Content uploaded by Yovela Murzello
Author content
All content in this area was uploaded by Yovela Murzello on Oct 30, 2022
Content may be subject to copyright.
Research Article
Transportation Research Record
1–25
ÓNational Academy of Sciences:
Transportation Research Board 2022
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/03611981221096666
journals.sagepub.com/home/trr
Hazard Perception in Driving: A
Systematic Literature Review
Shi Cao
1
,SibySamuel
1
, Yovela Murzello
1
,
Wen Ding
1
, Xuemei Zhang
2
, and Jianwei Niu
2
Abstract
Hazard perception (HP) is the process of detecting and identifying hazards. Drivers’ HP abilities are critical for driving safety.
This paper presents a systematic literature review of driver HP, including scientific measures of HP, major human factors
affecting HP, and training methods for improving HP skills. Sixty-nine peer-reviewed studies were identified and reviewed.
The results showed that common measures of HP include hazard scenario questionnaires, HP reaction time, hazard hit rate,
and eye fixation measures such as fixation probability, fixation reaction time, fixation duration, and fixation variance. Major
human factors that affect HP include experience, aging, fatigue, distraction, and the use of alcohol and drugs. Various training
methods have been developed to train and improve drivers’ HP skills. In general, there is evidence in the literature showing
the effectiveness of HP training. A combination of complementary training approaches such as instruction, expert demonstra-
tion, and active practice with feedback and attention support the use of picture-, video-, computer-, and simulator-based
training methods to improve HP performance in shorter HP reaction time, higher hazard hit rate, and better eye scan pat-
terns (more spread scan, more anticipatory scan). These findings could guide future work developing and designing HP train-
ing programs. Three future research areas are identified and discussed: the need for standardized HP tests, long-term testing
of HP training programs, and new HP questions and challenges brought by partially automated vehicles.
Keywords
pedestrians, bicycles, human factors, human factors of vehicles, driver behavior, hazard perception
The cognitive process of driving can be described as three
stages: perception of driving-related information, making
decisions, and taking actions to control the vehicle.
Hazard perception (HP) is within the perception stage. In
the surface transportation literature, a hazard generally
refers to ‘‘any object, situation, occurrence or combina-
tion of these that introduces the possibility of the individ-
ual road user experiencing harm’’ (1). It covers many
different types of hazards from different sources, such as
hazards from the driver (e.g., fatigue, alcohol, and dis-
tracted attention), hazards from the traffic (e.g., a leading
car that suddenly brakes, a pedestrian crossing the road,
and obstacles on the roadway), hazards from the natural
environment (e.g., fog, rain, and snow), and hazards
from the driver’s vehicle (e.g., engine malfunction, tire
explosion, and brake malfunction). Depending on the
threat imminence level, hazards can be grouped into
immediate hazards (or materialized hazards) and non-
immediate hazards (or unmaterialized, latent hazards).
Immediate hazards can be defined as hazards that require
a driver ‘‘to take immediate action (e.g., braking or swer-
ving) to avoid a dangerous interaction with another road
user,’’ whereas non-immediate hazards can be defined as
‘‘hazards that did not require immediate evasive action
but required attention in case they developed into imme-
diate hazards’’ (2). Depending on their visibility, non-
immediate hazards can be further grouped into potential
or overt non-immediate hazards (visible) and hidden or
covert non-immediate hazards (not visible) (3).
HP is the process of identifying hazards—also used to
refer to the skills or capabilities to identify hazards (4).
1
Department of Systems Design Engineering, University of Waterloo,
Waterloo, Canada
2
School of Mechanical Engineering, University of Science and Technology
Beijing, Beijing, China
Corresponding Author:
Jianwei Niu, niujw@ustb.edu.cn
Studies have shown that poorer HP skills are associated
with higher rates of accident involvement (5–7). It has
been shown that including HP tests in the driver licensing
process has benefits in reducing crash rate and enhancing
traffic safety (8). HP is related to situation awareness
(SA) (7). SA is about the degree to which the available
information in working memory meets the needs to suc-
cessfully perform the task (9). HP can be considered as
the SA of hazards (3). Corresponding to Endsley’s (10)
model, SA of hazards in driving is posited to comprise
three components: the ability to constantly perceive
hazards on the roadway, the comprehension of the sever-
ity of potential adverse events caused by the hazard, and
the ability to project the status of the hazard into the
near future.
Over the past few decades, many studies have exam-
ined factors that affect HP. They have found multiple
human factors such as driver’s knowledge and experience
that can affect HP. Studying and understanding HP are
important for promoting driving safety. Although novice
drivers have learned some basic skills for driving a vehi-
cle and passed the licensing exams, they could still lack
proper skill proficiency and skills for HP, which may
contribute to their overrepresentation in accidents (11).
Despite the great amount of literature in this field, there
is a lack of a comprehensive review that addresses multi-
ple core aspects pertaining to HP. The aim of this paper
is to provide an organized review as a reference for
researchers, by summarizing various approaches of HP
measurement and HP training under different hazard
scenarios. The objectives include (i) documenting the
methods used to measure HP, (ii) summarizing various
human factors that affect HP, (iii) highlighting different
training methods developed to improve HP, and (iv) dis-
cussing potential future research directions.
Methodology and Scope of Review
This paper presents a systematic review of literature on
HP in driving. Reviewed studies included both labora-
tory simulation and real-world driving studies. Articles
were included in the review if they met the following cri-
teria: (a) demonstrated clear alignment to the purpose of
the review in the four objectives specified above; (b) pre-
sented clear experimental evidence; (c) retrievable online;
and (d) written in English. Articles were excluded from
the review if they: (a) qualified as non-peer-reviewed lit-
erature and (b) included small samples.
A comprehensive search of various databases was
conducted to identify peer-reviewed, English language
publications from indexed electronic and digital sources.
The various electronic journals and databases that
were scoped include PubMed, TRID (Transportation
Research International Documentation), CiteSeer,
Scopus, Ref-Works, Web of Science, Mendeley, and
Google Scholar. Additional articles were searched, veri-
fied, and scoured using a snowballing approach of scop-
ing identified articles for additional references.
The keywords for the search process were driven by a
preliminary environmental scan. Specific keywords used
in the search process included ‘‘hazard perception’’;
‘‘hazard anticipation’’; ‘‘latent hazard anticipation’’;
‘‘hazard detection’’; ‘‘hazard perception training’’; ‘‘latent
hazard anticipation training’’; ‘‘tactical hazard percep-
tion’’; ‘‘strategic hazard anticipation’’; ‘‘measurement
hazard perception’’; ‘‘situation awareness’’; ‘‘SAGAT’’;
‘‘eye tracking in hazard perception’’; ‘‘age related differ-
ences in hazard perception.’’ These key phrases were
searched in the ‘‘title’’ and ‘‘abstract’’ fields of the data-
bases identified above.
Following quality assessment (two researchers combed
through the sources to ensure adherence to inclusion/
exclusion criteria), the systematic search yielded 69 peer-
reviewed studies that met all criteria and satisfied the
scope of the review. Fifty-six studies were laboratory
based (38 studies used videos or images, 15 studies used
driving simulators, and three studies used questionnaires
or focus groups), while the remaining five studies were
either conducted on the open road or on a closed-loop
track. Two studies were laboratory based as well as con-
ducted on-road, and six studies were meta-analyses or
reviews. Among the 69 studies, 39 focused on measure-
ment of HP, 28 studied the human factors affecting HP,
and 25 studies evaluated training-based methodologies to
improve HP. The findings from these studies are reported
and discussed in the subsequent sections.
Measuring Hazard Perception
Researchers studying HP usually focus on traffic
hazards, such as sudden emerging pedestrians or motor-
cycles. Anticipation and quick identification of traffic
hazards are not easy and require knowledge and experi-
ence (12). HP is often measured using tests showing
driver-perspective video of traffic situations. Some
videos were real footage recordings (e.g., Sagberg and
Bjørnskau [13]), whereas others were computer generated
video clips (e.g., Vlakveld [14] and Malone and Bru
¨nken
[15]). Test-takers are often asked to identify traffic
hazards presented in the video. In the instructions to
test-takers, hazards are explained as traffic conflicts—
situations in which the driver needs to take actions such
as braking, steering, or sounding the horn to avoid
potentially dangerous interaction with another road user
(e.g., Isler et al. [16] and Smith et al. [17]). Using a video-
based HP test, for example, a study (6) showed that
2Transportation Research Record 00(0)
drivers who failed the test were more likely to be
involved in crash during a one year period following the
test.
Driving simulators have also been used to test HP.
Simulators allow researchers to observe drivers’ steering,
accelerator, and brake actions, as well as other opera-
tions while simulated hazards are presented. For exam-
ple, Chan et al. (18) measured eye tracking results from
both newly licensed and experienced drivers in a driving
simulator. The results showed that novice drivers glanced
for longer periods of time inside the vehicle, and the eye
fixation patterns were similar to the authors’ knowledge
of results in real-world driving studies. Takahashi et al.
(19) measured older drivers’ responses to hazards in a
driving simulator and found the impact of decreased cog-
nitive function on HP and response.
Both video-based and simulator-based HP tests
require dynamic driving scenes. Alternatively, picture-
based tests that use static images as testing materials
have also been examined (20,21). Static test materials
are easier to make, but there was only a weak correlation
between static and dynamic test results (22), and there
has been no evidence to support that static tests can
replace dynamic tests. Nevertheless, all these tests are
safer than on-road testing because they can reliably
recreate traffic hazards and avoid exposing drivers and
their assessors to danger during on-road testing (23).
To assess HP, researchers have used both question-
naires and behavioral index measures. In some HP tests
using multiple-choice questionnaires (15,24), partici-
pants were given pre-recorded traffic scenarios to review
and then asked if they were aware of a potential hazard
within the observed scenario. If their answer was affir-
mative, they were presented with a multiple-choice ques-
tion where each alternative answer contained a short
description of a hazardous situation that could have
occurred within the scenario. Another scheme is to use
SA questionnaire techniques for measuring HP. The
Situation Awareness Global Assessment Technique
(SAGAT) (9) can be used, which involves freezing the
video clip before the hazard occurs at a pre-determined
time unknown to the participant and asking questions to
measure awareness about current and near-future situa-
tions. For example, researchers have used the SAGAT
method with questions such as ‘‘What was the source of
the hazard? What was the location of the hazard? What
happens next?’’ (3Ws). Participants’ responses were then
numerically scored by awarding two points for a correct
response, one point for a partially correct response, and
zero points for an incorrect response (25). Similar ques-
tions and scoring methods have been used in several
studies (e.g., [26–30]) as well as in the Multiple-Choice
Hazard Perception and Prediction test (31–33). In one
study, researchers (34) used questions to measure both
detection and cautiousness such as ‘‘Had you seen any
hazard at the moment when the video was cut?’’ (detec-
tion) and ‘‘What maneuver would you perform if you
were the driver of the vehicle?’’ (cautiousness). It is
important to note that only the detection question is
about HP. The cautiousness question is about decision
and responses that are beyond the scope of HP.
While questionnaires provide a direct way to measure
whether a driver has perceived the hazard, these methods
often interfere with the natural flow of driving. In addi-
tion, questionnaire methods could not examine how fast
drivers can recognize hazards. In this regard, behavioral
index measures such as reaction time can be used to
quantify HP without interference with the driving task.
Previous studies have used behavioral indexes, including
reaction time, hit rate, and eye tracking measures for
HP.
The reaction time of HP, also called HP time duration
or HP latency, is measured from the moment a conflict
becomes recognizable to the driver to the moment that it
is reported by a button press or a mouse click. For imme-
diate hazards, the timing usually begins when the hazard
is visible to the driver. For non-immediate hazards, the
timing may start at the moment when test designers
believe most drivers should anticipate potential risk from
an object or a big object blocking the view to potential
hidden hazards. Since non-immediate hazards pose no
immediate threat and do not require immediate evasive
action, HP reaction time is usually not an important
measure for non-immediate hazards (instead, HP hit rate
could be used). The validity of each response should be
checked to avoid participants guessing or cheating. Some
researchers have considered using time windows to
exclude very early responses (25), but there has been no
consensus on how to optimally calibrate these response
windows. Although it is usually assumed that HP time
duration should not be shorter than simple reaction time
duration (typically around 200–300 ms), it is still possible
for some drivers who utilize early cues to respond very
fast (23). Instead of time windows, a better validation
approach is to check the intention of responses by
recording participants’ verbal report for each hazard
identification and check if the identified object is a valid
target (e.g., Borowsky et al. [4] and Shahar et al. [35]).
The accuracy of HP is often measured as hit rate,
which is calculated as the ratio of the total number of
correctly identified hazards in a test to the total number
of hazards appearing in the test as defined by test
designers. From a signal detection theory perspective,
correct rejection rate might also be useful, but it is usu-
ally very difficult to quantify because it requires knowing
the total number of non-hazards appearing in the test
scenario. Since essentially all the objects other than
hazards are non-hazards, it is difficult to quantity the
Cao et al 3
total number of non-hazards because there are too
many. In some studies, only the total number of cor-
rectly identified hazards was reported; however, that pre-
vents comparison across studies because different studies
may use different numbers of hazards. The present
authors recommend comprehensive reporting of HP hit
rate as well as the total number of correctly identified
hazards and the total number of hazards appeared in the
test.
