Accident Analysis and Prevention 101 (2017) 8796

Contents lists available at ScienceDirect

Accident Analysis and Prevention

journal homepage: www.elsevier.com/locate/aap

The impact of walking while using a smartphone on pedestrians

awareness of roadside events
Ming-I Brandon Lin , Yu-Ping Huang
Department of Industrial and Information Management, National Cheng Kung University, No.1 University Road, Tainan City 701, Taiwan, ROC

a r t i c l e i n f o a b s t r a c t

Article history: Previous studies have shown that using a cell phone to talk or text while walking changes gait kinematics
Received 3 November 2016 and encourages risky street-crossing behaviors. However, less is known about how the motor-cognitive
Received in revised form 3 January 2017 interference imposed by smartphone tasks affects pedestrians situational awareness to environmental
Accepted 3 February 2017
targets relevant to pedestrian safety. This study systematically investigated the inuence of smartphone
use on detection of and responses to a variety of roadside events in a semi-virtual walking environment.
Keywords:
Twenty-four healthy participants completed six treadmill walking sessions while engaged in a concurrent
Pedestrian
picture-dragging, texting, or news-reading task. During distracted walking, they were required to simul-
Dual-task
Smartphone
taneously monitor the occurrence of road events for two different levels of event frequency. Performance
Attention measures for smartphone tasks and event responses, eye movements, and perceived workload and situ-
Roadside event ational awareness were compared across different dual-task conditions. The results revealed that during
Eye-tracking dual-task walking, the reading app was associated with a signicantly higher level of perceived workload,
and impaired awareness of the surrounding environment to a greater extent compared with the texting
or picture-dragging apps. Pedestrians took longer to visually detect the roadside events in the reading
and texting conditions than in the dragging condition. Differences in event response performances were
mainly dependent on their salient features but were also affected by the type of smartphone task. Texting
was found to make participants more reliant on their central vision to detect road events. Moreover, dif-
ferent gaze-scanning patterns were adopted by participants to better protect dual-task performance in
response to the changes in road-event frequency. The ndings of relationships between workload, dual-
task performances, and allocation strategies for visual attention further our understanding of pedestrian
behavior and safety. By knowing how attentional and motor demands involved in different smartphone
tasks affect pedestrians awareness to critical roadside events, effective awareness campaigns might be
devised to discourage smartphone use while walking.
2017 Elsevier Ltd. All rights reserved.

1. Introduction
Among the activities of daily living, walking is a basic and
common mode of transportation, especially when the movement
distance is within several blocks. However, pedestrians are more
likely to directly experience high impact forces than road users
in vehicles when a trafc accident occurs, and consequently,
pedestrians are more susceptible to severe injuries (Pucher and
Dijkstra, 2003). In the US, pedestrian fatalities as a proportion
of the total motor vehicle fatalities has steadily increased from
11% in 2005 to 15% in 2014, despite a substantial decrease in
Corresponding author.
E-mail addresses: brandonl@mail.ncku.edu.tw, mingilin@gmail.com
(M.-I.B. Lin).
total trafc fatalities over the past decade (National Highway
Trafc Safety Administration, 2016). Several risk factors that could
affect pedestrian injury severity or fatality have been cited in the
literature (Haleem et al., 2015; Jang et al., 2013), including envi-
ronmental characteristics (e.g., darkness/lack of illumination, high
trafc density), vehicle characteristics (e.g., large vehicles), and
pedestrian-related characteristics (e.g., very young or old age, cell
phone use, and other risky behaviors).
The Wireless Association suggested that in 2014, more than half
of the device connections in North America were smartphones,
and these accounted for 77% of the total wireless network traf-
c (The Wireless Association, 2015). With the advent of mobile
technology, smartphones allow users not only to make voice calls
and perform text messaging, but also to engage in social network-
ing, gaming, video streaming, and a variety of mobile applications
(apps). A recent survey of more than 2000 US adults by the

http://dx.doi.org/10.1016/j.aap.2017.02.005
0001-4575/ 2017 Elsevier Ltd. All rights reserved.
88 M.-I.B. Lin, Y.-P. Huang / Accident Analysis and Prevention 101 (2017) 8796
American Academy of Orthopaedic Surgeons (2015) reported that
nearly one-third of the respondents admitted to frequently using
a smartphone for non-speech activities while walking, whereas
85% stated they had seen others perform these behaviors. Further-
more, 26% of respondents had been involved in distracted-walking
incidents, ranging from bumping into something without injury,
falling, or experiencing sprains or fractures. Using data from the
US Consumer Product Safety Commission on injuries in hospital
emergency rooms from 2004 to 2010, Nasar and Troyer (2013)
found that the percentage of total pedestrian injuries related to
mobile phones steadily increased; in 2010, it exceeded the esti-
mate for drivers. However, unlike distracted driving (Collet et al.,
2010; Johannsdottir and Herdman, 2010; Kaber et al., 2012; Recarte
and Nunes, 2003; Strayer and Johnston, 2001), there is less pub-
lished research on the impact of using a smartphone while walking,
despite its imminent threat to the safety of pedestrians and other
shared-road users.
Several researchers using semi-immersive virtual environments
found that individuals tend to exhibit more risky street-crossing
behaviors (e.g., less likely to recognize and act on crossing oppor-
tunities, pay less attention to trafc, or more likely to look away
from the street environment) when either engaging in phone
conversation (Neider et al., 2010), listening to a mobile music
device (Neider et al., 2010), or texting (Schwebel et al., 2012) than
when undistracted. For example, Stavrinos et al. (2011) reported
that pedestrians conversing on a phone were less likely than
undistracted pedestrians to recognize crossing opportunities, even
though the former waited longer curbside. Schwebel et al. (2012)
also revealed that participants distracted by texting or music had
higher odds of being hit by vehicles in a simulated environment
than those who were undistracted. The performance decrements
observed in distracted pedestrians are in part due to competition
for limited attentional resources between street-crossing behaviors
and mobile phone activities (Wickens, 2008). In the primary-
secondary task paradigm, secondary task performance was found
to be associated with the amount of residual attention available and
has been used to assess mental workload (Parasuraman et al., 2008;
Young et al., 2015). The multi-dimensional nature of mental work-
loads also suggests that interferences as the result of dual-tasking
could occur at different information processing stages (e.g., detec-
tion, discrimination, or motor response) (Recarte and Nunes, 2003)
and selectively impair task performance. However, the extent to
which visual attention is affected by mobile phone use in distracted
walking is not clear due to the lack of detailed information regard-
ing oculomotor responses.
