575149 HFSXXX10.

1177/0018720815575149Human FactorsCognitive Distraction in the Automobile

SPECIAL SECTION: Peer Commentary on Assessing Cognitive Distraction in Driving

Assessing Cognitive Distraction in the Automobile

David L. Strayer and Jonna Turrill, University of Utah, Salt Lake City, Joel M. Cooper,
Precision Driving Research, Salt Lake City, Utah, James R. Coleman, University of
Utah, Salt Lake City, Nathan Medeiros-Ward, University of Illinois, Urbana-Champaign,
and Francesco Biondi, University of Padova, Padova, Italy

Objective: The objective was to establish a sys- Introduction

tematic framework for measuring and understanding
Driver distraction is increasingly recognized
cognitive distraction in the automobile.
Background: Driver distraction from secondary
as a significant source of injuries and fatalities
in-vehicle activities is increasingly recognized as a sig- on the roadway. In fact, the National Highway
nificant source of injuries and fatalities on the roadway. Traffic Safety Administration (NHTSA) esti-
Method: Across three studies, participants com- mated that inattention accounted for 25% of
pleted eight in-vehicle tasks commonly performed by all police-reported crashes (Ranney, Mazzae,
the driver of an automobile. Primary, secondary, sub- Garrott, & Goodman, 2000; Wang, Knipling,
jective, and physiological measures were collected and & Goodman, 1996). Other estimates have sug-
integrated into a cognitive distraction scale. gested that inattention was a factor in as many
Results: In-vehicle activities, such as listening to as 35% to 50% of all crashes (Sussman, Bishop,
the radio or an audio book, were associated with a low Madnick, & Walker, 1985). More recently, data
level of cognitive workload; the conversation activities
from the 100-Car Naturalistic Driving Study
of talking to a passenger in the vehicle or conversing
with a friend on a handheld or hands-free cell phone
(Dingus et al., 2006) showed that inattention
were associated with a moderate level of cognitive was a factor in 78% of all crashes and near
workload; and using a speech-to-text interfaced e-mail crashes, making it the single largest crash causa-
system involved a high level of cognitive workload. tion factor in the analysis. However, in each of
Conclusion: The research established that there these analyses, the classification of inattention
are significant impairments to driving that stem from was a catchall category encompassing a vari-
the diversion of attention from the task of operating ety of phenomena, including fatigue, driving-
a motor vehicle and that the impairments to driving related distractions (such as glances to mirrors
are directly related to the cognitive workload of these during a merge), nonspecific eye glances away
in-vehicle activities. Moreover, the adoption of voice-
from the forward roadway, and distraction from
based systems in the vehicle may have unintended con-
secondary in-vehicle activities.
sequences that adversely affect traffic safety.
Application: These findings can be used to help Distraction from secondary in-vehicle activi-
inform scientifically based policies on driver distraction, ties stems from a combination of three sources
particularly as they relate to cognitive distraction stem- (Strayer, Watson, & Drews, 2011). Impairments
ming from the diversion of attention to other concur- to driving can arise from a competition for visual
rent activities in the vehicle. processing, for example, when a driver takes his
or her eyes off the road to interact with a device.
Keywords: cognitive workload, cognitive distraction, Impairments can also arise from manual inter-
driving, EEG, visual scanning behavior, divided atten-
ference, as in cases when drivers take their hands
tion, multitasking
off the steering wheel to manipulate a device.
Address correspondence to David L. Strayer, Department
Finally, cognitive sources of distraction occur
of Psychology, University of Utah, 380 S. 1530 E. BEHS when attention is withdrawn from the process-
502, Salt Lake City, UT 84112, USA; e-mail: David. ing of information necessary for the safe opera-
[email protected]. tion of a motor vehicle. These three sources of
HUMAN FACTORS
distraction can operate independently, but they
Vol. 57, No. 8, December 2015, pp. 1300­–1324 are not mutually exclusive, and therefore inter-
DOI: 10.1177/0018720815575149 acting with different in-vehicle devices can
Copyright © 2015, Human Factors and Ergonomics Society. result in impairments from one, two, or all three
Cognitive Distraction in the Automobile 1301