Speed–accuracy trade-off is a frequently observed
human behavioral characteristic. It means that people
can choose to focus on attaining either faster reaction
speed or higher accuracy, but it is difficult to achieve
both at the same time (20). Since reaction time and hit
rate are two separate measures, some researchers have
proposed ways to combine them into a single measure.
One of the methods creates a time duration value for
missed trials by assigning it the maximum reaction time
value observed from the participants (35–37), so the over-
all reaction time of HP will cover both hit and miss trials.
Another method is to formulate a scoring scheme that
reflects both reaction time and hit rate. For example,
faster responses in a trial yield higher scores (up to five
points), and missed or wrong responses yield a score of
zero (38). Other researchers have also suggested calculat-
ing the z-scores of reaction time and hit rate separately
and then calculating the average of z-scores for each par-
ticipant as the combined measure (39). However, there
has been no consensus on the optimal method for com-
bining reaction time and hit rate measures. The authors’
recommendation is to report both to provide a compre-
hensive representation of HP performance.
As to eye tracking measures for HP, previous studies
have used fixation probability, fixation reaction time,
fixation duration, and fixation variance. Eye fixation
means maintaining the visual gaze on a single location.
To determine whether a hazard is fixated, a region of
interest is defined surrounding the hazard, and then
researchers calculate whether a fixation falls within this
region of interest. For example, in one study using a
video-based HP test (40), the region of interest for pedes-
trian hazards was defined as a rectangle surrounding the
pedestrian with an additional width and height of 1°
visual angle, roughly representing the area of foveal
vision. Fixation measures rely on the assumptions that
fixation means hazard perceived, and no fixation means
hazard not perceived. These assumptions are often true
but not always. On the one hand, people may fixate on
familiar objects but fail to perceive them because of inat-
tentional blindness (41). On the other hand, peripheral
vision can also be used to perceive simple objects such as
light, simple symbols, and moving cars (42–44).
Therefore, fixation measures for HP should be inter-
preted with caution, and the authors recommend the use
of verbal confirmation, that is, participants should verb-
ally report the name of the hazard they see.
In the literature, fixation probability, or glance prob-
ability, has been used to measure the percentage of
hazards perceived by drivers (12,40,45,46). A successful
fixation, that is, a fixation that falls in the predefined
area of interest, is denoted by ‘‘1,’’ while ‘‘0’’ represents
failure. Fixation probability is the success rate of fixation
over all hazards. Some researchers have emphasized the
importance of anticipation in HP and argued that a
hazard perceived too late should not be considered as a
success. Therefore, they proposed the use of launch
zones based on the position of the driver’s vehicle. A
launch zone is defined for each hazard and represents
the area of the roadway where the driver should begin
glancing at the hazard to be able to successfully antici-
pate and mitigate the threat (47). When the launch zone
method is used, fixation on a hazard is only counted as a
success if the driver’s vehicle is within the launch zone
defined for the hazard. In general, a higher fixation
probability represents better HP. But there has been no
established standard about how to set the launch zone
for each hazard.
Fixation reaction time, measured as the duration from
the onset of a hazard to the moment when participants
first fixate on the hazard, has also been used (12).
Theoretically, the fixation reaction time should be
shorter than the reaction time of HP measured by man-
ual responses because of the extra time needed for man-
ual processes. The measure of fixation reaction time has
an advantage over manual response measure of HP reac-
tion time because fixation reaction does not require any
key press and therefore avoids adding interference with
natural driving.
Some studies have also reported fixation duration,
which refers to the duration of each fixation on hazards
(12,40). Typically the average of fixation duration is
around 150 to 300 ms (48). The relationship between
fixation duration and HP performance is not straightfor-
ward (49). In one study (12), experienced drivers pro-
duced longer average fixation duration than learners; in
contrast, another study found that experienced drivers
produced shorter average fixation duration than learners
(40). While longer fixation duration generally means a
higher amount of attention devoted to perceiving the
hazard, the effectiveness of the fixation also depends on
drivers’ knowledge and skills. Drivers need to divide
visual attention resources (represented by fixation dura-
tion) properly across multiple visual targets. Ideally, each
fixation duration should be long enough to allow the
extraction of important information concerning a poten-
tial hazard, but not too long to hinder the processing of
other visual targets. It is generally regarded that fixation
duration within 100 to 500ms is appropriate to process
4Transportation Research Record 00(0)
the information (48), but the proper duration is affected
by many factors such as complexity of the road environ-
ment and the lighting conditions (50,51). Since the opti-
mal fixation duration has not been determined, more
studies are needed, and researchers are encouraged to
report fixation duration to accumulate data for future
meta-analysis.
Another measure related to drivers’ visual scan pat-
tern is fixation variance (in degrees of visual angle
squared), which refers to the variance of fixation loca-
tions (in degree) along the vertical and horizontal median
(52). Studies have shown that experienced drivers had
greater spread of visual search represented by larger fixa-
tion variance (52,53). Since it is generally preferable to
have visual attention spread across the visual field to
scan more hazards, larger fixation variance is usually
regarded as an indicator for better HP.
While the above eye fixation measures are clearly
applicable to immediate hazards, their application to
non-immediate hazards requires more careful thought
and definition. In the case of potential or overt non-
immediate hazards that are visible but have not materia-
lized into a course of collision with the driver’s vehicle, it
is important to define the critical time window when the
driver needs to scan these visual targets to anticipate and
become aware of these non-immediate hazards, and then
valid eye fixations within this time window can be
regarded as indication of good HP. In the case of hidden
or covert non-immediate hazards that are not visible,
such as pedestrians hidden behind a bus or oncoming
traffic hidden behind a truck, it is important to define
both the critical time window when the driver needs to
scan them and the area where such hidden hazards may
appear. Eye fixations in this area within this time win-
dow are indication of good HP. This does not mean that
all the big objects such as trucks and bushes should be
defined as areas of hidden hazards that need to be
scanned. Only the big objects in scenarios with clear cues
of hidden hazards should be considered; for example, the
cues could be school crossing signs, bus stop with cross-
walk, intersection, and left turn waiting area. Eye fixa-
tions on non-immediate hazards are sometimes referred
to as anticipatory glances (54,55). Since there is no
imminent threat from non-immediate hazards, reaction
time measures are usually not considered for non-
immediate hazards.
In summary, HP can be measured using both dynamic
tests (video-based and simulator-based) and static tests
(picture-based). Although picture- and video-based tests
are easier to conduct, simulator-based tests are better at
capturing multiple sources of mental demands similar to
real driving scenarios. In driving simulators, HP can be
examined under limited attention resources while some
attention resources are used to drive the vehicle as in on-
road driving. Video-based tests are generally better than
picture-based tests, because HP involves anticipating
potential hazards using precursors of hazards (12), which
are typically lacking in static tests. Table 1 provides a
summary of all the behavioral measures of HP discussed
above with notes and recommended practice.
Human Factors Affecting Hazard
Perception
From the literature, experience, aging, fatigue, distrac-
tion, and the use of alcohol and drugs were identified as
major factors that affect HP.
Experience
Researchers generally believe that HP skills improve with
the accumulation of driving experience (56). This is evi-
dent in studies that examined the validity of HP tests and
driving simulators by comparing performance between
novice and experienced drivers (20,23,40,49). Many
studies have found that experienced drivers have better
HP than less experienced drivers, with faster HP reaction
time (13,15,20,21,57), faster fixation reaction time
(12), higher hazard hit rate (14,15,27), and higher fixa-
tion probability (12,46). The visual scan pattern of expe-
rienced drivers showed greater fixation variance (wider
spread of fixation locations horizontally and vertically),
indicating that they are better at spreading attention to a
wider range of visual targets (52,53). In general, previous
studies suggested that the knowledge and skills developed
as drivers gain more experience include the visual search
strategy to allocate attention across multiple objects,
knowledge about the characteristics of hazards, and
knowledge about the link between precursors and
hazards, all of which are the reasons for better HP per-
formance. These results underly the foundation for HP
training.
Factors Affecting Cognitive Process
Since the process of HP relies on visual search, attention
resources, and pattern recognition, factors affecting cog-
nitive process will also affect HP performance. In the lit-
erature, a group of factors belong to this category,
including aging, fatigue, distraction, alcohol, and drugs.
Aging. Horswill et al. (58) measured HP reaction time
from older drivers (118 participants aged 65–84 years,
each having at least 10 years of driving experience). The
HP test used edited video recordings from real-world
traffic scenarios that contained immediate hazards. The
results showed that HP reaction time correlated with age,
with older drivers having longer HP reaction time. The
Cao et al 5
drivers’ simple reaction time, visual contrast sensitivity,
and visual useful field of view were also measured and
included in a regression model to predict HP reaction
time. The result showed that ‘‘contrast sensitivity, useful
field of view, and simple reaction time could account for
the variance in hazard perception, independent of one
another and of individual differences in age’’ (58), which
means that within this older age group, the negative
impact of aging on HP could be caused by slow down
and degradation in the cognitive and vision processes.
Fatigue. Examining HP reaction time in a video-based
test, researchers (17) found that sleepiness significantly
increased HP reaction time for novice drivers (aged 17–
24 years), whereas HP reaction time from experienced
drivers (aged 28–36) was not significantly affected by
sleepiness. This result suggests that experienced drivers
may have more robust skills for HP that are more resili-
ent to fatigue. With age and experience, it is difficult to
control and separate their effects because older drivers
typically have more experience than younger drivers
(59). With older adults (65years and above), the effect of
aging could be stronger than the effect of experience;
with younger adult groups, such as the groups in this
fatigue study (no more than 36years old), the effect of
experience is expected to be stronger than the effect of
aging.
Distraction. Distraction during driving may present signif-
icant risk to drivers. Distraction could be visual (e.g.,
Table 1. Summary of Behavioral Measures of Hazard Perception (HP)
Measure name Definition Notes and recommendation References
Response reaction time The duration of time from the
moment a conflict becomes
recognizable to the moment it
is reported, usually by pressing
a button
It is recommended for drivers to
verbally report the hazard they
see, to avoid guessing and
incorrect objects being identified.
Reaction time should be reported
together with hit rate.
Jackson et al. (25)
Wetton et al. (23)
Borowsky et al. (4)
Shahar et al. (35)
Hit rate The ratio of the total number of
correctly identified hazards in
a test to the total number of
hazards appearing in the test
as defined by test designers
It is recommended to report hit
rate as well as the total number of
correctly identified hazards and
the total number of hazards
appearing in the test.
Shahar et al. (35)
McKenna et al. (36)
Markkula et al. (37)
Hoffman and Rosenbloom (38)
Eye tracking measures
Fixation probability The ratio of the number of
hazards that have been fixated
to the total number of hazards
It is recommended to ask
participants to verbally report the
name of the hazard they see as a
confirmation.
Borowsky et al. (40)
Crundall et al. (12)
Hajiseyedjavadi et al. (45)
Pradhan et al. (46)
Samuel and Fisher (47)
Fixation reaction time The duration of time from the
onset of a hazard to the
moment when participant first
fixates on the hazard
It has an advantage over manual
response measure of HP reaction
time because it does not require
any key press and therefore
avoids adding interference with
natural driving.
Crundall et al. (12)
Fixation duration The duration of time of each
fixation on hazards
Since the optimal fixation duration
has not been determined, more
studies are needed, and
researchers are encouraged to
report fixation duration to
accumulate data for future meta-
analysis.
Borowsky et al. (40)
Crundall et al. (12)
Fixation variance The variance of fixation
locations along the vertical and
horizontal median.
Since it is generally preferable to
have visual attention spread
across the visual field to scan
more hazards, larger fixation
variance is usually regarded as an
indicator for better HP.
Crundall and Underwood (52)
Mourant and Rockwell (53)
6Transportation Research Record 00(0)
reading emails), auditory (e.g., listening to radio news),
mental (e.g., thoughts not related to driving), speech
(e.g., speaking over the phone), manual (e.g., pressing
button on an interface), or the combination of the above.
With regard to visual and manual distraction such as
texting, studies using driving simulators have found that
texting while driving significantly increased HP reaction
time (60,61). For mental distraction, a study using
video-based HP tests found that concurrent mental tasks
(solving puzzles) also significantly increased HP reaction
time (62). However, results for auditory and speech dis-
traction were mixed. While some studies found that con-
versation during driving reduced reaction time (63–66),
other studies found that conversation while driving
resulted in increased HP reaction time and decreased hit
rate (67,68). In summary, most non-driving in-vehicle
tasks are expected to impair HP with the exception of
verbal conversation in some driving situations.