Recent ndings from the dual-task literature suggest that
walking while performing a task may alter neutral correlates of
executive function and dynamic postural control (Hamacher et al.,
2015; Lin and Lin, 2016). With a concurrent non-gait task, partic-
ipants change gait patterns, reduce low-limb joint variability, or
experience reduced ability to maintain path direction (Kao et al.,
2015; Lamberg and Muratori, 2012; Lin and Lin, 2016; Schabrun
et al., 2014). Meanwhile, it has been argued that some modications
of these joint kinematics and spatiotemporal gait characteristics
(e.g., wider stride width and greater margins of stability) are pro-
tective walking strategies adopted by pedestrians to compensate
for the increased stability challenges during dual-tasking (Kao et al.,
2015; Schabrun et al., 2014).
In addition to examining pedestrians behaviors and posture
controls during ambulation, it is crucial to understand how dis-
tracted walking inuences the ability of individuals to maintain
awareness of the trafc environment, so as to devise efcient
interventions and policies to improve pedestrian safety. Vision has
showed to play an important role in guiding anticipatory locomo-
tor adjustments to avoid potential perturbation (Higuchi, 2013).
Empirical research indicated that engaging in a mobile phone con-
versation can cause pedestrians to neglect out-of-place objects
(i.e., an UNSAFE sign, a boot, a cat, pieces of fake vomit, and a chalk
sketch of an ostrich) placed along their walking route (Nasar et al.,
2008). Lim et al. (2015) recently investigated the effects of tex-
ting on situational awareness during walking and found that the
reduced detection of visual cues depended on the nature of the
visual information provided. Nevertheless, some popular current
activities entice pedestrians to interact with their smartphones, and
may tax attentional resources in a manner different from texting
or conversing (Rogers, 2016). It remains unknown whether individ-
uals distracted by these smartphone tasks would selectively alter
their gaze behaviors to better perceive environmental cues that, if
unattended, could place the individual at risk of collision. This gap in
the current understanding is particularly pertinent given that indi-
viduals who tend to show distracted walking seem overcondent
of their multitasking abilities (American Academy of Orthopaedic
Surgeons, 2015).
The main purpose of this study was to systematically investi-
gate the effects of smartphone use on the situational awareness of
roadside events relevant to pedestrian safety. While walking along
a virtual path and performing either a picture-dragging, instant
messaging, or news reading task on a smartphone, young par-
ticipants were instructed to continuously monitor the onset of
designated roadside events and discriminate them accordingly.
Laboratory-based experiments were designed to allow measure-
ment of performances of both app tasks and event responses, gaze
patterns, and perceived task workload and situational awareness
across different levels of road event density. We hypothesized that
participants would experience unequal levels of task workload and
awareness of the trafc environment when engaging in different
smartphone tasks while walking. Furthermore, the inuence of
dual-task walking on the detectability of road events was expected
to be more profound for a smartphone task that demanded more
attentional resources or induced more cognitive-motor interfer-
ence. Further, response times to road events were predicted to
depend on the stimulus characteristics and to be further affected
by trafc density. Finally, we also hypothesized that participants
would adopt different strategies of attention allocation to better
preserve the desired task performances when the frequency of
occurrence of road-events increased.
2. Method
2.1. Participants
Twenty-four healthy young adults (12 males;
age = 23.5 1.1 years; height = 168.4 6.7 cm) were recruited
from local universities for the study. Individuals were excluded
if they had any known neurological or musculoskeletal abnor-
malities that would limit their ability to walk or text on the
touchscreen of a smartphone. This particular age group was cho-
sen as they experience the highest incidence of pedestrian injuries
(National Center for Injury Prevention and Control, 2014) and are
signicantly more likely to report frequent smartphone use while
crossing the roads than other age groups (Lennon et al., 2017). All
participants were right-handed, had normal or correct-to-normal
vision, and had possessed a smartphone for more than one year.
Before participation, they provided written informed consent for
procedures, which were approved by the Human Research Ethics
Committee of the National Cheng Kung University (#103-030-2).
2.2. Instruments
The virtual pedestrian walking environment consisted of a
surrounding display formed by two 42 liquid-crystal display
 M.-I.B. Lin, Y.-P. Huang / Accident Analysis and Prevention 101 (2017) 8796
Fig. 1. (a) Photograph of the simulated environment that can record smartphone
task performances, eye movements, and roadside-event responses simultaneously,
(b) Experimental set-up for the treadmill and the two-screen display with areas of
interests dened, and (c) The onset locations and movement patterns for six types
of roadside events on the display (1: for pedestrians approaching from the front; 2:
for pedestrians approaching from the back; 3/4: for green and red pedestrian trafc
lights; 5: for stop signs; 6: for cars crossing the front path). (For interpretation of the
references to color in this gure legend, the reader is referred to the web version of
this article.)
televisions (LC-42D69U, Sharp; 1920 1080 pixels; 60 Hz) angled
toward each other with the center of screens located 1.8 m above
the ground, 2 m in front of a treadmill (MX-T5, Matrix, Taiwan;
Fig. 1a). The center display area, enclosed by the lines of sight from
participants walking position on the treadmill to the centers of
each screen, subtended a horizontal visual angle of 30 (Fig. 1b). The
system allowed for the animation of a pedestrian path and road-
side events displayed during the testing sessions that resembled an
actual street scene perceived from a distance of 10 m.
Participants were instructed to monitor different locations
within the display space for a variety of roadside events while walk-
ing on the treadmill, and to respond with designated hand gestures
(waves). Specically, six types of visual stimuli (simple geometric
shapes) that represented road events critical to pedestrian safety
were included in the animation of the walking environment dis-
played on the screen: 1) a pedestrian approaching from the front,
recognized by waving right or left based on the event location;
2) a pedestrian approaching from the back, recognized by waving
right or left; 3) a red pedestrian trafc light, recognized by waving
upward; 4) a green pedestrian trafc light, recognized by waving
downward; 5) a stop sign, recognized by waving right or left; and
89
6) a car crossing the front path, recognized by waving upward. The
movement patterns of individual road events on the two-screen
display are depicted in Fig. 1c.
A mobile eye-tracking system (Eye Tracking Glasses, Senso-
Motoric Instruments, Germany) was used to record participants
ocular movements at 30 Hz. Information regarding eye xations
and associated transition matrices for areas of interest (AOIs)
were automatically calculated by a proprietary analysis software
(BeGaze v3.4, SensoMotoric Instruments, Germany) dedicated
for the eye-tracking data recorded by our system. The gesture
responses to road events were captured by a synchronized Sony
HDR-CX150 digital camcorder for 1920 1080 interlaced video at
a 59.94-Hz eld rate. The timing information of initiations of road
events and their following hand responses were then manually
coded in the BeGaze environment. Customized scripts written in
MATLAB R2014a (The Mathworks, Inc. Natick, MA, USA) were used
to calculate the measures for task performances and attentional
allocation strategies (described later in the data analysis section).