sources. The focus of this article is on develop- and ocular), and subjective workload assess-
ing a valid and sensitive tool for reliably mea- ments. Prior research in aviation psychology has
suring inattention arising from cognitive sources used these measures to assess the cognitive
of distraction. workload of pilots (Kramer, Sirevaag, & Braun,
Standardized efforts to evaluate sources of 1987; Sirevaag et al., 1993). For example,
distraction are not new. Indeed, NHTSA has Kramer et al. (1987) examined pilots’ workload
issued driver distraction guidelines to address by comparing their flight performance during
visual and manual sources of distraction (cf. takeoff, level flight, holding a heading, and land-
NHTSA, 2012). Like other published standards, ing. In this study, flight performance, subjective
these guidelines specify a number of methods measurements, and brain-based physiological
for evaluating the visual (and to a lesser extent, measures all reflected changes in mental work-
manual) demand of secondary-task interactions. load as the primary task of piloting became more
However, there are currently no published stan- difficult.
dards that explicitly and exclusively apply to Following the lead from aviation psychology,
cognitive distraction (but see Engström et al., Strayer et al. (2003; Strayer & Drews, 2007a)
2013, for an evolving discussion on driver inat- used an eye tracker in conjunction with an inci-
tention). Moreover, with regard to the use of sec- dental recognition memory paradigm to deter-
ondary in-vehicle tasks, experimental studies mine what information in the driving scene par-
using driving simulators or instrumented vehi- ticipants attended. The authors found that par-
cles (Caird, Willness, Steel, & Scialfa, 2008; ticipants were more than twice as likely to
Horrey & Wickens, 2006; Strayer, Drews, & recognize objects encountered in the single-task
Johnston, 2003; Strayer & Johnston, 2001; see driving condition as when they were driving and
also the epidemiological studies by Redelmeier concurrently talking on a hands-free cell phone.
& Tibshirani, 1997, and McEvoy et al., 2005) Even when the participants’ eyes were directed
have produced strikingly different estimates of at objects in the driving environment for the
driving impairment and crash risk than the cor- same duration, they were less likely to remem-
relation-based naturalistic studies of driving ber them if they had been conversing on a cell
(Dingus et al., 2006; Klauer et al., 2014), so the phone. Strayer and Drews (2007a; see also
precise crash risk associated with cell phone use Strayer, Cooper, & Drews, 2004) suggested that
is far from settled. using a cell phone induces a form of inattention
In fact, cognitive distraction is the most diffi- blindness whereby the cell phone conversation
cult of the three sources of distraction to assess diverts attention from processing the informa-
because of the problems associated with observ- tion necessary to safely operate a motor vehicle.
ing what a driver’s brain (as opposed to hands or To evaluate cognitive driver distraction, reac-
eyes) is doing. Furthermore, changes in driving tion time (RT) is typically measured using
performance associated with cognitive distrac- ­sudden-onset stimuli (such as a braking lead
tion have been shown to be qualitatively differ- vehicle or a flashing dashboard light) that require
ent from those associated with visual distraction an immediate response from drivers. Results are
(Angell et al., 2006; Engström, Johansson, & often interpreted as an indication of a driver’s
Östlund, 2005). For example, visual distraction ability to quickly and safely respond to the sud-
has been shown to increase the variability of lane den appearance of a threat. RT measures show a
position, whereas cognitive distraction has been great deal of consistency, regardless of whether
shown to decrease the variability of lane position drivers are responding to a lead braking vehicle,
(Cooper, Medeiros-Ward, & Strayer, 2013; to peripherally flashing lights, or to the appear-
Medeiros-Ward, Cooper, & Strayer, 2013). ance of unexpected objects. In all cases, drivers
In order to assess cognitive workload, prior engaged in secondary in-vehicle activities are
experimental research has typically employed slower to react than drivers who are paying
some combination of primary-task and attention to the roadway. Because of the consis-
secondary-­task behavioral measures, physiolog- tent sensitivity of RT measures, a new effort is
ical measures (e.g., neurological, cardiovascular, being considered by the International Standards
1302 December 2015 - Human Factors