Driving Under Influence. Driving under the influence (DUI)
of alcohol or drugs is a common problem (69–71). A
review study (72) showed strong evidence that blood
alcohol concentration (BAC) 0.05% or higher can signif-
icantly impair driving performance and increase crash
rate. A recent review (73) showed that studies using can-
nabis dose around 10 to 20 mg D9-tetrahydrocannabinol
(THC) all found significantly negative impact on driving
performance. The combination of low BAC (0.04%) and
low THC (around 70 ng/ml pre-drive) has been found to
have an additive effect to produce an additive decrement
on driving performance (74). While previous studies mea-
sured driving performance such as lateral control and
speed control, fewer studies have particularly examined
the effect of DUI on HP as an aspect of driving perfor-
mance. A study found that HP reaction time was
significantly increased by 0.3 s (in 0.025% BAC condi-
tion) and 0.7 s (in 0.05% BAC condition) by alcohol.
However, one study found that very light consumption
of alcohol (0.015% BAC) increased HP hit rate in com-
parison with 0% BAC (75). For cannabis, a study found
no significant impact of cannabis dose around 10 to
20 mg THC on HP in comparison with a control group
with 0.035 mg THC (76).
In summary, HP skills develop with driving experi-
ence. Research suggests that important knowledge and
skills supporting HP include the visual search strategy to
allocate attention across multiple locations and targets,
knowledge about the characteristics of hazards, and
knowledge about the link between precursors and
hazards. Previous studies have examined many human
factors affecting HP, including experience, aging, fatigue,
distraction, and the use of alcohol and drugs. These find-
ings have provided the foundation for the design of HP
training programs (77,78).
Hazard Perception Training
In most countries, driver training is provided by experi-
enced driving instructors. Novice drivers’ knowledge and
skills are usually developed through reading materials,
viewing instructors’ demonstration, and supervised prac-
tice with feedback from instructors (79). Traditional
driver training often focuses on vehicle control skills.
This alone is insufficient to support the development of
HP skills (11,80), and learners in traditional driver train-
ing usually cannot experience the full range of hazardous
situations. To address this issue, specialized HP training
programs have been developed (e.g., Horswill et al. [78]).
HP training involves various training materials and is
based on several training methodologies and techniques
Figure 1. Example of a computer-based hazard perception (HP) training program by Samuel et al. (90). The program provides simulated
drives where users can interact by clicking on hazards (e.g., a car pulling out, shown by red solid arrow).
Cao et al 7
that are summarized as follows. From a skill training
methodology point of view, HP skills can be acquired
through instructions and active practice. Instructions
provide trainees with declarative knowledge such as
safety facts and what they should and should not do,
and then repeated practice allows trainees to apply the
declarative knowledge in performing the tasks and gra-
dually form procedural knowledge through skill acquisi-
tion (81).
Instructional training methods generally include
instructions, demonstration, and expert commentary.
For example, Meir et al. (82) used written instructions
with pictures of hazardous situations to conceptually
teach the types of hazards and where they are most likely
to appear. Ivancic and Hesketh (83) used video clips that
demonstrated HP errors such as failure to identify a hid-
den car and failure to identify a red light that resulted in
collision or police tickets. McKenna et al. (36) and
Wetton et al. (84) used video clips of various traffic
situations recorded from the driver’s perspective with
expert commentary training that explained potential
hazards and where to look for them to maximize the
chance of identifying them. In another unique study,
Castro et al. (28) examined the effect of proactive listen-
ing to a training commentary. In this training, partici-
pants listened to a speech commentary with information
about how to allocate their attention in complex driving
scenes, and the real consequences of the scenes were
revealed at the same time. It resulted in improved HP
performance for the participants on a post-intervention
What Happens Next (WHN) Assessment. In the work of
Isler et al. (16), young drivers received video-based road
commentary training, and their HP abilities were tested
afterwards. Trained young drivers detected and identi-
fied significantly more hazards (Mean = 77.2, Standard
Deviation = 6.5) than the age and driving experience
matched control group (M = 62.5, SD = 11.6); the two
groups had the same HP performance before the training.
Different types of attention support can be used in
instructional training and active practice to reduce parti-
cipants’ cognitive load on driving activities. Practice
without any guide or support, equivalent to self-learning
without training, is often used as a baseline condition to
be compared with different training methods. According
to the cognitive load theory (85), novice drivers’ mental
resources are mostly occupied by basic driving tasks such
steering and speed control, and they have limited mental
resources for learning HP. Therefore, HP training pro-
grams should be designed to reduce trainees’ cognitive
load from activities unrelated to HP training and help
trainees focus on self-reflection and learning, for exam-
ple, by guiding drivers’ attention (86). As trainees
advance in the training process, guidance and support of
the trainee’s attention can be gradually reduced and
eventually removed to allow the adaptation to actual
driving.
Video-based and picture-based HP tests can help
reduce cognitive load during HP training, because the
vehicle control components of driving are not included
in these tests. For example, researchers used a video-
based HP test that included 63 video clips with different
kinds of hazards and traffic environments for training pur-
poses (82). The video playing speed may be slowed, for
example, at half speed to further reduce task difficulty and
cognitive load (87). Some researchers have also used meth-
ods that freeze video clips at critical moments to ensure
enough time for deep thinking and reflection (84,87).
Trainees could be probed using questionnaires, similar to
SAGAT (9), to strengthen their thoughts on what has just
happened and what will happen next. Continuous video
clips may be simplified into a sequence of pictures (screen-
shots). For example, Pradhan et al. (88)usedsuch3s
screenshots to present driving scenes from the driver’s per-
spective looking straight ahead, while left or right views
were also provided at intersections. Trainees were asked to
click on the areas to which they would pay attention for
potential hazards. However, screenshots make it difficult
for trainees to judge the driving speed. Petzoldt et al. (89)
developed a computer-based training (CBT) module that
complemented existing driver training programs. In the
study, video sequences of potentially hazardous driving
situations were used followed by multiple-choice questions
with adaptive feedback for improving the understanding
of the scenarios. The results of a simulator test confirmed
that participants using this CBT had quicker glances
toward critical cues and relevant areas in their visual field
than participants who received paper-based training with
similar content or no training at all. This result demon-
strated the potential of this CBT module in supporting the
development of HP skills. Figure 1 shows an example of a
computer-based HP program.
Some training techniques were found useful in train-
ing drivers to enhance HP in driving. Repeated exposure
to specific hazards has been shown to improve HP. In a
study by Kahana-Levy et al. (91), both inexperienced
and experienced drivers participated in a training session
that involved viewing repetitions of HP video clips each
representing one type of hazard, embedded among filler
videos. The eye movement data and performance in the
subsequent transference sessions where video clips with
novel hazards are used indicated that the trained groups
had an earlier fixation on hazards than untrained groups.
The results suggested that repetitive training was effective
for both inexperienced and experienced drivers.
Another way to support trainees’ skill learning is to
direct their attention toward hazards. During normal
driving, drivers need to divide their attention across mul-
tiple task components such as steering control, speed
8Transportation Research Record 00(0)
control, road monitoring, and trip planning. Drivers may
also engage in thoughts not related to driving, that is,
mind wandering (92). One technique that has been used
to help trainees focus on HP is commentary training,
where participants are asked to continuously verbalize
their thoughts about hazards that they detect. Participant
commentary training was often applied while trainees
were watching driver-perspective video clips (16,87). In
one study (80), trainees were asked to continuously ver-
balize their HP process as well as what they would do to
avoid risky situations while they drove on the road with
an instructor. While the commentary is expected to help
trainees focus their attention on perceiving hazards, it
creates additional workload to them as drivers and could
be detrimental to driving performance and safety (93).
Feedback is also important to learning and self-reflec-
tion. Horswill et al. (94) showed that adding feedback on
drivers’ performance in a video-based HP training task
resulted in an improvement in HP time performance in
comparison with the control group with no feedback. In
addition, Chapman et al. (87) used visual cues to guide
trainees’ visual attention to critical areas where hazards
were likely to appear. In the video clips, areas of interest
were circled in blue or red color. These attention support
methods could help trainees in the initial stages of
learning.
Through instructional training, trainees can quickly
learn a wide range of hazardous situations that would
otherwise take a long time to acquire from on-road driv-
ing practice. However, instructional training does not
provide such a deep level of involvement as active cogni-
tive processing does in practice training, and active prac-
tice is necessary for the acquisition of cognitive skills
(95). Therefore, the two approaches should be combined
to achieve maximal effects. Wetton et al. (84) reported a
training package combining video-based ‘‘what happens
next’’ exercises and self-generated commentary training.
Participants who received the full training package had
a significantly larger reduction in HP response time
(M = 24.32 s, Standard Error = 0.43) compared with
those in the control group (M= 20.14 s, SE = 0.41).
Chapman et al. (87) reported a training package that
included participant commentary training and practice
with task simplification and support (slowing down,
freezing, and visual attention guide). The materials were
video clips of potentially dangerous driving situations.
Training for about 50 min produced notable changes in
the participants’ visual search patterns, and some of the
changes were still detectable after a few months (87).
To incorporate active practice, feedback, and atten-
tion guidance into instructional training, some training
programs have used an error-based training approach.
For example, Risk Awareness and Perception Training
(RAPT) is a CBT program that focuses on anticipation
of hazards and combines active practice and instructions
(88,96). The practice is based on HP tests using screen-
shots. The hazards focused on hidden road users and
abrupt movement of other road users, for example, a
bicyclist or a pedestrian hidden behind a hedge, cars
obscured by bushes or other vegetation, cars hidden
behind hills, and cars that abruptly change lanes.
Trainees were required to click on the areas to which
they would pay attention for potential hazards. If they
failed to click on the correct areas, plan views (top-down
views showing positions of the driver’s car and other
objects) of hazardous situations and accompanying writ-
ten instructions were shown to explain potential hazards
and where to look for them. It does not mean all big
objects such as trees and bushes should be defined as
areas of hidden hazards that need to be scanned. Only
the big objects in scenarios with clear cues of hidden
hazards should be considered; for example, the cues
could be school crossing signs, bus stop with crosswalk,
intersection, and left turn waiting area.
Trainees who received training for about 45 min had
significantly higher hazard hit rate (64.4%) than the
untrained control group (37.4%) (88), and this increased
hit rate after training had not diminished after about
eight months (97). Subsequent versions of RAPT admi-
nistered on web and tablet were found to be effective (in
increased hazard hit rate measured in HP tests) immedi-
ately after training and for up to sevenmonths after
training (90,98–100). However, these studies did not col-
lect data on any potential effect on crash rate.
Additionally, the effect of the RAPT program on newly
licensed young drivers has been investigated (101). After
12 months of tracking post-licensure, analyses of pre-test
and post-test data indicated performance improvements
in trainees who completed the RAPT program.
However, the increase in the number of correct responses
after the training did not necessarily indicate an increase
in risk perception or hazard recognition knowledge.
Gender differences were also observed with a decrease in
the crash rate for males, but not females. Nevertheless,
the findings suggest that RAPT can have a positive influ-
ence on the driving safety of young novice drivers.
In comparison with CBT programs, simulator-based
training can further utilize driving simulation to train
and practice HP skills in dynamic driving situations. In
CBT, the displayed environment and drives are fixed and
do not change as a function of the participant’s inputs.
In simulator-based training, participants take control as
the driver of the simulated car, and they could experience
in real time the consequences of failing to detect a
hazard. Research by Vlakveld et al. (55) revealed that
simulator-based training can enhance novice drivers’
visual search for non-immediate hazards. The partici-
pants underwent Simulator-based Risk Awareness and
Cao et al 9
Perception Training (SimRAPT) with elicited crashes
and near-crashes in a driver training simulator and were
then assessed on an advanced driving simulator. In the
post-training evaluation, both near transfer (where the
hazards were similar to the trained situations but in a dif-
ferent environment) and far transfer scenarios (where the
hazards were different from the trained situations) were
tested. Eye tracking results showed that, in comparison
with untrained drivers, the trained drivers had significantly
higher hazard fixation probability (i.e., better HP) in both
near and far transfer scenarios. Note that in this study
(55), drivers’ glances toward the general areas where non-
immediate hazards may appear and materialize were used
as indication of good HP (hazard anticipation).
Moreover, researchers have started to combine multi-
ple training methods to enhance training effectiveness.
Isler et al. (80) examined a five-day training package that
combined participant commentary training with expert
feedback (on-road driving), video-based HP training,
and on-road driving practice (without commentary). The
results showed that the combined training package pro-
duced significant improvement on HP hit rate; in con-
trast, the vehicle handling training group and the control
group showed no such improvement.
Meir et al. (82) examined Act and Anticipate Hazard
Perception Training (AAHPT). The AAHPT hybrid
mode combined instructional training and active prac-
tice. The instructional training explained hazards using
example situations, where immediate and non-immediate
hazards were circled and highlighted for easier identifica-
tion. The practice training used a video-based HP task
that included different kinds of hazards, roads, and traf-
fic environments. Trainees were required to press a
response button each time they detected a hazard.