2.3. Experimental design
The study employed a 3 2 factorial design with both the
effect of smartphone tasks (App Type: picture drag and drop game,
instant messaging, news/blog reading) and the effect of foot trafc
(Event Frequency: high and low frequency of occurrence of road-
side events) as within-subject factors. Participants had to divide
their attention between the two-screen display channels at the
front and a smartphone while walking.
For the smartphone tasks, three mobile applications were pro-
grammed and executed on the same smartphone (iPhone 4S, iOS
5.1, Apple, Cupertino, CA, USA). The game screen of the drag & drop
app (Dragging) displayed a picture (75 75 pixels) in the center,
surrounded by eight different potentially matching images. In each
3-s round, the participant had to drag the central picture to the
matching image. The instant messaging app (Texting) consisted
of a user interface identical to the native iOS Message app with
the autocorrect and autoll functions turned off. This customized
app prompted an inquiry every 10 s and waited for the participant
to type a simple text response back. The inquiries were phrased
in Chinese and carefully designed to be answerable using simple
English vocabulary (selected from 1200 basic words required for
junior-high school levels in Taiwan). The news reading task was
also performed with a dedicated app (Reading) used to simulate
how people read news or weblogs from the Internet. Every 60 s, the
participant had to read an excerpted article in Chinese from online
news in the rst screen page and answer two related inquires on
the next screen (one was a Yes/No question, another was a multiple
choice question).
Regarding the visual task for the roadside events, during a test-
ing session, at any given time, only one type of road event could
occur. Events persisted for 10 s, during which they could poten-
tially be detected and recognized. The order of occurrence and
onset location (either the right or left side of the display) of the
visual stimuli were randomized in each testing condition. In the
high event-frequency condition, thirty roadside events with six
instances per type were assigned in the 5-min period. In the low
event-frequency condition, only fteen roadside events with three
instances per type were presented with the constraint that every
1-min period contained exactly three stimuli.
2.4. Procedures
After reporting to the laboratory, the participant was briefed
regarding the details of the experiment and required responses
for the smartphone and road-event tasks. Based on the method
proposed by Dingwell and Marin (2006), the individual was rst
90 M.-I.B. Lin, Y.-P. Huang / Accident Analysis and Prevention 101 (2017) 8796
required to walk on a treadmill several times without performing
any additional activity to determine their preferred walking speed.
The preferred walking speed would be further reduced by 33% as his
or her constant speed setting on the treadmill for all following test-
ing conditions. This protocol was chosen to better accommodate
the additional visual and motor demands experienced by partic-
ipants during distracted walking (Lamberg and Muratori, 2012).
Each participant then wore the calibrated eye-tracking system and
had 30 min to practice the individual tasks while seated in a chair
(smartphone tasks) or while standing still on the treadmill (road-
event tasks). Participants were instructed to comfortably hold the
smartphone in portrait mode with the right hand and make task
responses using the thumb of the same hand. The left hand was
reserved for making gestures associated with the detection of road
events when necessary. Finally, participants performed two 3-
min warm-up trials for each smartphone task while concurrently
monitoring the occurrences of roadside events in the virtual pedes-
trian walking environment. During the familiarization session, the
road-event frequency was set to four events per minute and the
smartphone apps utilized content not shown in the following for-
mal testing sessions.
After a 10-min break, each participant completed six formal
5-min testing sessions while walking on the treadmill. Sessions
consisted of different combinations of smartphone and road-event
tasks. Participants were instructed to perform the smartphone task
as accurately as possible while concurrently monitoring the virtual
walking environment to capture the potential occurrences of road-
side events. The participant had 10 min to relax after the session
concluded. The order of presentation of conditions was random-
ized across participants. During the intermission, task workload
and situational awareness of the visual walking environment expe-
rienced by the participant during the preceding testing session
were subjectively assessed by the NASA Task Load Index (NATA-
TLX) (Hart and Staveland, 1988) and the Situational Awareness
Rating Technique (SART) (Taylor, 1990), respectively. The NASA-
TLX is a widely used multi-dimensional assessment tool that allows
users to quantify the perceived workload with six subscale rat-
ings (i.e., mental, physical, and temporal demands, performance,
effort, and frustration). The NASA-TLX uses weighting schemes to
account for individual differences and provides a single overall
workload score ranging from low to high on a scale from 0
to 100. Dissociation between subjective workload ratings and task
performances has been reported in the literature, along with pos-
sible explanations (Kaber et al., 2012; Vidulich and Tsang, 2012;
Young et al., 2015). The SART also adopts a similar post-trial self-
rating technique to assess SA using 10 generic items, which can
be grouped into three dimensions (i.e., demand on attentional
resources, supply of attentional resources, and understanding of the
situation provided). An overall SART score was calculated as SA (cal-
culated) = Understanding (Demand Supply) with a higher value
indicating better awareness. This subjective SA measure has been
found to be associated with subjective performance and workload
measures (Endsley, 1998; Endsley and Jones, 2011; Selcon et al.,
1991).
2.5. Data analysis
2.5.1. Measures of smartphone task performance
The total percentage of response errors and average reaction
times were calculated from individual log les generated by the
mobile apps to evaluate smartphone task performance across con-
ditions.
2.5.2. Measures of visual task performance
The detection time of a particular roadside event was dened as
the time elapsed between the event onset time and the time when
it received its rst following gaze xation. The time delay in event
recognition was dened as the time elapsed between the pedes-
trians rst xation on the road event and the time at which the
correct hand gesture was initiated. Error rates for the road event
task for a particular session were computed by dividing the num-
ber of events without correct gesture responses by the number of
events presented.
2.5.3. Measures of attentional allocation strategies
The percentage usage of central vision during target detection
within a testing session was estimated as the number of road
events receiving gaze xation divided by the total number of road
events with gesture responses. The distribution of visual attention
in the virtual walking environment was also examined by compar-
ing the percentage dwell time across four different AOIs: the area
encompassing the smartphone and gesture reaction regions, and
the central, upper far, and lower far areas of the two-screen display
(Fig. 1b). Specically, the central area was the middle section, con-
sisting of each half of the right and left screens demarcated by the
central vertical line. The rest of the outer display channels was fur-
ther separated as the far upper and the far lower areas with respect
to the center horizontal line.
