 779–792.
Ball, K., Owsley, C., Stalvey, B., Roenker, D.L., Sloane, M.E., Graves, M., 1998.
Driving avoidance and functional impairment in older drivers. Accid. Anal.
Prev. 30, 313–322.
Bella, F., 2005. Validation of a driving simulator for work zone design. Transp.
Res. Rec. 1937, 136–144.
Blaauw, G.J., 1982. Driving experience and task demands in simulator and
instrumented car: a validation study. Hum. Factors 24, 473–486, 2004.
Chovan, J.D., Tijerian, L., Everson, J.H., Pierowicz, J.A., Hendricks, D.L.,
1994. Examination of intersection left turn across path crashes and IVHS
countermeasures (DOT HS 808 154). National Highway Traffic Safety
Administration.
Cooper, D.F., Storr, P.A., Wennell, J., 1977. Traffic studies at T-junctions: the
effect of speed on gap acceptance and conflict rate. Traffic Eng. Control 18,#p#分頁標題#e#
110–112.
http://www.400190.com/dissertation_writing/Darzentas, M.R., McDowell, C., Cooper, D., 1980. Minimum acceptable gaps
and conflict involvement in a simple crossing maneuver. Traffic Eng. Control
21 (2), 58–61.
Davis, G.A., Swenson, T., 2004. Field study of gap acceptance by left-turning
drivers. Transp. Res. Rec. 1899, 71–75.