To comprehensively evaluate the effectiveness of HP
training, both near (apply directly to situations learned in
training) and far (apply to a new situation) transfer needs
to be considered (83,102). McDonald et al. (103)reviewed
19 studies published between 1980 and 2013 and high-
lighted evidence to support the effectiveness of various HP
training methods for both near and far transfer. However,
the authors also noted several limitations in these studies,
such as a lack of standardized tests and immediate assess-
ments. Only one study evaluated training effects after
about eight months (97). Since HP skills are likely to decay
at the early learning stage (84), long-term skill retention
studies are needed to confirm that training effects can be
sustained over a long period of time.
On the measures used to assess HP training effective-
ness, previous studies used HP measures such as HP
reaction time, hit rate, and eye tracking measures. These
measures focus on measuring only the perception and
awareness of hazards by excluding decision making and
vehicle control actions that follow the process of
perception. However, it is important to note that the
assessment of real-world benefits on reducing accidents
and crashes will require the examination of all three pro-
cesses including perception, decision making, and
response actions. There is a lack of studies examining
whether the expected benefits of HP training can trans-
late into reduction in crash rates (103).
In a driving simulator study, Zhang et al. (104) exam-
ined the effectiveness of a CBT program (SAFE-T) (105)
using hazard mitigation measures including velocity and
acceleration when the driver’s car is approaching the
hazard. The four tested scenarios included Bus Bicyclist,
Stop and Turn Left, Opposing Lane Roadwork, and Car
in Parking Lane, where other road users such as bicy-
clists, pedestrians, and vehicles were potential non-
immediate hazards (visible), and the driver should slow
down for them. The results showed that ‘‘the only signifi-
cant difference between placebo and trained drivers in
vehicular control is that trained drivers maintained a
larger deceleration when approaching the potential
hazard in the Bus Bicyclist Scenario’’ (104). This result
suggests that HP training needs to be combined with
hazard response and mitigation training to strengthen
the effects on driving safety and crash reduction. Future
studies need to use more holistic approaches that include
all three processes of perception, decision making, and
response actions to examine the effect of hazard training
on reducing accident and crash rates, using simulator
studies or real-world driving data analysis.
In summary, previous studies showed that a combina-
tion of complementary training approaches such as
instruction, expert demonstration, and active practice
with feedback and support is important and effective for
HP training (84). Table 2 provides a summary of the HP
training methods reviewed in this paper. In early training
stages, more instructions and simpler training methods
(e.g., picture- and video-based training) could help reduce
trainees’ mental workload; in later stages of training, the
combination of multiple training methods including
simulator-based training could further strengthen the
knowledge transfer from training courses to real-world
scenarios. Some evidence has been found to support the
effectiveness of the reported training programs in
improved HP performance. Methodological studies are
still needed to establish standard tests and indexes that
allow comparison across different training methods.
Future studies are also needed to examine long-term
effects of training to connect the short-term effects to
potential reduction in crash rates.
Discussion and Future Work
HP is often measured by hazard scenario questionnaires,
HP reaction time, hazard hit rate, and eye fixation
10 Transportation Research Record 00(0)
Table 2. Summary of Hazard Perception (HP) Training Methods Reviewed in This Paper
Paper
Content of training (hazards scenarios
involved in training or testing) Variables (groups) Evaluation methods Conclusion
Ivancic and Hesketh (83) In a left-hand traffic environment,
training scenarios include:
1. Left lane blocked with oncoming
vehicles.
2. Traffic signals at an intersection with
cross traffic.
3. Road sign with curved arrow,
followed by road curving sharply to
the right.
4. Both right lanes blocked with
oncoming vehicles.
5. ‘‘High wind’’ road sign followed by
speed limit sign.
6. Left lane blocked by a concrete
barrier.
Two training groups:
Error training group; Errorless
training group.
1. Number of errors (either a
crash or a police ticket)
committed on the transfer
tests; 2. Driving speed; 3. The
number of strategies recalled;
4. Self-perceived confidence
level.
Compared with errorless
learning, error training leads
to significantly better transfer
to driving tests and novel
driving situations.
Chapman et al. (87) In the training, a series of videos based
on films which include potentially
dangerous driving situations were
used as training material. No specific
scenarios were mentioned in the
paper. Participants needed to
anticipate the situation, comment on
scenarios or do button response
during the training.
Two groups of drivers were
evaluated:
Training intervention group;
Control group.
1. Driving measurements:
a. Free speed; b. Time headway.
2. Visual search measures:
a. Fixation duration; b.
Horizonal variance; c. Vertical
variance.
The training intervention
produced notable changes in
the drivers’ search patterns,
though some changes were no
longer detectable after a few
months.
McKenna et al. (36) The training method is commentary
drive.
The trained group was required to
watch a 21 min. video of various
road and traffic situations, with a
recorded commentary from a police
driver training program. The
commentary referred to potentially
hazardous events and how to
identify them.
Two groups:
Trained group (trained with
video and recorded
commentary); Untrained
group.
Two types of measures done
after the training:
1. Risk-taking measures:
Motorway speed; Questionnaire
speed; Video speed; Violation
questionnaire; Normal
following distance;
Uncomfortably close distance;
Gap acceptance.
2. Hazards perception measures:
Averaged reaction times over
each hazard.
Participants who received
training responded significantly
faster in the HP test than
those who did not.
(continued)
11
Table 2. (continued)
Paper
Content of training (hazards scenarios
involved in training or testing) Variables (groups) Evaluation methods Conclusion
Pollatsek et al. (96) In the training scenario, 10 scenarios
would be presented including left
fork, adjacent truck left turn,
intersection, and so forth.
Drivers moved symbols to indicate
hazards, and they would receive
feedback screens of each scenario.
In the simulator test, drivers followed
a leading vehicle and operated the
car just as they did in the real world
with an eye tracking device
mounted.
Two groups:
Trained group; Control group.
In training session, the
performance was measured by
probability risk recognized.
In simulator test session, a
series of criteria were
developed base on
participants’ eye fixation
behaviors. And a score of 0 or
1 would be given depending
on whether the risk was
identified.
Trained novice drivers were
almost twice as likely as
untrained drivers to fixate on
the area which contains
information about potential
hazards for both near and far
transfer scenarios.
Pradhan et al. (88) Drivers were trained using RAPT-3
which contains nine driving scenarios
selected from a set used in prior
studies. For example, the hidden
sidewalk scenario, left fork scenario,
left turn (reveal) scenario, right turn
(reveal) scenario and abrupt lane
change scenario, and so forth.
Two groups of drivers were
involved:
Trained with RAPT-3 group;
Control group.
A binary scoring method was
employed.
Drivers were given a score of 1
if a target area of potential
risk (Area of Interest) was
fixated on by them when they
were in the launch zone. A
score 0 was given if not.
The trained drivers were
significantly more likely to gaze
at areas of the roadway that
contained information
pertinent to risk reduction
(64.4%) than were the
untrained drivers (37.4%).
Significant training effects were
observed even in far transfer
scenarios.
Isler et al. (16) The HP dual task (both baseline test
and post-training test) requires the
participants to search for immediate
hazards on video-based traffic
scenarios while doing a tracking task
at the same time.
In video-based road commentary
training, participants were required
to provide a running verbal
commentary about any hazards they
detected.
Three groups for trained
drivers:
Two groups doing video-based
road commentary training.
One group needs to provide
verbal commentary and one
did not. One control group
watched commercial tapes.
In addition to that, one
experienced driver group did
not receive training.
In both dual task post-training
testing, two main measures
are used:
1. Number of hazards identified.
2. Reaction time.
After the road commentary
training, the detected hazards
percentage of the young
drivers improved to the level
of the experienced drivers and
was significantly higher than
the corresponding control
group.
Damm et al. (79) Five prototypical accident scenarios
were involved in testing:
1. Overtaking scenario
2. Pedestrian scenario
3. Opposite vehicle crossing scenario
4. Left crossroads scenario
5. Parked vehicle scenario
Three groups of drivers were
tested:
Traditionally trained; Novice
early-trained; Experienced
drivers.
To describe driving behavior:
1. Speed
2. Lateral positions (LPs)
To evaluate performance:
1. Reaction time
2. Obstacle avoidance/collision
No difference was detected
across groups for reaction
time. But in some scenarios,
position control by
traditionally trained drivers
was more conservative, and
early-trained drivers were far
more likely to respond with
efficient evasive action.
(continued)
12
Table 2. (continued)
Paper
Content of training (hazards scenarios
involved in training or testing) Variables (groups) Evaluation methods Conclusion
Isler et al. (80) Out of three groups, one, the higher-
order driving skill training group,
received training about HP. This
group would receive a series of
training activities including road
commentary, video-based HP on a
computer, on-road self-evaluation
driving exercise, and focus group-
based discussions.
All participants are allocated
into three groups:
Higher-order driving skill
training; Vehicle handling skill
training; Control group.
1. On-road driving assessment:
Visual Search; Speed Control;
Direction Control.
Each of these three skills is
classified into some
predefined levels.
2. HP test:
Percentage of Hazards
Detected; Percentage of
Action to Hazards.
The participants who received
higher-order driving skill
training showed a statistically
significant improvement in HP.
The participants who received
vehicle handling skill training
showed significant
improvements in on-road
direction control, but showed
no improvement in HP.
Taylor et al. (97) The RAPT training was evaluated in
this study. RAPT consisted of 11
scenarios. The training program
displayed sequences of photographs
from the drivers’ first-person
perspective in real-world conditions
(Amherst, MA). Participants had to
click critical locations in the scenario.
Participants were assigned to one of
two groups to explore the retention
effects of training a year following
training. Assessments were
conducted on-road (13 km field route
in Greenfield, MA).
Two groups of young drivers:
Trained group (RAPT training);
Placebo group.
One experienced driver group
(no training provided).
Eye movements were analyzed
to assess whether participants
anticipated a potential hazard.
The effects of training were
found to persist over time.
Immediately after training, the
trained group was found to
anticipate 65.8% of the hazards
(47.3% for placebo group).
Trained group anticipated
61.9% (37.7% for the control
group) of the hazards when
evaluated up to a year
following training.
Vlakveld et al. (55) A low-cost, fixed-base simulator
training program (SimRAPT) was
developed and evaluated to improve
drivers’ ability to anticipate potential
hazards. The training contained 10
scenarios (seven scenarios with
latent hazards and three scenarios
with no high priority hazards).
Participants drove three versions
(hazard detection drive, error drive,
and improvement drive) of each
hazard anticipation scenario. In the
hazard detection drive, no hazards
materialized, while in the error drive
the hazard materialized aggressively.
In the improvement drive, the latent
hazard manifested less aggressively.
Two groups—SimRAPT-Trained
group and untrained group.
The eye movements of both
groups were measured.
Nineteen scenarios were
evaluated of which seven were
near transfer and 12 were far
transfer.
The scores of the critical
scenarios were internally
consistent (alpha = 0.83)
implying that the 19 potential
hazard scenarios test and
measure one concept.
Compared with the placebo
group, the trained group
anticipated significantly more
hazards in near transfer
scenarios (83.61% versus
56.91%), far transfer scenarios
(70.95% versus 53.49%) and all
scenarios (75.60% versus
54.73%) together.
(continued)
13
Table 2. (continued)
Paper
Content of training (hazards scenarios
involved in training or testing) Variables (groups) Evaluation methods Conclusion
Petzoldt et al. (89) The computer-based training (CBT)
video sequence includes three
aspects: Pre-test on theoretical
knowledge; An instructional phase;
Actual training phase.
The actual training phase includes
short clips of traffic scenes
generated by artificial animations and
each video sequence contained two
or three relevant situations,
whereby sequences were paused at
various positions and questions were
presented.
In the training section, three
groups of participants
completed:
(a) CBT of cognitive driving
skills;
(b) A paper-based training of
cognitive driving skills;
(c) No learning intervention.
The main measurement is glance
behaviors. A glance sequence
is defined as the eye
movement from an unspecific
hazard indicator directly to a
relevant area.
The time between the
occurrence of the hazard
indicator and the first
completion of the glance
sequence was used as a
measurement.
The experiment confirmed that
CBT participants exhibited
earlier glances toward critical
cues and relevant areas than
the participants from paper-
based training group and
control group.
Wetton et al. (84) The training package used in this study
was inspired by the package
developed by Poulsen et al. (106),
which included both hybrid
commentary and what happens next
exercises. In his study, three new
packages were developed, which
focus on: 1. expert commentary
drive exercises; 2. hybrid
commentary drive exercises; and 3.
what happens next exercises.
Participants are divided into five
training groups:
What happens next training;
Expert commentary training;
Hybrid commentary training
(i.e., expert plus self-
generated commentaries);
The full training package (i.e.,
what happens next plus hybrid
commentary training); The
placebo control condition.
After the training, reaction time
was used as the measurement
for HP tests.
All training interventions
significantly improved HP
response times immediately
after the training. The full
training interventions resulted
in the largest improvement,
and the what happens next
training intervention led to the
least improvement.
Samuel et al. (90) In the RoadAware
TM
(RA) training
program, a program-controlled
simulation drive was shown to the
trainees from the driver’s
perspective, and trainees could see
the surrounding environment and
click the hazards they saw. Trainees
would receive auditory and visual
feedback on whether they have found
critical potential hazards. Both near
and far transfer scenarios were
included in the test simulated
hazardous scenarios.