2.5.4. Statistical analysis
All outcome variables were compared among different con-
ditions of distracted walking using repeated-measures analysis
of variance (ANOVA) in SAS via linear mixed model (v9.4, SAS
Institute, Cary, NC, USA). Box-Cox power transformation was rst
performed for the measures for which data substantially deviated
from normality. The most parsimonious variancecovariance struc-
ture of the response measures was then selected based on the
Bayesian information criterion (Oberfeld and Franke, 2013). Differ-
ences in performance metrics of the smartphone tasks, error rates
for road-event responses, and percentage usages of central vision
among conditions were analyzed separately via 3 App Type 2
Event Frequency ANOVA models, as were perceived ratings of task
workload and SA of the walking environment. The effects of road
event type on event-detection times and event-recognition delays
were also evaluated by applying separate 3 App Type 2 Event Fre-
quency 6 Road Event Type ANOVAs. Finally, strategies of visual
attention allocation among conditions were investigated based on
percentage dwell times of different locations in the virtual walk-
ing environment using a 3 App Type 2 Event Frequency 4 AOI
ANOVA. All the aforementioned factors were modeled as xed
effects in individual ANOVAs, with participant included as an addi-
tional random effect. Potential inuences of gender, walking speed,
and iOS system familiarity were further controlled by treating them
as covariates in the analyses. Post-hoc tests between different lev-
els of a xed factor were performed using Tukeys method if the
results of omnibus F-tests were considered statistically signicant
(p < 0.05).
3. Results
3.1. Situational awareness and task workload
As expected, when walking participants were required to con-
duct app tasks on the smartphone and concurrently detect road
events on the screens, the type of smartphone task had a signi-
cant inuence on their SA levels (F2,115 = 22.66, p < 0.001). Walking
while using the dragging-game app led to the highest SART score
(24.2 5.7; Dragging vs Texting, t115 = 2.96, p = 0.010; Dragging vs
Reading, t115 = 6.72, p < 0.001), followed by Texting (21.9 5.4; Tex-
ting vs Reading, t115 = 3.76, p < 0.001), and Reading (19.2 5.5).
Moreover, participants also exhibited decreased situational aware-
 M.-I.B. Lin, Y.-P. Huang / Accident Analysis and Prevention 101 (2017) 8796
Fig. 2. Error rates for road-event responses as a function of smartphone task and
event occurrence frequency.
Under the same condition of event occurrence frequency, a signicant dif-
ference (p < 0.05) between the target smartphone task condition and the
dragging/texting/reading task condition was denoted by the symbol D/T/R, respec-
tively.
ness when the frequency of road events was high (High vs Low:
19.8 5.4 vs 23.8 5.6, F2,115 = 36.46, p < 0.001).
Similarly, there were signicant App Type (F2,115 = 34.66,
p < 0.001) and Event Frequency main effects (F2,115 = 9.33, p = 0.003)
on the measured NASA-TLX ratings. Participants experienced
the lowest workload from the dragging app (44.2 17.3; Drag-
ging vs Texting, t115 = 4.50, p < 0.001; Dragging vs Reading,
t115 = 8.32, p < 0.001) and the highest workload from the reading
app (65.1 12.7), with scores from the texting app (55.8 17.7;
Texting vs Reading, t115 = 3.82, p < 0.001) in between. On the
other hand, more frequent roadside events in the detection task
resulted in a higher TLX score (58.2 18.3) than in the low-
frequency condition (51.8 17.5). The frequency inuence of
road-event occurrences on the subjective workload ratings was
particularly obvious in the texting conditions (Low: 49.4 16.1;
High: 62.1 17.2; t115 = 3.47, p < 0.001). There was no signicant
App Frequency interaction for either SART or TLX scores.
3.2. Task performance
Analysis of the performance measures for the smartphone tasks
revealed a signicant App Type main effect on RTs (F2,115 = 15612.7,
p < 0.001) and error rates (F2,115 = 66.19, p < 0.001). Participants
responded faster and made fewer errors when playing the drag-
ging game (reaction time: 831 99 ms; error rate: 1.3% 1.5) than
when texting on the phone (reaction time: 6447 750 ms; error
rate: 9.1% 7.2); the latter performance was superior to that noted
with the reading app (reaction time: 47095 7103 ms; error rate:
18.3% 14.0). All pairwise comparisons had p-values less than
0.001. No signicant differences were found between the low- and
high-frequency conditions for RTs and error rates.
Regarding visual-task performance, our participants exhibited a
signicantly lower error rate when dual-tasking with the dragging
app than the other two apps (F2,115 = 9.46, p < 0.001; Dragging vs
Texting, t115 = 4.24, p < 0.001; Dragging vs Reading, t115 = 2.96,
p = 0.010; Fig. 2). However, for a given smartphone app, the event
occurrence frequency in the virtual walking environment did not
signicantly change overall response errors. The ANOVA of partic-
ipants detection times of road events also indicated a signicant
main App Type effect (F2,3829 = 47.30, p < 0.001). Concurrently per-
forming the texting or reading task on the smartphone led to a
91
longer event-detection time than that associated with the drag-
ging task (Dragging vs Texting, t3829 = 8.01, p < 0.001; Dragging vs
Reading, t3829 = 8.79, p < 0.001), whereas no difference was found
between the texting and reading tasks. Similarly, the subsequent
time delay in event recognition was also signicantly affected by
the type of smartphone task (F2,3829 = 37.34, p < 0.001). As shown
in Fig. 3, the average times required to initiate the correct hand
gestures in response to road events were longer in the texting
condition than in either reading or dragging conditions (Dragging
vs Texting, t3829 = 7.56, p < 0.001; Texting vs Reading, t3829 = 7.41,
p < 0.001). Nevertheless, event frequency minimally inuenced the
time needed for event detection or recognition.
It is worth noting that the type of roadside event signi-
cant affected detection time (F5,3829 = 439.10, p < 0.001) and the
time delay for event recognition (F5,3829 = 24.41, p < 0.001). Fig. 3a
presents the mean detection time for event types across different
app tasks. Notably, participants took longer to visually detect small
events appearing on the upper far regions of the display (e.g., the
stop sign and red trafc light) than to detect a target that subtended
a large visual angle (e.g., the car crossing the path). Accordingly,
events that occurred within the near peripheral visual-eld (e.g.,
the pedestrian approaching from the rear and green trafc light)
had the shortest mean detection times during distracted walk-
ing. Moreover, our data show that the effect of App Type on the
time delay in event recognition was signicantly different across
roadside event types (F10,3829 = 2.94, p = 0.001). In the texting con-
dition, most types of roadside events (except the green trafc light)
required more than 400 extra milliseconds for the participant to
issue the correct hand gesture after initial detection (Fig. 3b). In
contrast, in the reading and dragging conditions, only stop sign and
red light events led to substantial recognition delays.