Two groups:
Trained group; Untrained group.
A logistic regression model was
used within the framework of
generalized estimating
equations (GEE), including
three binary fixed effects:
1. Driver group (RA and
placebo)
2. Hazard type (materialized
versus unmaterialized hazard)
3. Visibility of hazard instigator
(visible or hidden).
The results showed that trained
drivers were more likely to
anticipate hazards than
untrained drivers, both in near
transfer scenarios and far
transfer scenarios.
(continued)
14
Table 2. (continued)
Paper
Content of training (hazards scenarios
involved in training or testing) Variables (groups) Evaluation methods Conclusion
Meir et al. (82) Active members observed video-based
traffic scenes with button response.
Instructional members watched a
tutorial including both written
material and video-based examples of
HP. Hybrid members received a
theoretical component followed by a
succinct active component. In the
training session, participants
observed more materialized
situations to strengthen their
weakness.
Four groups:
1. Three AAHPT mode groups:
Instructional, Active, Hybrid
groups
2. Control group.
In the HP test, various
measurements were applied:
1. Eye tracking measures:
Minimum fixation duration;
Minimum dispersion
considered a fixation;
Maximum consecutive sample
loss.
2. Other HP measures:
Response time, Response
sensitivity, Searching
strategies; Traffic scene
classification.
The active and hybrid groups
were more aware of potential
hazards relative to the control
group.
McDonald et al. (103) A literature review was conducted on
hazard anticipation training for young
drivers. Studies were only included if
they involved an assessment of
training outcomes and included at
least one group of younger drivers
(\21 years).
Only studies with a younger
driver (\21) group were
included.
A critical review was
implemented on 19 peer-
reviewed studies. Training
programs, outcome measures,
study designs, length of follow
up and driving experience
were captured.
Studies were found to have used
a variety of training methods
ranging from interactive
computer programs and
videos to simulation and
commentary training. Four
studies were found to include
an on-road evaluation. Most
studies were found to have
evaluated short-term
outcomes.
Castro et al. (27) The training uses the complete version
of the same 16 videos which were
used in the Spanish Hazard
Perception test done before,
revealing the hazards with a voice
containing relevant information about
where to allocate attention in the
complex driving scene.
Three independent variables:
1. Training condition (Trained
and Untrained group)
2. Experience (learners, novices,
and experienced drivers)
3. Recidivism condition (non-
offenders versus re-offenders)
A scoring system was
developed. For example, when
answering ‘‘What is the
hazard?’’, participants received
two points if they gave an
exact description of the hazard
(e.g., red car in the left lane),
one point if they gave a
partially correct answer (e.g., a
car on the left, but missing
critical details) and zero points
for an incorrect answer.
This training shows significant
positive effects for all types
and groups of participants
(continued)
15
Table 2. (continued)
Paper
Content of training (hazards scenarios
involved in training or testing) Variables (groups) Evaluation methods Conclusion
Zafian et al. (98) The Engaged Driver Training System
(EDTS) is an iPad-based training
program for hazard anticipation and
engagement consisting of eight
training scenarios. The user grasps
the iPad in both hands as one would
with a steering wheel and steering is
handled clockwise or
counterclockwise by tilting the iPad.
Acceleration/deceleration are
handled by the right thumb sliding a
gas pedal icon up and down a slide
bar on the right side of the interface.
Similarly, scanning left and right is
handled by sliding the eye icon
horizontally with the left thumb on a
horizontal slide bar. Users can
identify hazards by pausing the drive
and then tapping on it with their
finger. Correct features are
highlighted with a glowing yellow
outline.
Four groups: Trained teen
group, Trained parent dyad,
Placebo teen and Parent
placebo dyad.
On-road evaluation was
conducted in the town of
Amherst along a 2.3-mile
route with 14 pre-identified
scenarios of key interest
(including various types of
intersections, turns, curves,
crosswalks and turns across
path situations). A logistic
regression model was used
within the framework of
generalized estimating
equations (GEE), including
three binary fixed effects:
1. Type of training (RA and
placebo)
2. Group (solo and dyad)
3. Scenarios
The EDTS-trained group was
found to be markedly better at
anticipating hazards compared
with the placebo group (71%
versus 44%).
Thomas et al. (101) This study evaluated the impact of PC-
based RAPT on young drivers’
crashes and traffic violations. Nine
training scenarios were evaluated.
RAPT was reprogrammed in Adobe
Air.
Total of 5,251 young drivers
(RAPT group or Comparison
group). Comparison group
received a pre-test but no
training.
An analysis of equivalency was
utilized to demonstrate that
the group assignment was
effective at producing
equivalent groups. Analyses of
pre-test and post-test data
were conducted to assess
trainee performance.
Researchers used Cox
regression analysis to evaluate
the number of weeks after
licensure at which each
participant had their first crash
(time to first crash).
A significant treatment effect
was found for males but not
for females. RAPT-trained
males showed an
approximately 23.7% lower
crash rate than the male
comparison group.
(continued)
16
Table 2. (continued)
Paper
Content of training (hazards scenarios
involved in training or testing) Variables (groups) Evaluation methods Conclusion
Young et al. (93) As with Young et al. (107), the training
video consists of footage of driving
around Nottingham (UK) filmed from
the perspective of the driver.
Participants were divided into
four groups based on whether
exposed to commentary
during training and whether
required to give commentary
during testing.
Two measures of behavioral data
employed:
1. Percentage of predefined
hazards correctly identified;
2. Response times.
Two measures of eye tracking
data employed:
1. Number of fixations;
2. Mean fixation durations;
3. Vertical and horizontal
variance in fixation location.
Giving a live commentary is
detrimental to HP.
Commentary exposure resulted
in an initial increase in the
accuracy of HP responses, but
this effect only lasted a very
limited amount of time.
Horswill et al. (94) In the HP tests, participants were
asked to watch a series of video clips
which contained potential traffic
conflicts, on a computer screen.
Participants in all four groups re-
watched all of the video clips used in
the first HP test. In graph feedback
condition, participants saw a bar
graph. In a video feedback condition,
participants watched videos with
superimposed annotations. In the
last group, the participants received
all kinds of feedback.
Four groups are divided by the
way of providing feedback:
(a) The graph-based feedback;
(b) The video-based feedback;
(c) Both; (d) No feedback.
In this study, participants’
reaction times were used to
measure their HP ability. A
standard score was gained
from the reaction time
through a standardization
process.
All three types of feedback
resulted in an improvement in
HP performance. Also, the
combination of video-based
and graph-based feedback
resulted in the largest
improvement in HP
performance.
Zafian et al. (99) This study explored the longitudinal
evaluation of the EDTS training
program. Participants were trained
on eight scenarios as in Zafian et al.
(98). Drivers were evaluated a week
after training and a second time,
seven months after training.
Two groups: Trained group and
Placebo group.
A logistic regression model was
used within a GEE framework
to analyze the binomially
distributed, binary-coded data.
Type of training (EDTS or
placebo) and type of
evaluation (after one week or
after seven months) were
included as fixed effects in the
model.
Seven months after training, the
placebo group’s hazard
anticipation performance was
found to have increased to
that observed for EDTS-
trained teens a week after
training. Overall EDTS-trained
teens anticipated significantly
more latent hazards than
placebo-trained teens. These
differences were found to be
consistent for both near
transfer and far transfer.
(continued)
17
Table 2. (continued)
Paper
Content of training (hazards scenarios
involved in training or testing) Variables (groups) Evaluation methods Conclusion
Unverricht et al. (100) This paper explored a meta-analysis of
all studies that have explored the
effectiveness of latent hazard
anticipation training programs. The
review focused on 19 peer-reviewed
training studies that utilized eye
movements to measure
improvements in drivers’ latent
hazard anticipation following training.
Meta-analysis and literature
review.
Meta-analysis explored the role
of four moderating factors: 1.
mode of delivery: PC-based or
non-PC-based; 2. presentation
of training: egocentric or
exocentric; 3. method of
evaluation: on-road or driving
simulator; 4. age of sample:
teen drivers (16–17 years) or
young drivers (18–21 years).
The meta-analysis suggested that
superficial improvements in
training do not necessarily
improve training effectiveness.
Training programs with both
egocentric and exocentric
training views achieved greater
levels of hazard anticipation
performance compared with
training programs with either
view but not both.
Zhang et al. (104) A study was undertaken to determine
whether the effectiveness of a
training program at improving hazard
anticipation and mitigation skills is
moderated by driving style. A
computer-based training called SAFE-
T was administered and evaluated.
SAFE-T consists of four training
modules and a total of 12 training
scenarios.
Two groups: Trained and
untrained groups.
Drivers were classified as careful
or careless based both on
measures designed to evaluate
two general traits (sensation
seeking and aggressiveness)
and two driving-specific
behaviors (aggressive driving
behaviors, and driving
violations and errors).
Analysis showed that training
improved the latent hazard
anticipation behavior of careful
drivers, but not careless
drivers. Across all scenarios,
the main effect of training was
consistent. Trained careful
drivers anticipated 84.4% of
the hazards (compared with
58.9% for placebo careful
drivers).
Kahana-Levy et al. (91) Both training phase and transfer phase
use videos of real-world driving
filmed from a driver’s perspective as
material. In training section, each
movie was presented three times. In
transfer phase, the process of an
unmaterialized hazardous scenario
becoming materialized was displayed.
Two groups for young
inexperienced drivers: Trained
group; Control group.
In addition, there was another
group of experienced drivers.
Four measures are employed:
1. Number of fixations;
2. Reaction time;
3. Horizontal spread of search;
4. Vertical spread of search.
In training session, young
inexperienced drivers
gradually increased their focus
on visible materialized hazards.
In the transfer session, both
trained groups focused on
hazards earlier compared with
untrained drivers.
Horswill et al. (78) Five types of newly designed activities
listed below were involved in the
training courses and they were
presented in a recursive and
progressive order.
1. What Happens Next. Added real
crash clips.
2. Crash Analysis.
3. Commentary Drive. Added follow
up presentation to each exercise.
4. Video Review Feedback with
feedback provided.
5. Real World Transfer undertaken
during real driving between online
sessions.
Two groups of participants:
Trained group and Waitlist
control group.
Two ways of HP measurements:
1. Response time
2. A scoring system in HP test.
Score of 1 for an answer that
matched expert’s prediction.
There are also a few additional
measurements of other
driving behaviors.
The study found that this
training course can significantly
improve drivers’ HP response
time and hazard prediction
scores.
18
measures including fixation probability, fixation reaction
time, fixation duration, and fixation variance. Previous
studies have examined human factors affecting HP
including experience, aging, fatigue, distraction, and the
use of alcohol and drugs. Various training intervention
methods have been used to improve HP. In general, there
is evidence in the literature showing the effectiveness of
HP training for shorter HP reaction time, higher hazard
hit rate, and better eye scan patterns (more spread scan,
more anticipatory scan). A combination of complemen-
tary training approaches such as instruction, expert
demonstration, and active practice with feedback and
support improved measured behaviors. This review has
identified the following areas for future work.
Standardized HP Tests
A variety of different tests and measures have been used
in the literature to measure HP. The lack of standard
tests prevents the comparison of different training meth-
ods across different studies. Transportation authorities
need standardized tests for driver licensing programs.
Some countries such as United Kingdom and Australia
have implemented HP tests in the driver licensing pro-
cess, but many countries such as China, India and the
United States have not. The majority of HP tests
reported in the literature are video-based, which has the
benefits of lower cost and being easier to implement than
simulator-based tests. However, simulator-based tests
are expected to be more accurate at measuring novice
drivers’ HP skills because video-based tests do not have
the requisite vehicle control components. Novice drivers’
skill limitation is more likely to be exposed in driving
simulators when they must concurrently work on both
vehicle control and HP.
Another limitation of the current HP measures is that
they focus on examining a single early moment when a
hazard is first attended or recognized. However, as sug-
gested by Markkula et al. (37), the process of hazard
response is a continuous flow of assessing, responding,
and reassessing the situation. Future studies need to
develop methods that can continuously measure HP dur-
ing the entire process of hazard response. More holistic
approaches could be used to measure the combination of
HP and hazard mitigation processes by including mea-
sures such as speed loss and lane keeping measures (104,
108,109).
Improving HP Training Programs
Future studies need to compare different training meth-
ods and improve the design of training programs by inte-
grating the most effective training approaches. Most
training programs reported in the literature were short
(within one hour), and trainees received the training only
once. Repeated training with multiple sessions could
improve training effects and help the effects to be sus-
tained over a longer period of time. Most of the existing
training programs were developed in developed countries
such as Australia, the United Kingdom, and the United
States. Traffic situations and rules are different in other
countries, such as China and India, where drivers have
more interaction with other road users including cyclists
and pedestrians. This means that training programs need
to consider cultural difference and adapt to different
countries. In addition, although many researchers have
emphasized the need for anticipatory glances on non-
immediate hazards to gain HP and being prepared for
any non-immediate hazard turning into an immediate
hazard, there is a lack of research and data showing the
benefits of anticipatory HP training in impact on crash
rate. Future studies are needed to establish evidence to
support the effectiveness of training on crash outcomes
using both simulator studies and road crash data
analysis.
While most HP training programs focused on improv-
ing anticipation and awareness of non-immediate
hazards before they become immediate hazards, fewer
studies have designed HP training focusing on proper
recognition of immediate hazards. Forensic research and
simulator studies (110–112) that analyzed driver behavior
in different types of crashes such as left turn across path
from opposite direction and lead vehicle front-to-rear (or
rear-end) crashes have shown human limitations in
recognizing immediate hazards. Even after the immediate
hazards are visible to a driver, they may not be able to
judge the speed, gap, and future positions of the vehicles
correctly for proper assessment of an imminent collision.
Future studies are needed to improve HP training includ-
ing speed and gap judgment in this type of immediate
hazard scenario that frequently causes fatal crashes.
New Questions Brought by Autonomous Vehicles
Autonomous vehicles are expected to be the future. In
partially and fully automated driving, the role of drivers
is shifted from operational control to monitoring and
supervising. This means a reduced need of vehicle control
skills and an increased importance of HP skills. Since
autonomous vehicles are equipped with sensors and algo-
rithms that can monitor the environment for hazards,
drivers’ workload on monitoring hazards is expected to
reduce. However, autonomous vehicles may present new
hazards when algorithms fail or when the driving sce-
nario exceeds the design limits of the algorithms. A recent
study showed that after a period of staying in the auto-
mated driving mode, when the automation gives vehicle
control back to the human driver, drivers need a long
Cao et al 19
time (at least about 8 s) to regain awareness of the envi-
ronment and be prepared to take over control smoothly
(113). To successfully perceive hazards in the condition
of partially automated vehicles, drivers need to know the
capabilities of autonomous vehicles as well as their lim-
itations. Autonomous vehicles bring many new questions
for HP, and more future studies are needed to answer
these questions.
Conclusion
Many studies have been conducted to quantify HP in
driving and factors affecting it. The major approach is to
separate HP from hazard mitigation and analyze them
individually. Previous tests and measures are good at
measuring HP alone, but more holistic approaches are
needed to combine the assessment of HP and hazard
mitigation to connect hazard perception directly to driv-
ing safety and crash rate. Various HP training programs
have been developed using pictures, computers, and driv-
ing simulators. Their effects on improving the HP pro-
cess are supported by evidence from measured behaviors,
but future studies are needed to further examine the
impact on crash data. The current review could provide a
quick reference for the best practice in HP measures and
a summary of existing HP training methods for future
researchers and designers of training courses to consider.
Author Contributions
The authors confirm contribution to the paper as follows: study
conception and design: S. Cao, J. Niu; data collection: S. Cao,
S. Samuel, Y. Murzello, X. Zhang; analysis and interpretation
of results: S. Cao, S. Samuel, W. Ding, J. Niu; draft manuscript
preparation: S. Cao, S. Samuel, X. Zhang, J. Niu. All authors
reviewed the results and approved the final version of the
manuscript.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The authors disclosed receipt of the following financial support
for the research, authorship, and/or publication of this article:
This work was supported by National Key Technology R&D
Program of China (2014BAK01B01), State Scholarship Fund
from China Scholarship Council (201208110144), and
Fundamental Research Funds for the Central Universities,
China (FRF-TP-14-026A2) to J.N., Natural Sciences and
Engineering Research Council of Canada Discovery Grant
(RGPIN-2019-05304) to S.S., and Natural Sciences and
Engineering Research Council of Canada Discovery Grant
(RGPIN-2015-04134) to S.C.
ORCID iDs
Shi Cao https://orcid.org/0000-0002-6448-6674
Siby Samuel https://orcid.org/0000-0002-2168-8479
Yovela Murzello https://orcid.org/0000-0002-4563-2373
Data Accessibility Statement
Data available on request.
References
1. Haworth, N., M. Symmons, and N. Kowadlo. Hazard Per-
ception by Inexperienced Motorcyclists. Monash Univer-
sity, Melbourne, Victoria, Australia, 2000.
2. White, M. J., L. C. Cunningham, and K. Titchener. Young
Drivers’ Optimism Bias for Accident Risk and Driving
Skill: Accountability and Insight Experience Manipula-
tions. Accident Analysis & Prevention, Vol. 43, No. 4, 2011,
pp. 1309–1315.
3. Borowsky, A., and T. Oron-Gilad. Exploring the Effects
of Driving Experience on Hazard Awareness and Risk Per-
ception via Real-Time Hazard Identification, Hazard Clas-
sification, and Rating Tasks. Accident Analysis and
Prevention, Vol. 59, 2013, pp. 548–565. https://doi.org/10.
1016/j.aap.2013.07.008.
4. Borowsky, A., D. Shinar, and T. Oron-Gilad. Age, Skill,
and Hazard Perception in Driving. Accident Analysis and
Prevention, Vol. 42, No. 4, 2010, pp. 1240–1249.
https://doi.org/10.1016/j.aap.2010.02.001.
5. Cheng, A. S. K., T. C. K. Ng, and H. C. Lee. A Compari-
son of the Hazard Perception Ability of Accident-Involved
and Accident-Free Motorcycle Riders. Accident Analysis
and Prevention, Vol. 43, No. 4, 2011, pp. 1464–1471.
https://doi.org/10.1016/j.aap.2011.02.024.
6. Horswill, M. S., A. Hill, and M. Wetton. Can a Video-
Based Hazard Perception Test Used for Driver Licensing
Predict Crash Involvement? Accident Analysis and Preven-
tion, Vol. 82, 2015, pp. 213–219. https://doi.org/10.1016/j.
aap.2015.05.019.
7. Horswill, M. S., and F. P. McKenna. Drivers’ Hazard Per-
ception Ability: Situation Awareness on the Road. In A
Cognitive Approach to Situation Awareness: Theory and
Application (Banbury, S., and S. Tremblay, eds.), Ashgate,
Aldershot, UK, 2004, pp. 155–175.
8. Horswill, M. S. Hazard Perception in Driving. Current
Directions in Psychological Science, Vol. 25, No. 6, 2016,
pp. 425–430.
9. Endsley, M. R. Situation Awareness Global Assessment
Technique (SAGAT). Proc., IEEE 1988 National Aero-
space and Electronics Conference, Dayton, OH, IEEE,
New York, 1988.
10. Endsley, M. R. Toward a Theory of Situation Awareness
in Dynamic Systems. Human Factors: The Journal of the
Human Factors and Ergonomics Society, Vol. 37, No. 1,
1995, pp. 32–64. https://doi.org/10.1518/0018720957790
49543.
11. Deery, H. A. Hazard and Risk Perception Among Young
Novice Drivers. Journal of Safety Research, Vol. 30, No. 4,
20 Transportation Research Record 00(0)
1999, pp. 225–236. https://doi.org/10.1016/S0022-4375(99)
00018-3.
12. Crundall, D., P. Chapman, S. Trawley, L. Collins, E. Van
Loon, B. Andrews, and G. Underwood. Some Hazards
Are More Attractive Than Others: Drivers of Varying
Experience Respond Differently to Different Types of
Hazard. Accident Analysis and Prevention, Vol. 45, 2012,
pp. 600–609. https://doi.org/10.1016/j.aap.2011.09.049.
13. Sagberg, F., and T. Bjørnskau. Hazard Perception and
Driving Experience Among Novice Drivers. Accident Anal-
ysis and Prevention, Vol. 38, No. 2, 2006, pp. 407–414.
https://doi.org/10.1016/j.aap.2005.10.014.
14. Vlakveld, W. P. A Comparative Study of Two Desktop
Hazard Perception Tasks Suitable for Mass Testing in
Which Scores Are not Based on Response Latencies. Trans-
portation Research Part F: Traffic Psychology and Beha-
viour, Vol. 22, 2014, pp. 218–231. https://doi.org/10.1016/j.
trf.2013.12.013.
15. Malone, S., and R. Bru
¨nken. The Role of Ecological Valid-
ity in Hazard Perception Assessment. Transportation
Research Part F: Traffic Psychology and Behaviour, Vol.
40, 2016, pp. 91–103. https://doi.org/10.1016/j.trf.2016.04.
008.
16. Isler, R. B., N. J. Starkey, and A. R. Williamson. Video-
Based Road Commentary Training Improves Hazard Per-
ception of Young Drivers in a Dual Task. Accident Analy-
sis and Prevention, Vol. 41, No. 3, 2009, pp. 445–452.
https://doi.org/10.1016/j.aap.2008.12.016.
17. Smith, S. S., M. S. Horswill, B. Chambers, and M. Wetton.
Hazard Perception in Novice and Experienced Drivers: The
Effects of Sleepiness. Accident Analysis and Prevention, Vol.
41, No. 4, 2009, pp. 729–733. https://doi.org/10.1016/j.aap.
2009.03.016.
18. Chan, E., A. K. Pradhan, A. Pollatsek, M. A. Knodler,
and D. L. Fisher. Are Driving Simulators Effective Tools
for Evaluating Novice Drivers’ Hazard Anticipation, Speed
Management, and Attention Maintenance Skills? Transpor-
tation Research Part F: Traffic Psychology and Behaviour,
Vol. 13, No. 5, 2010, pp. 343–353. https://doi.org/10.1016/j.
trf.2010.04.001.
19. Takahashi, R., M. Kobayashi, T. Sasaki, Y. Yokokawa,
H. Momose, and T. Ohhashi. Driving Simulation Test for
Evaluating Hazard Perception: Elderly Driver Response
Characteristics. Transportation Research Part F: Traffic
Psychology and Behaviour, Vol. 49, 2017, pp. 257–270.
https://doi.org/10.1016/j.trf.2017.07.003.
20. Scialfa, C. T., D. Borkenhagen, J. Lyon, M. Descheˆ nes,
M. Horswill, and M. Wetton. The Effects of Driving Expe-
rience on Responses to a Static Hazard Perception Test.
Accident Analysis and Prevention, Vol. 45, 2012,
pp. 547–553. https://doi.org/10.1016/j.aap.2011.09.005.
21. Wetton, M. A., M. S. Horswill, C. Hatherly, J. M. Wood,
N. A. Pachana, and K. J. Anstey. The Development and
Validation of Two Complementary Measures of Drivers’
Hazard Perception Ability. Accident Analysis and Preven-
tion, Vol. 42, No. 4, 2010, pp. 1232–1239. https://doi.
org/10.1016/j.aap.2010.01.017.
22. Scialfa, C. T., D. Borkenhagen, J. Lyon, and M.
Descheˆ nes. A Comparison of Static and Dynamic Hazard
Perception Tests. Accident Analysis and Prevention, Vol.
51, 2013, pp. 268–273. https://doi.org/10.1016/j.aap.2012.
12.006.
23. Wetton, M. A., A. Hill, and M. S. Horswill. The Develop-
ment and Validation of a Hazard Perception Test for Use
in Driver Licensing. Accident Analysis and Prevention, Vol.
43, No. 5, 2011, pp. 1759–1770. https://doi.org/10.1016/j.
aap.2011.04.007.
24. Malone, S., and R. Bru
¨nken. Hazard Perception Assess-
ment – How Much Ecological Validity is Necessary? Proce-
dia Manufacturing, Vol. 3, 2015, pp. 2769–2776. https://doi.
org/10.1016/j.promfg.2015.07.713.
25. Jackson, L., P. Chapman, and D. Crundall. What Happens
Next? Predicting Other Road Users’ Behaviour as a Func-
tion of Driving Experience and Processing Time. Ergo-
nomics, Vol. 52, No. 2, 2009, pp. 154–164. https://doi.
org/10.1080/00140130802030714.
26. Castro, C., J. L. Padilla, J. Roca, I. Benı
´tez, P. Garcı
´a-Fer-
na
´ndez, B. Este
´vez, M. F. Lo
´pez-Ramo
´n, and D. Crundall.
Development and Validation of the Spanish Hazard Per-
ception Test. Traffic Injury Prevention, Vol. 15, No. 8,
2014, pp. 817–826. https://doi.org/10.1080/15389588.2013.
879125.
27. Crundall, D. Hazard Prediction Discriminates Between
Novice and Experienced Drivers. Accident Analysis and
Prevention, Vol. 86, 2016, pp. 47–58. https://doi.org/10.
1016/j.aap.2015.10.006.
28. Castro, C., P. Ventsislavova, E. Pen
˜a-Suarez, A. Gugliotta,
P. Garcia-Fernandez, E. Eisman, and D. Crundall. Proac-
tive Listening to a Training Commentary Improves Hazard
Prediction. Safety Science, Vol. 82, 2016, pp. 144–154.
https://doi.org/10.1016/j.ssci.2015.09.018.
29. Gugliotta, A., P. Ventsislavova, P. Garcia-Fernandez, E.
Pen
˜a-Suarez, E. Eisman, D. Crundall, and C. Castro. Are
Situation Awareness and Decision-Making in Driving
Totally Conscious Processes? Results of a Hazard Predic-
tion Task. Transportation Research Part F: Traffic Psychol-
ogy and Behaviour, Vol. 44, 2017, pp. 168–179. https://doi.
org/10.1016/j.trf.2016.11.005.