3.3. Oculomotor responses
Fig. 4 presents the percentage usage of central vision in tar-
get detection as a function of event frequency for each app
condition. Participants apparently tailored their visual-scanning
strategies to detect road events in response to dissimilar atten-
tional and motor demands imposed by the concurrent app tasks
(F2,115 = 17.56, p < 0.001). When the event frequency was low, the
use of central vision in the texting condition was 12% greater than
in the dragging or reading conditions (low-frequency condition:
Dragging vs Texting, t115 = 3.02, p = 0.009; Texting vs Reading,
t115 = 2.70, p = 0.022). The discrepancy further increased to over
19% in the high-frequency condition (high-frequency condition:
Dragging vs Texting, t115 = 4.46, p < 0.001; Texting vs Reading,
t115 = 4.31, p < 0.001; Fig. 4). However, the effect of Event Frequency
was not signicant in the ANOVA.
Averaged over all six testing conditions, 93% of gaze time was
spent xating on the smartphone, with the remaining 7% allocated
to different areas of the display. The percentage dwell time was
substantially greater in the central area of the display than in the
other two outer areas (F3,529 = 5418.23, p < 0.001; Fig. 5). The change
in the frequency of occurrence of roadside events induced differ-
ent patterns of attention distribution within the virtual walking
environment, as indicated by a signicant AOI Frequency inter-
action (F3,529 = 7.61, p < 0.001). For instance, as shown in Fig. 5,
the percentage dwell times of the upper far areas were larger in
the high-frequency condition than the low-frequency condition
(Low vs High, t529 = 2.34, p = 0.020), whereas the percentage dwell
times of the central area decreased when the event frequency
increased (Low vs High, t529 = 4.15, p < 0.001). This suggests that
in the high-frequency condition, participants may have used top-
down expectancy mechanisms to bias overt attention toward the
areas in which most road events rst occurred (high-frequency
condition: upper far area vs lower far area, t529 = 2.21, p = 0.028).
92 M.-I.B. Lin, Y.-P. Huang / Accident Analysis and Prevention 101 (2017) 8796

Fig. 3. (a) Road-event detection time and (b) time delay in event recognition as a function of smartphone task and roadside-event type.
Under the same roadside-event type, a signicant difference (p < 0.05) between the target smartphone task condition and the dragging/texting/reading task condition was
denoted by the symbol D/T/R, respectively.

Meanwhile, participants relied more heavily on covert attention to
monitor the central area of the display. The main effect of App Type
also reached signicance (F2,529 = 4.34, p = 0.014). This was primar-
ily due to a higher percentage dwell time observed in the central
area during the Reading condition than during the Texting or Drag-
ging condition (Texting vs Reading, t529 = 2.23, p = 0.026; Dragging
vs Reading, t529 = 3.40, p = 0.002).
4. Discussion
This study investigated the impact of concurrent smartphone
tasks on road-event detectability while walking against different
trafc frequencies. The ndings conrmed our hypotheses that
participants ability to perceive and then distinguish changes in
the forward view was affected by the level of distraction caused
by the smartphone apps. Longer detection times and recognition
delays were observed for conditions in which the smartphone
task either depleted substantial attentional resources or induced
severe motor-cognitive interference. The exact magnitude of the
aforementioned performance decrement depended on the intrin-
sic properties of the road events. Attention to small-size roadside
events and those occurring outside participants eld of view was
generally impaired to a considerable extent by the additional work-
load imposed by the concurrent texting or reading tasks. More
importantly, analyses of gaze behaviors further revealed that par-
ticipants tended to deploy their attention allocation strategies to
 M.-I.B. Lin, Y.-P. Huang / Accident Analysis and Prevention 101 (2017) 8796
Fig. 4. Percentage usage of central vision during road-event detection as a function of smartphone task and event occurrence frequency.
Under the same condition of event occurrence frequency, a signicant difference (p < 0.05) between the target smartphone task condition and the dragging/texting/reading
task condition was denoted by the symbol D/T/R, respectively.
balance dual-task performance in response to the manipulation of
road-event frequency in the simulated-walking environment.
4.1. Situational awareness and task workload
Since developing and maintaining SA in the current study
required frequent updates to participants visuospatial sketchpad,
the presence of a concurrent smartphone task reduced the amount
of attention spent on the trafc environment (Fig. 5). The results
are in agreement with ndings from previous studies in which
individuals talking (Stavrinos et al., 2011) or using mobile Internet
on a phone (Byington and Schwebel, 2013) were found to check
oncoming trafc less frequently before crossing a street compared
to undistracted individuals. Furthermore, our data also suggest
that participants tended to perceive the greatest decrement in SA
when using an app that imposed the highest level of task work-
load. This observation could be at least partly explained using the
multiple resource theory proposed by Wickens (2008). First, the
reading and texting tasks require participants to either gaze upon
the displayed article or select characters from the miniaturized vir-
tual keyboard. Therefore, spare visual attention available for the
road-event detection task is expected to be more limited when per-
forming those two tasks than playing the picture-dragging game.
Second, the diverse cognitive components of the reading task (e.g.,
maintenance of information in working memory, comprehension,
reasoning, decision-making) could make participants more suscep-
tible to cognitive tunneling and substantially affect their abilities
to relocate attentional resources dynamically (Vidulich and Tsang,
2012).
One unexpected nding of our study is the negligible effect of
road-event frequency on performance outcomes of the smartphone
93
and event detection tasks. Although the mechanisms underpinning
these phenomena are difcult to determine, it might imply that our
participants inclined to cope actively with changes in the frequency
of road-event occurrence, so as to protect both task performances
(Hockey, 1997). Specically, additional mental efforts were devoted
in the high-frequency condition and consequently led to a concomi-
tant rise in the NATA-TLX workload ratings (Young et al., 2015). As
suggested by Kahneman (1973), an individuals evaluation of task
performances, along with enduring dispositions and momentary
intentions, are three important factors governing allocation policies
during multitasking.
4.2. Detection time and delay in hand response
The signicant difference in detection times across the three
apps suggests that the amount of multitasking interference while
walking involves cognitive processes that vary along multiple
dimensions (Wickens, 2008). Previous research revealed that indi-
viduals holding a phone conversation tend to take more time to
initiate crossing behaviors. This was attributed to possible poor
performance in encoding visual information (approaching cars)
into working memory (Neider et al., 2010). In the current study,
participants also spent more time detecting cars when reading or
texting than when playing the picture-dragging game (Fig. 3a).
In the texting and reading conditions, participants tended to x-
ate their gazes on the displayed sentences during information
acquisition/comprehension processes. This presumably increased
visual-cognitive interference, with much attention diverted away
from the surroundings. Consequently, we found longer detection
times for road events, degraded SA, and increased task workload
during these conditions. Elucidating the nature of interference
94 M.-I.B. Lin, Y.-P. Huang / Accident Analysis and Prevention 101 (2017) 8796

Fig. 5. Percentage dwell time as a function of areas of interests and event occurrence frequency.