30. Ventsislavova, P., D. Crundall, T. Baguley, C. Castro, A.
Gugliotta, P. Garcia-Fernandez, W. Zhang, Y. Ba, and Q.
Li. A Comparison of Hazard Perception and Hazard Pre-
diction Tests Across China, Spain and the UK. Accident
Analysis and Prevention, Vol. 122, 2019, pp. 268–286.
https://doi.org/10.1016/j.aap.2018.10.010.
31. Ventsislavova, P., A. Gugliotta, E. Pena
˜-Suarez, P. Garcia-
Fernandez, E. Eisman, D. Crundall, and C. Castro. What
Happens When Drivers Face Hazards on the Road? Acci-
dent Analysis and Prevention, Vol. 91, 2016, pp. 43–54.
https://doi.org/10.1016/j.aap.2016.02.013.
32. Crundall, D., and V. Kroll. Prediction and Perception of
Hazards in Professional Drivers: Does Hazard Perception
Skill Differ Between Safe and Less-Safe Fire-Appliance
Drivers? Accident Analysis and Prevention, Vol. 121, 2018,
pp. 335–346. https://doi.org/10.1016/j.aap.2018.05.013.
33. Ventsislavova, P., and D. Crundall. The Hazard Prediction
Test: A Comparison of Free-Response and Multiple-Choice
Formats. Safety Science, Vol. 109, 2018, pp. 246–255.
https://doi.org/10.1016/j.ssci.2018.06.004.
Cao et al 21
34. Castro, C., J. L. Padilla, P. Doncel, P. Garcia-Fernandez,
P. Ventsislavova, E. Eisman, and D. Crundall. How Are
Distractibility and Hazard Prediction in Driving Related?
Role of Driving Experience as Moderating Factor. Applied
Ergonomics, Vol. 81, 2019, p. 102886. https://doi.org/10.
1016/j.apergo.2019.102886.
35. Shahar, A., C. F. Alberti, D. Clarke, and D. Crundall.
Hazard Perception as a Function of Target Location and
the Field of View. Accident Analysis and Prevention, Vol.
42, No. 6, 2010, pp. 1577–1584. https://doi.org/10.1016/j.
aap.2010.03.016.
36. McKenna, F. P., M. S. Horswill, and J. L. Alexander. Does
Anticipation Training Affect Drivers’ Risk Taking? Journal
of Experimental Psychology: Applied, Vol. 12, No. 1, 2006,
pp. 1–10. https://doi.org/10.1037/1076-898X.12.1.1.
37. Markkula, G., J. Engstro
¨m, J. Lodin, J. Ba
¨rgman, and T.
Victor. A Farewell to Brake Reaction Times? Kinematics-
Dependent Brake Response in Naturalistic Rear-End
Emergencies. Accident Analysis & Prevention, Vol. 95,
2016, pp. 209–226.
38. Hoffman, Y., and T. Rosenbloom. Driving Experience
Moderates the Effect of Implicit Versus Explicit Threat
Priming on Hazard Perception Test. Accident Analysis and
Prevention, Vol. 92, 2016, pp. 82–88. https://doi.org/10.
1016/j.aap.2016.03.019.
39. Chignell, M., T. Tong, S. Mizobuchi, and W. Walmsley.
Combining Speed and Accuracy Into a Global Measure of
Performance. Proceedings of the Human Factors and Ergo-
nomics Society, Vol. 58, No. 1, 2014, pp. 1442–1446.
40. Borowsky, A., T. Oron-Gilad, A. Meir, and Y. Parmet.
Drivers’ Perception of Vulnerable Road Users: A Hazard
Perception Approach. Accident Analysis and Prevention,
Vol. 44, No. 1, 2012, pp. 160–166. https://doi.org/10.1016/j.
aap.2010.11.029.
41. Rees, G., C. Russell, C. D. Frith, and J. Driver. Inatten-
tional Blindness Versus Inattentional Amnesia for Fixated
but Ignored Words. Science, Vol. 286, No. 5449, 1999,
pp. 2504–2507. https://doi.org/10.1126/science.286.
5449.2504.
42. Crundall, D., G. Underwood, and P. Chapman. Attending
to the Peripheral World While Driving. Applied Cognitive
Psychology, Vol. 16, 2002, pp. 459–475. https://doi.org/10.
1002/acp.806.
43. Sperling, G., and M. J. Melchner. The Attention Operating
Characteristic: Examples From Visual Search. Science,
Vol. 202, No. 4365, 1978, pp. 315–318. https://doi.org/10.
1126/science.694536.
44. Summala, H., D. Lamble, and M. Laakso. Driving Experi-
ence and Perception of the Lead Car’s Braking When
Looking at In-Car Targets. Accident Analysis and Preven-
tion, Vol. 30, No. 4, 1998, pp. 401–407. https://doi.org/10.
1016/S0001-4575(98)00005-0.
45. Hajiseyedjavadi, F., T. Zhang, R. Agrawal, M. Knodler,
D. Fisher, and S. Samuel. Effectiveness of Visual Warnings
on Young Drivers Hazard Anticipation and Hazard Miti-
gation Abilities. Accident Analysis and Prevention, Vol. 116,
2017, pp. 41–52. https://doi.org/10.1016/j.aap.2017.11.037.
46. Pradhan, A. K., K. R. Hammel, R. DeRamus, A. Pollat-
sek, D. A. Noyce, and D. L. Fisher. Using Eye Movements
to Evaluate Effects of Driver Age on Risk Perception in a
Driving Simulator. Human Factors, Vol. 47, No. 4, 2005,
pp. 840–852. https://doi.org/10.1518/001872005775570961.
47. Samuel, S., and D. L. Fisher. Evaluation of the Minimum
Forward Roadway Glance Duration. Transportation
Research Record: Journal of the Transportation Research
Board, 2015. 2518: 9–17.
48. Albert, B., and T. Tullis. Measuring the User Experience:
Collecting, Analyzing, and Presenting Usability Metrics.
Newnes, London, UK, 2013.
49. Underwood, G., D. Crundall, and P. Chapman. Driving
Simulator Validation With Hazard Perception. Transporta-
tion Research Part F: Traffic Psychology and Behaviour,
Vol. 14, No. 6, 2011, pp. 435–446. https://doi.org/10.1016/j.
trf.2011.04.008.
50. Yan, Y., X. Wang, L. Shi, and H. Liu. Influence of Light
Zones on Drivers’ Visual Fixation Characteristics and
Traffic Safety in Extra-Long Tunnels. Traffic Injury Pre-
vention, Vol. 18, No. 1, 2017, pp. 102–110.
51. Shinar, D., E. D. McDowell, and T. H. Rockwell. Eye
Movements in Curve Negotiation. Human Factors, Vol.
19, No. 1, 1977, pp. 63–71.
52. Crundall, D. E., and G. Underwood. Effects of Experience
and Processing Demands on Visual Information Acquisi-
tion in Drivers. Ergonomics, Vol. 41, No. 4, 1998,
pp. 448–458. https://doi.org/10.1080/001401398186937.
53. Mourant, R. R., and T. H. Rockwell. Strategies of Visual
Search by Novice and Experienced Drivers. Human Fac-
tors, Vol. 14, No. 4, 1972, pp. 325–335. https://doi.org/10.
1177/001872087201400405.
54. Stahl, P., B. Donmez, and G. A. Jamieson. Eye Glances
Towards Conflict-Relevant Cues: The Roles of Anticipa-
tory Competence and Driver Experience. Accident Analysis
& Prevention, Vol. 132, 2019, p. 105255.
55. Vlakveld, W., M. R. E. Romoser, H. Mehranian, F. Diete,
A. Pollatsek, and D. L. Fisher. Do Crashes and Near
Crashes in Simulator-Based Training Enhance Novice
Drivers’ Visual Search for Latent Hazards? Transportation
Research Record: Journal of the Transportation Research
Board, 2011. 2265: 153–160.
56. Bonfiglio, D. J., Y. C. Lee, N. La Voie, and J. Parker. Pro-
gression of Hazard Perception Knowledge, Licensure Sta-
tus and Driving Experience Among Teen Drivers.
Proceedings of the Human Factors and Ergonomics Society,
Vol. 58, No. 1, 2014, pp. 2136–2140.
57. Cao, S., Y. Qin, X. Jin, L. Zhao, and M. Shen. Effect of
Driving Experience on Collision Avoidance Braking: An
Experimental Investigation and Computational Modelling.
Behaviour and Information Technology, Vol. 33, No. 9,
2014, pp. 929–940. https://doi.org/10.1080/0144929X.2014.
902100.
58. Horswill, M. S., S. A. Marrington, C. M. McCullough, J.
Wood, N. A. Pachana, J. McWilliam, and M. K. Raikos.
The Hazard Perception Ability of Older Drivers. Journals
of Gerontology Series B: Psychological Sciences and Social
Sciences, Vol. 63, No. 4, 2008, pp. P212–P218. https://doi.
org/10.1093/geronb/63.4.P212.
59. Groeger, J. A. Youthfulness, Inexperience, and Sleep Loss:
The Problems Young Drivers Face and Those They Pose
22 Transportation Research Record 00(0)
for Us. Injury prevention, Vol. 12, Supplement 1, 2006,
pp. i19–i24.
60. Burge, R., and A. Chaparro. The Effects of Texting and
Driving on Hazard Perception. Proceedings of the Human
Factors and Ergonomics Society, Vol. 56, No. 1, 2012,
pp. 715–719.
61. He, J., A. Chaparro, B. Nguyen, R. J. Burge, J. Crandall,
B. Chaparro, R. Ni, and S. Cao. Texting While Driving: Is
Speech-Based Text Entry Less Risky Than Handheld Text
Entry? Accident Analysis & Prevention, Vol. 72, 2014,
pp. 287–295. https://doi.org/10.1016/j.aap.2014.07.014.
62. Savage, S. W., D. D. Potter, and B. W. Tatler. Does Preoc-
cupation Impair Hazard Perception? A Simultaneous EEG
and Eye Tracking Study. Transportation Research Part F:
Traffic Psychology and Behaviour, Vol. 17, 2013, pp. 52–62.
https://doi.org/10.1016/j.trf.2012.10.002.
63. Shinar, D., N. Tractinsky, and R. Compton. Effects of
Practice, Age, and Task Demands, on Interference From a
Phone Task While Driving. Accident Analysis & Preven-
tion, Vol. 37, No. 2, 2005, pp. 315–326.
64. Smiley, A., T. Smahel, D. Donderi, J. Caird, S. Chisholm,
J. Lockhart, and E. Teteris. The Effects of Cellphone and
CD Use on Novice and Experienced Driver Performance.
Insurance Bureau of Canada, Toronto, Ontario, 2007.
65. Ranney, T., G. S. Watson, E. N. Mazzae, Y. E. Papelis, O.
Ahmad, and J. R. Wightman. Examination of the Distrac-
tion Effects of Wireless Phone Interfaces Using the National
Advanced Driving Simulator - Final Report on a Freeway
Study. National Highway Traffic Safety Administration,
East Liberty, OH, 2005.
66. Cooper, J. M., and D. L. Strayer. Effects of Simulator
Practice and Real-World Experience on Cell-Phone-
Related Driver Distraction. Human Factors, Vol. 50, No.
6, 2008, pp. 893–902.
67. Caird, J. K., S. M. Simmons, K. Wiley, K. A. Johnston,
and W. J. Horrey. Does Talking on a Cell Phone, With a
Passenger, or Dialing Affect Driving Performance? An
Updated Systematic Review and Meta-Analysis of Experi-
mental Studies. Human Factors, Vol. 60, No. 1, 2018,
pp. 101–133. https://doi.org/10.1177/0018720817748145.
68. Ishigami, Y., and R. M. Klein. Is a Hands-Free Phone
Safer Than a Handheld Phone? Journal of Safety Research,
Vol. 40, No. 2, 2009, pp. 157–164. https://doi.org/10.1016/j.
jsr.2009.02.006.
69. Aitken, C., M. Kerger, and N. Crofts. Drivers Who Use
Illicit Drugs: Behaviour and Perceived Risks. Drugs: Edu-
cation, Prevention and Policy, Vol. 7, No. 1, 2000,
pp. 39–50. https://doi.org/10.1080/dep.7.1.39.50.
70. Greenfield, T. K., and J. D. Rogers. Alcoholic Beverage
Choice, Risk Perception and Self-Reported Drunk Driv-
ing: Effects of Measurement on Risk Analysis. Addiction,
Vol. 94, No. 11, 1999, pp. 1735–1743. https://doi.org/10.
1046/j.1360-0443.1999.9411173510.x.
71. Kelly, E., S. Darke, and J. Ross. A Review of Drug Use
and Driving: Epidemiology, Impairment, Risk Factors and
Risk Perceptions. Drug and Alcohol Review, Vol. 23, No. 3,
2004, pp. 319–344.