An asterisk (*) denoted a signicant difference (p < 0.05) between the low and high frequency condition with respect to the same area of interests. Under the same condition
of event occurrence frequency, a signicant difference (p < 0.05) between the target area of interests and the smartphone/upper far/lower far/central area was denoted by
the symbol S/U/L/C, respectively.

in visual-information processing provides additional insights that
extend the current understanding of distracted walking beyond
phone conversations (Neider et al., 2010; Stavrinos et al., 2011) or
text entry (Byington and Schwebel, 2013; Schwebel et al., 2012).
4.3. Strategies for allocating visual attention
Fundamental research on visual attention has revealed several
determinants of perception or search performance by manipulat-
ing features of visual objects (Goldstein, 2014). For example, visual
saliency, which denotes that attention is generally captured by
salient objects in the visual eld, is known to affect task perfor-
mance due to bottom-up inuences related to various physical
properties of the stimulus, such as color, size, brightness, orien-
tation, onset, motion, and contrast (Carmi and Itti, 2006; Wolfe
and Horowitz, 2004). Moreover, target detectability is known to
decrease when visual information is located further toward the
peripheral visual eld or in separate regions of a display, when
searching involves long scans (Johnston et al., 2015; Stelzer and
Wickens, 2006). However, it remains unclear to what extent visual
features inuence target-detection performance in ecologically
valid experiments that are valid and directly relevant to pedestrian
safety. To the best of our knowledge, ours is the rst work to exam-
ine the dual-task effects of smartphone use on pedestrians SA with
respect to a variety of common road events with different visual fea-
tures. Consistent with ndings from studies in aviation and driving
(Johnston et al., 2015; Shahar et al., 2010; Steelman et al., 2013),
we found that small road events that emerged in regions typically
outside the participants visual eld during their smartphone inter-
actions were generally detected more slowly than salient events
that occurred within the near visual eld (Fig. 3a). Therefore, in
the present research, the differences in detection times across road
events provides direct evidence of how overt visual attention is
inuenced by bottom-up mechanisms during distracted walking.
The analysis of ocular responses across conditions also demon-
strated how top-down attentional controls driven by observers
expectancies and experiences inuenced gaze scanning patterns
in response to task demands (Geng and Behrmann, 2005; Steelman
et al., 2013). Our data suggest that when the frequency of occur-
rence of visual targets changed, smartphone users were able to
maintain task performance by adopting different strategies for
attention allocation. Specically, in the high event-rate condition,
more frequent sampling and updating of visual information in the
upper far regions of the display was observed, which was accompa-
nied by reduced percentage dwell times in the central area (Fig. 5).
These ndings may suggest top-down effects that bias attention
to better facilitate the detection of onsets, albeit with additional
efforts required as a consequence of the resulting long scans. Using
a dual task consisting of a central ight task and peripheral digi-
tal detection task, Steelman et al. (2013) examined the interaction
between bottom-up and top-down control of visual processing.
They found that top-down processes could mitigate costs of eccen-
tricity for targets located in high-expectancy regions versus those
in low-expectancy regions. Consistent with their ndings, we found
that more overt attention was allocated to the areas in which
most events rst occurred. Consequently, detection times of our
least-salient event (stop sign) decreased by 610 ms when event
frequency increased.
Another interesting gaze pattern observed was that to main-
tain awareness of the walking environment, participants modied
their attentional monitoring behaviors in response to dissimilar
demands imposed by different smartphone tasks. For example,
 M.-I.B. Lin, Y.-P. Huang / Accident Analysis and Prevention 101 (2017) 8796
while text messaging, participants relied more heavily on cen-
tral vision to detect road events than when performing the other
two app tasks (Fig. 4). Previous research into distracted walking
revealed reduced neck range of motion in the sagittal plane during
texting versus reading on a phone (Schabrun et al., 2014). Moving
the head in-phase with the thorax apparently compromises pedes-
trians ability to monitor changes in the surroundings using the
upper visual eld, which could cause substantial and frequent head
movements to overtly shift attention between the smartphone and
display. This unique task demand may partially account for the
higher percentage use of central vision in the texting condition than
in the dragging or reading condition. Meanwhile, the spatial gaze
concentration elicited by the added task workload while texting
might also contribute to this nding as a consequence of reducing
the functional eld of view (Atchley and Dressel, 2004; Recarte and
Nunes, 2003).
4.4. Limitations
Several limitations of this experimental study have to be taken
into account before generalizing the ndings to everyday living.
First, it was conducted in a laboratory using a semi-immersive vir-
tual pedestrian environment to precisely quantify participants SA
to roadside events while they were distracted by smartphone use.
Simulated behaviors may not completely represent pedestrians
responses in the real world where failure to detect road events
could seriously endanger their safety. Secondly, walking on a tread-
mill could restrict participants adaptive strategies available for
postural control during dual-task walking. Further evidence from
research conducted in naturalistic walking environments is needed
to verify our ndings. Additionally, aural cues are important in
judging the safety of pedestrian environments (Barton et al., 2013);
these were not considered in our visual environment. Therefore,
it is unclear how much the distraction elicited by the concur-
rent smartphone tasks disrupted auditory information processing
and hence hindered road event detection. We also did not include
baseline (single-task) conditions for road-event detection in the
experimental design. Consequently, how participants respond to
environment cues in the absence of smartphone distractions can-
not be estimated. Moreover, the current ndings were based on
a limited number of participants. The small sample size could
impede the ability of statistical tests to detect signicant differ-
ences across testing conditions. Therefore, we should be cautious
when generalizing the conclusions to different distracted walking
scenarios. Finally, during the experiment, we implicitly reinforced
task engagement by instructing participants to accurately perform
the smartphone tasks and informed them of their task perfor-
mance. The negligible effect of event frequency on the error rates
for smartphone and detection tasks conrmed the success of this
strategy. Nevertheless, given the multifaceted nature of human
information-processing and strategic management, impairment in
pedestrians SA would be likely more severe when distracted by
smartphone activities that are inherently more engaging and that
engender attentional tunneling, such as highly addictive gaming
apps (Rogers, 2016).
4.5. Conclusions
The primary contribution of the present study is our empirical
evidence that the decrement in pedestrians behavior performances
to detect roadside events while walking is substantially inuenced
by the amount and type of attentional resources demanded by
the concurrent smartphone tasks. Texting or reading news from
social networking sites on a smartphone while walking should be
avoided because it has a high perceived task workload and impairs
the pedestrians awareness of the surrounding environment to a
95
considerable extent. Our ndings have several implications for
improving pedestrian safety. From a research perspective, this is
the rst study to examine the adverse effects of distracted walking
with respect to information-processing performance of road-event
detection and recognition. Future research should distinguish how
different types and amounts of attentional resources demanded
by smartphone use affect an individuals abilities to detect, recog-
nize, and respond to environment cues critical to pedestrian safety.
New experimental paradigms for research in distracted walking are
needed to better time-lock various stages of interactions (e.g., ini-
tiate, monitor, and respond to). From a prevention perspective, the
ndings demonstrate that performing a task with considerable cog-
nitive demands (e.g., reading news posted on social network sites)
could impair pedestrians awareness to an extent similar to (if not
greater than) texting. A recent study revealed that pedestrians with
favorable attitudes toward smartphone use while crossing a road
were more likely to actually do that (Lennon et al., 2017). There-
fore, public educational campaigns should exploit these ndings to
strengthen pedestrians negative attitudes toward performing cog-
nitively distracted smartphone activities. Last, ndings regarding
the specicity of visual attention allocation strategies associated
with smartphone interactions might provide valuable insights for
designers to devise an intelligent technology that could better warn
distracted pedestrians about imminent dangers.
Acknowledgments
We appreciated the help in data collection from graduate stu-
dents, Xin-Min Ke, Wen-Xian Li, Wan-Chen Lin, and Chi-Heng
Chung. This work was partially funded by the Ministry of Science
and Technology of Taiwan (MOST 102-2221-E-006-181-).
References
American Academy of Orthopaedic Surgeons, 2015. Distracted Walking Study,
Retrieved from http://orthoinfo.aaos.org/topic.cfm?topic=A00748.
Atchley, P., Dressel, J., 2004. Conversation limits the functional eld of view. Hum.
Factors 46 (4), 664673.
Barton, B.K., Lew, R., Kovesdi, C., Cottrell, N.D., Ulrich, T., 2013. Developmental
differences in auditory detection and localization of approaching vehicles.
Accid. Anal. Prev. 53, 18, http://dx.doi.org/10.1016/j.aap.2012.12.040.
Byington, K.W., Schwebel, D.C., 2013. Effects of mobile Internet use on college
student pedestrian injury risk. Accid. Anal. Prev. 51, 7883, http://dx.doi.org/
10.1016/j.aap.2012.11.001.
Carmi, R., Itti, L., 2006. Visual causes versus correlates of attentional selection in
dynamic scenes. Vision Res. 46 (26), 43334345, http://dx.doi.org/10.1016/j.
visres.2006.08.019.
Collet, C., Guillot, A., Petit, C., 2010. Phoning while driving I: a review of
epidemiological, psychological, behavioural and physiological studies.
Ergonomics 53 (5), 589601, http://dx.doi.org/10.1080/00140131003672023.
Dingwell, J.B., Marin, L.C., 2006. Kinematic variability and local dynamic stability of
upper body motions when walking at different speeds. J. Biomech. 39 (3),
444452, http://dx.doi.org/10.1016/j.jbiomech.2004.12.014.
Endsley, M.R., Jones, D.G., 2011. Designing for Situation Awareness: An Approach
to User-Centered Design, 2nd ed. CRC Press, Boca Raton, FL.
Endsley, M.R., 1998. A comparative analysis of SAGAT and SART for evaluations of
situation awareness. In: Paper Presented at the Human Factors and
Ergonomics Society 42nd Annual Meeting, Chicago, Illinois.
Geng, J.J., Behrmann, M., 2005. Spatial probability as an attentional cue in visual
search. Percept. Psychophys. 67 (7), 12521268.
Goldstein, E.B., 2014. Sensation and Perception, 9th ed. Wadsworth Cenage
Learning, New York.
Haleem, K., Alluri, P., Gan, A., 2015. Analyzing pedestrian crash injury severity at
signalized and non-signalized locations. Accid. Anal. Prev. 81, 1423, http://dx.
doi.org/10.1016/j.aap.2015.04.025.
Hamacher, D., Herold, F., Wiegel, P., Hamacher, D., Schega, L., 2015. Brain activity
during walking: a systematic review. Neurosci. Biobehav. Rev. 57, 310327,
http://dx.doi.org/10.1016/j.neubiorev.2015.08.002.
Hart, S.G., Staveland, L.E., 1988. Development of NASA-TLX (task load index):
results of empirical and theoretical research. In: Peter, A.H., Najmedin, M.
(Eds.), Advances in Psychology, vol. 52. North-Holland, pp. 139183.
Higuchi, T., 2013. Visuomotor control of human adaptive locomotion:
understanding the anticipatory nature. Front. Psychol. 4, 277, http://dx.doi.org/
10.3389/fpsyg.2013.00277.
96 M.-I.B. Lin, Y.-P. Huang / Accident Analysis and Prevention 101 (2017) 8796
Hockey, G.R., 1997. Compensatory control in the regulation of human performance
under stress and high workload; a cognitive-energetical framework. Biol.
Psychol. 45 (13), 7393.
Jang, K., Park, S.H., Kang, S., Song, K.H., Kang, S., Chung, S., 2013. Evaluation of
pedestrian safety pedestrian crash hot spots and risk factors for injury severity.
Transp. Res. Rec., 104116, http://dx.doi.org/10.3141/2393-12.
Johannsdottir, K.R., Herdman, C.M., 2010. The role of working memory in
supporting drivers situation awareness for surrounding trafc. Hum. Factors
52 (6), 663673, http://dx.doi.org/10.1177/0018720810385427.
Johnston, J.C., Ruthruff, E., Lien, M.C., 2015. Visual information processing from
multiple displays. Hum. Factors 57 (2), 276297, http://dx.doi.org/10.1177/
0018720814545974.
Kaber, D.B., Liang, Y.L., Zhang, Y., Rogers, M.L., Gangakhedkar, S., 2012. Driver
performance effects of simultaneous visual and cognitive distraction and
adaptation behavior. Transp. Res. F 15 (5), 491501, http://dx.doi.org/10.1016/
j.trf.2012.05.004.
Kahneman, D., 1973. Attention and Effort, 1st ed. Prentice-Hall.
Kao, P.C., Higginson, C.I., Seymour, K., Kamerdze, M., Higginson, J.S., 2015. Walking
stability during cell phone use in healthy adults. Gait Posture 41 (4), 947953,
http://dx.doi.org/10.1016/j.gaitpost.2015.03.347.
Lamberg, E.M., Muratori, L.M., 2012. Cell phones change the way we walk. Gait
Posture 35 (4), 688690, http://dx.doi.org/10.1016/j.gaitpost.2011.12.005.