72. Fell, J. C., and R. B. Voas. The Effectiveness of a 0.05
Blood Alcohol Concentration (BAC) Limit for Driving in
the United States. Addiction, Vol. 109, No. 6, 2014,
pp. 869–874.
73. Alvarez, L., R. Colonna, S. Kim, C. Chen, K. Chippure, J.
Grewal, C. Kimm, T. Randell, and V. Leung. Young and
Under the Influence: A Systematic Literature Review of the
Impact of Cannabis on the Driving Performance of Youth.
Accident Analysis & Prevention, Vol. 151, 2021, p. 105961.
74. Downey, L. A., R. King, K. Papafotiou, P. Swann, E.
Ogden, M. Boorman, and C. Stough. The Effects of Can-
nabis and Alcohol on Simulated Driving: Influences of
Dose and Experience. Accident Analysis and Prevention,
Vol. 50, 2013, pp. 879–886. https://doi.org/10.1016/j.aap.
2012.07.016.
75. Frick, U., J. Rehm, A. Knoll, M. Reifinger, and J. Has-
ford. Perception of Traffic Accident Risk and Decision to
Drive Under Light Alcohol Consumption - A Double-
Blind Experimental Study. Journal of Substance Abuse,
Vol. 11, No. 3, 2000, pp. 241–251. https://doi.org/10.
1016/S0899-3289(00)00024-9.
76. Sexton, B., R. Tunbridge, N. Brook-Carter, P. Jackson, K.
Wright, M. Stark, and K. Englehart. The Influence of Can-
nabis on Driving. TRL Report, Vol. 477, 2000, p. 106.
77. Alvarez, L., and R. Colonna. Characteristics of a Technol-
ogy-Based Intervention for Young Drivers’ Hazard Percep-
tion: A Delphi Study. Journal of Transport & Health, Vol.
19, 2020, p. 100951.
78. Horswill, M. S., A. Hill, L. Silapurem, and M. O. Watson.
A Thousand Years of Crash Experience in Three Hours:
An Online Hazard Perception Training Course for Drivers.
Accident Analysis & Prevention, Vol. 152, 2021, p. 105969.
79. Damm, L., C. Nachtergae
¨le, M. Meskali, and C. Berthe-
lon. The Evaluation of Traditional and Early Driver Train-
ing With Simulated Accident Scenarios. Human Factors,
Vol. 53, No. 4, 2011, pp. 323–337. https://doi.org/10.
1177/0018720811413765.
80. Isler, R. B., N. J. Starkey, and P. Sheppard. Effects of
Higher-Order Driving Skill Training on Young, Inexper-
ienced Drivers’ On-Road Driving Performance. Accident
Analysis and Prevention, Vol. 43, No. 5, 2011,
pp. 1818–1827. https://doi.org/10.1016/j.aap.2011.04.017.
81. Anderson, J. R. Acquisition of Cognitive Skill. Psychologi-
cal Review, Vol. 89, No. 4, 1982, pp. 369–406. https://doi.
org/10.1037/0033-295X.89.4.369.
82. Meir, A., A. Borowsky, and T. Oron-Gilad. Formation
and Evaluation of Act and Anticipate Hazard Perception
Training (AAHPT) Intervention for Young Novice Drivers.
Traffic Injury Prevention, Vol. 15, No. 2, 2014, pp. 172–180.
https://doi.org/10.1080/15389588.2013.802775.
83. Ivancic, K., and B. Hesketh. Learning From Errors in a
Driving Simulation: Effects on Driving Skill and Self-Con-
fidence. Ergonomics, Vol. 43, No. 12, 2000, pp. 1966–1984.
https://doi.org/10.1080/00140130050201427.
84. Wetton, M. A., A. Hill, and M. S. Horswill. Are What
Happens Next Exercises and Self-Generated Commentaries
Useful Additions to Hazard Perception Training for
Cao et al 23
Novice Drivers? Accident Analysis and Prevention, Vol. 54,
2013, pp. 57–66. https://doi.org/10.1016/j.aap.2013.02.013.
85. Paas, F., A. Renkl, and J. Sweller. Cognitive Load Theory
and Instructional Design: Recent Developments. Educa-
tional Psychologist, Vol. 38, No. 1, 2003, pp. 1–4.
86. Paas, F. G. W. C., and J. J. G. Van Merrie
¨nboer. Instruc-
tional Control of Cognitive Load in the Training of Com-
plex Cognitive Tasks. Educational Psychology Review, Vol.
6, No. 4, 1994, pp. 351–371. https://doi.org/10.1007/
BF02213420.
87. Chapman, P., G. Underwood, and K. Roberts. Visual
Search Patterns in Trained and Untrained Novice Drivers.
Transportation Research Part F: Traffic Psychology and
Behaviour, Vol. 5, No. 2, 2002, pp. 157–167. https://doi.
org/10.1016/S1369-8478(02)00014-1.
88. Pradhan, A. K., A. Pollatsek, M. Knodler, and D. L.
Fisher. Can Younger Drivers Be Trained to Scan for Infor-
mation That Will Reduce Their Risk in Roadway Traffic
Scenarios That Are Hard to Identify as Hazardous? Ergo-
nomics, Vol. 52, No. 6, 2009, pp. 657–673. https://doi.
org/10.1080/00140130802550232.
89. Petzoldt, T., T. Weiß, T. Franke, J. F. Krems, and M.
Bannert. Can Driver Education Be Improved by Computer
Based Training of Cognitive Skills? Accident Analysis and
Prevention, Vol. 50, 2013, pp. 1185–1192. https://doi.
org/10.1016/j.aap.2012.09.016.
90. Samuel, S., T. Zafian, A. Borowsky, M. R. E. Romoser,
and D. L. Fisher. Can Young Drivers Learn to Anticipate
Hidden Hazards: A Driving Simulator Study Simulator
Study. Proc., 7th International Driving Symposium on
Human Factors in Driver Assessment, Training, and Vehicle
Design, Bolton Landing, New York, 2013.
91. Kahana-Levy, N., S. Shavitzky-Golkin, A. Borowsky, and
E. Vakil. The Effects of Repetitive Presentation of Specific
Hazards on Eye Movements in Hazard Perception Train-
ing, of Experienced and Young-Inexperienced Drivers.
Accident Analysis and Prevention, Vol. 122, 2019,
pp. 255–267. https://doi.org/10.1016/j.aap.2018.09.033.
92. He, J., E. Becic, Y. C. Lee, and J. S. McCarley. Mind
Wandering Behind the Wheel: Performance and Oculomo-
tor Correlates. Human Factors, Vol. 53, No. 1, 2011,
pp. 13–21. https://doi.org/10.1177/0018720810391530.
93. Young, A. H., D. Crundall, and P. Chapman. Commentary
Driver Training: Effects of Commentary Exposure, Prac-
tice and Production on Hazard Perception and Eye Move-
ments. Accident Analysis and Prevention, Vol. 101, 2017,
pp. 1–10. https://doi.org/10.1016/j.aap.2017.01.007.
94. Horswill, M. S., M. Garth, A. Hill, and M. O. Watson.
The Effect of Performance Feedback on Drivers’ Hazard
Perception Ability and Self-Ratings. Accident Analysis and
Prevention, Vol. 101, 2017, pp. 135–142. https://doi.org/10.
1016/j.aap.2017.02.009.
95. Van Merrie
¨nboer, J. J. G. Training Complex Cognitive
Skills: A Four-Component Instructional Design Model for
Technical Training. Educational Technology, Englewood
Cliffs, NJ, 1997.
96. Pollatsek, A., V. Narayanaan, A. Pradhan, and D. L.
Fisher. Using Eye Movements to Evaluate a PC-Based
Risk Awareness and Perception Training Program on a
Driving Simulator. Human Factors, Vol. 48, No. 3,
2006, pp. 447–464. https://doi.org/10.1518/00187200677860
6787.
97. Taylor, T., K. Masserang, A. Pradhan, G. Divekar, S.
Samuel, J. Muttart, A. Pollastsek, and D. Fisher. Long-
Term Effects of Hazard Anticipation Training on Novice
Drivers Measured on the Open Road. Proc., 2011 Interna-
tional Driving Symposium on Human Factors in Driver
Assessment, Training, and Vehicle Design, Olympic Valley,
CA, 2011, pp. 187–194.
98. Zafian, T. M., S. Samuel, J. Coppola, E. G. O’Neill, M.
R. E. Romoser, and D. L. Fisher. On-Road Effectiveness
of a Tablet-Based Teen Driver Training Intervention. Pro-
ceedings of the Human Factors and Ergonomics Society,
Vol. 60, No. 1, 2016, pp. 1926–1930.
99. Zafian, T. M., R. Agrawal, and S. Samuel. On-Road
Assessment of Retention Effects on Hazard Anticipation
Training for Novice Drivers. Transportation Research
Board, Washington, D.C., 2017.
100. Unverricht, J., S. Samuel, and Y. Yamani. Latent Hazard
Anticipation in Young Drivers: Review and Meta-Analy-
sis of Training Studies. Transportation Research Record:
Journal of the Transportation Research Board, 2018. 2672:
11–19.
101. Thomas, F. D., S. Rilea, R. D. Blomberg, R. C. Peck,
and K. T. Korbelak. Evaluation of the Safety Benefits of
the Risk Awareness and Perception Training Program for
Novice Teen Drivers. Report No. DOT HS 812 235. Dun-
lap and Associates, Inc., Stamford, CT, 2016.
102. Wang, Y. B., W. Zhang, and G. Salvendy. A Comparative
Study of Two Hazard Handling Training Methods for
Novice Drivers. Traffic Injury Prevention, Vol. 11, No. 5,
2010, pp. 483–491. https://doi.org/10.1080/15389588.2010.
489242.
103. McDonald, C. C., A. H. Goodwin, A. K. Pradhan, M. R.
E. Romoser, and A. F. Williams. A Review of Hazard
Anticipation Training Programs for Young Drivers. Jour-
nal of Adolescent Health, Vol. 57, No. 1, 2015, pp. s15–s23.
https://doi.org/10.1016/j.jadohealth.2015.02.013.
104. Zhang, T., F. Hajiseyedjavadi, Y. Wang, S. Samuel, X.
Qu, and D. Fisher. Training Interventions Are Only Effec-
tive on Careful Drivers, not Careless Drivers. Transporta-
tion Research Part F: Traffic Psychology and Behaviour,
Vol. 58, 2018, pp. 693–707. https://doi.org/10.1016/j.trf.
2018.07.004.
105. Hamid, M. Effect of Total Awake Time on Drivers’ Perfor-
mance and Evaluation of Training Interventions to Mitigate
Effects of Total Awake Time on Drivers’ Performance.
Department of Mechanical and Industrial Engineering,
University of Massachusetts, Amherst, 2013.
106. Poulsen, A. A., M. S. Horswill, M. A. Wetton, A. Hill,
and S. M. Lim. A brief office-based hazard perception
intervention for drivers with ADHD symptoms. Austra-
lian & New Zealand Journal of Psychiatry, Vol. 44, No. 6,
2010, pp. 528–534.
107. Young, A. H., P. Chapman, and D. Crundall. Producing
a commentary slows concurrent hazard perception
responses. Journal of Experimental Psychology: Applied,
Vol. 20, No. 3, 2014, p. 285.
24 Transportation Research Record 00(0)
108. Cantwell, S. J. The Role of Hazard Perception in Speed
Choice. PhD thesis. The University of Waikato, Hamil-
ton, New Zealand, 2021.
109. Muttart, J. W. Identifying Hazard Mitigation Behaviors
That Lead to Differences in the Crash Risk Between Expe-
rienced and Novice Drivers. PhD thesis. University of Mas-
sachusetts, Amherst, 2020.
110. Dinakar, S., and J. Muttart. Driver Behavior in Left Turn
Across Path From Opposite Direction Crash and Near
Crash Events From SHRP2 Naturalistic Driving.SAE
Technical Paper 2019-01-0414. Society of Automotive
Engineers, Detroit, 2019.
111. Muttart, J. W., S. Dinakar, J. Suway, M. Kuzel, S. Gern-
hard, M. Rackers, T. Schafer, T. Vadnais, and J. Fischer.
Influence of Taillight Width on the Ability to Recognize
Closing Speed, Closing Distance, and Closing Versus
Separating. Proceedings of the Human Factors and Ergo-
nomics Society, Vol. 61, No. 1, 2017, pp. 1894–1898.
112. Muttart, J., S. Dinakar, and D. Edewaard. Drivers’
Responses to Lead Vehicles: Thresholds for Triggering an
Emergency Response, Age Differences, Crash Risks, and
Influence of Secondary Task Engagement. SAE Technical
Paper 2021-01-0898. Society of Automotive Engineers,
Detroit, 2021.
113. Samuel, S., A. Borowsky, S. Zilberstein, and D. L. Fisher.
Minimum Time to Situation Awareness in Scenarios
Involving Transfer of Control From an Automated Driv-
ing Suite. Transportation Research Record: Journal of the
Transportation Research Board, 2016. 2602: 115–120.
Cao et al 25