Lennon, A., Oviedo-Trespalacios, O., Matthews, S., 2017. Pedestrian self-reported
use of smart phones: positive attitudes and high exposure inuence intentions
to cross the road while distracted. Accid. Anal. Prev. 98, 338347, http://dx.doi.
org/10.1016/j.aap.2016.10.028.
Lim, J., Amado, A., Sheehan, L., Van Emmerik, R.E., 2015. Dual task interference
during walking: the effects of texting on situational awareness and gait
stability. Gait Posture 42 (4), 466471, http://dx.doi.org/10.1016/j.gaitpost.
2015.07.060.
Lin, M.I.B., Lin, K.H., 2016. Walking while performing working memory tasks
changes the prefrontal cortex hemodynamic activations and gait kinematics.
Front. Behav. Neurosci. 10, ARTN 9 10.3339/fnbeh.2010.0009.
Nasar, J.L., Troyer, D., 2013. Pedestrian injuries due to mobile phone use in public
places. Accid. Anal. Prev. 57, 9195, http://dx.doi.org/10.1016/j.aap.2013.03.
021.
Nasar, J., Hecht, P., Wener, R., 2008. Mobile telephones, distracted attention, and
pedestrian safety. Accid. Anal. Prev. 40 (1), 6975, http://dx.doi.org/10.1016/j.
aap.2007.04.005.
National Center for Injury Prevention and Control, 2014. WISQARSTM Nonfatal
Injury Reports, 20102014, Retrieved 12/19/2016, from Centers for Disease
Control and Prevention https://webappa.cdc.gov/sasweb/ncipc/nrates2001.
html.
National Highway Trafc Safety Administration, 2016. Trafc Safety Facts 2014
Data. Department of Transportation, Washington DC (DOT HS 812 270).
Neider, M.B., McCarley, J.S., Crowell, J.A., Kaczmarski, H., Kramer, A.F., 2010.
Pedestrians, vehicles, and cell phones. Accid. Anal. Prev. 42 (2), 589594,
http://dx.doi.org/10.1016/j.aap.2009.10.004.
Oberfeld, D., Franke, T., 2013. Evaluating the robustness of repeated measures
analyses: the case of small sample sizes and nonnormal data. Behav. Res.
Methods 45 (3), 792812, http://dx.doi.org/10.3758/s13428-012-0281-2.
Parasuraman, R., Sheridan, T.B., Wickens, C.D., 2008. Situation awareness, mental
workload, and trust in automation: viable, empirically supported cognitive
engineering constructs. J. Cognit. Eng. Decis. Making 2 (2), 140160, http://dx.
doi.org/10.1518/155534308x284417.
Pucher, J., Dijkstra, L., 2003. Promoting safe walking and cycling to improve public
health: lessons from The Netherlands and Germany. Am. J. Public Health 93 (9),
15091516, http://dx.doi.org/10.2105/Ajph.93.9.1509.
Recarte, M.A., Nunes, L.M., 2003. Mental workload while driving: effects on visual
search, discrimination, and decision making. J. Exp. Psychol. 9 (2), 119137.
Rogers, J., 2016. Death by Pokemon? Public safety fears mount as Pokemon GO
craze continues (7/14/2016) Retrieved from http://www.foxnews.com/tech/
2016/07/14/death-by-pokemon-public-safety-fears-mount-as-pokemon-go-
craze-continues.html.
Schabrun, S.M., van den Hoorn, W., Moorcroft, A., Greenland, C., Hodges, P.W.,
2014. Texting and walking: strategies for postural control and implications for
safety. PLoS One 9 (1), e84312, http://dx.doi.org/10.1371/journal.pone.
0084312.
Schwebel, D.C., Stavrinos, D., Byington, K.W., Davis, T., ONeal, E.E., de Jong, D.,
2012. Distraction and pedestrian safety: how talking on the phone, texting,
and listening to music impact crossing the street. Accid. Anal. Prev. 45,
266271, http://dx.doi.org/10.1016/j.aap.2011.07.011.
Selcon, S.J., Taylor, R.M., Koritsas, E., 1991. Workload or situational awareness? TLX
vs. SART for aerospace systems design evaluation. In: Paper Presented at the
Human Factors Society 35th Annual Meeting San Francisco, CA, USA, 9/19/6.
Shahar, A., Alberti, C.F., Clarke, D., Crundall, D., 2010. Hazard perception as a
function of target location and the eld of view. Accid. Anal. Prev. 42 (6),
15771584, http://dx.doi.org/10.1016/j.aap.2010.03.016.
Stavrinos, D., Byington, K.W., Schwebel, D.C., 2011. Distracted walking: cell phones
increase injury risk for college pedestrians. J. Safety Res. 42 (2), 101107,
http://dx.doi.org/10.1016/j.jsr.2011.01.004.
Steelman, K.S., McCarley, J.S., Wickens, C.D., 2013. Great expectations: top-down
attention modulates the costs of clutter and eccentricity. J. Exp. Psychol. 19 (4),
403419, http://dx.doi.org/10.1037/a0034546.
Stelzer, E.M., Wickens, C.D., 2006. Pilots strategically compensate for display
enlargements in surveillance and ight control tasks. Hum. Factors 48 (1),
166181, http://dx.doi.org/10.1518/001872006776412225.
Strayer, D.L., Johnston, W.A., 2001. Driven to distraction: dual-task studies of
simulated driving and conversing on a cellular telephone. Psychol. Sci. 12 (6),
462466, http://dx.doi.org/10.1111/1467-9280.00386.
Taylor, R., 1990. Situation awareness rating technique (SART): the development of
a tool for aircrew systems design (AGARD-CP-478). In: Paper Presented at the
Situational Awareness in Aerospace Operations (Chapter 3), Neuilly sur-Seine,
France.
The Wireless Association, 2015. Wireless Quick Facts, Retrieved from http://www.
ctia.org/your-wireless-life/how-wireless-works/wireless-quick-facts.
Vidulich, M.A., Tsang, P.S., 2012. Mental workload and situation awareness. In:
Salvendy, G. (Ed.), Handbook of Human Factors and Ergonomics. , 4th ed.
Wiley, pp. 243273.
Wickens, C.D., 2008. Multiple resources and mental workload. Hum. Factors 50 (3),
449455, http://dx.doi.org/10.1518/001872008x288394.
Wolfe, J.M., Horowitz, T.S., 2004. What attributes guide the deployment of visual
attention and how do they do it? Nat. Rev. Neurosci. 5 (6), 495501, http://dx.
doi.org/10.1038/nrn1411.
Young, M.S., Brookhuis, K.A., Wickens, C.D., Hancock, P.A., 2015. State of science:
mental workload in ergonomics. Ergonomics 58 (1), 117, http://dx.doi.org/10.
1080/00140139.2014.956151.