Organization (ISO) to standardize a protocol for been fully established (but see Angell et al.,
RT measurement while driving (ISO, 2012). 2006, for a related project that focused primarily
This technique will be discussed in more detail on visual/manual tasks and simple cognitive
later. tasks, such as listening to a book on tape).
Cognitive distraction can also be measured A second knowledge gap with respect to cog-
through a variety of physiological techniques. nitive distraction is that there is no comprehen-
Among these, direct measures of brain activity sive way for assessing the distraction potential of
may be the most compelling. One approach is to any single activity and relating that to the distrac-
use time-locked signals of electroencephalo- tion potential of other in-vehicle activities. What
graphic (EEG) activity, referred to as event- is needed is a comprehensive method for assess-
related brain potentials (ERPs). This technique ing secondary-task cognitive distraction and a
provides a window into the brain activity that is method to integrate the results into a simple,
associated with responses to imperative driving meaningful metric. This metric would allow
events (e.g., brake lights on a lead vehicle). researchers to make definitive statements about
Using this technique, Strayer and Drews (2007a) how one source of cognitive distraction com-
found that the brain activity associated with pro- pares to another. Although it is clear that activi-
cessing the information necessary for the safe ties, such as conversing on a cell phone, degrade
operation of a motor vehicle was suppressed certain aspects of driving, it is not clear how to
when drivers were talking on a cell phone. These interpret the magnitude of the findings. Is the
data help to explain why drivers using a cell cognitive distraction of cell phone conversation
phone fail to see information in the driving scene so severe that it is clearly incompatible with safe
and why their response time to scene events is driving, or is it sufficiently benign that it is nearly
slowed; they do not encode the information as indistinguishable from listening to the radio?
well as they do when they are not distracted by a In this article, we present the results from
cell phone conversation. In situations when the three experiments designed to systematically
driver is required to react quickly, the ERP data measure cognitive workload in the automobile.
suggest that those drivers using a cell phone will The first experiment served as a control in which
be less able to do so because of the diversion of participants performed eight different tasks
attention from driving to the phone conversation. without the concurrent operation of a motor
It is important to note that the demonstrations vehicle. In the second experiment, participants
of inattention blindness provide a pure measure performed the same eight tasks while operating
of cognitive distraction because participants’ a high-fidelity driving simulator. In the third
eyes were on the road and they were not manu- experiment, participants performed the eight
ally manipulating the phone in dual-task condi- tasks while driving an instrumented vehicle in a
tions. However, one shortcoming of the litera- residential section of a city.
ture on cognitive distraction is that it has often In each experiment, the order of the eight
assessed various secondary tasks in a piecemeal tasks was counterbalanced, and the tasks
fashion. Although many forms of cognitive dis- involved (a) a baseline single-task condition
traction have been evaluated (e.g., listening to (i.e., no concurrent secondary task), (b) concur-
the radio, talking on a cell phone, talking to a rent listening to a radio, (c) concurrent listening
passenger, interacting with a speech-to-text sys- to an audio book, (d) concurrent conversation
tem), no single study has yet involved analyzing with a passenger seated next to the participant,
a comprehensive set of common real-world (e) concurrent conversation on a handheld cell
tasks using the same experimental protocol. A phone, (f) concurrent conversation on a hands-
number of studies have demonstrated the sensi- free cell phone, (g) concurrent interaction with a
tivity of cognitive distraction metrics to grada- speech-to-text interfaced e-mail system, and (h)
tions in artificial task difficulty (Mehler, Reimer, concurrent performance with an auditory ver-
& Coughlin, 2012; Reimer, Mehler, Pohlmeyer, sion of the Operation Span (OSPAN) task. The
Coughlin, & Dusek, 2006), yet sensitivity to OSPAN task is a demanding complex span task
gradations in real-world cognitive tasks has not developed by Turner and Engle (1989) that
Cognitive Distraction in the Automobile 1303

requires participants to simultaneously perform Experiment 1

a math and memorization task and was chosen to Experiment 1 was designed to provide a
anchor the highest level of cognitive workload. baseline assessment of the eight tasks described
Each task allows the driver to keep his or her earlier. In this controlled assessment, partici-
eyes on the road and, with the exception of the pants were seated in front of a computer monitor
handheld cell phone condition, both hands on that displayed a static fixation cross, and they
the steering wheel, so that any impairment to performed the conditions without the added task
driving must stem from cognitive sources asso- of driving. The objective was to establish the
ciated with the diversion of attention from the cognitive workload associated with each activ-
task of operating the motor vehicle. Based upon ity and to thereby predict, given the capacity
prior research (Strayer et al., 2011), these tasks limits of attention, the accompanied cognitive
were hypothesized to reflect increasing levels of distraction from performing that activity while
cognitive workload and were selected because operating a motor vehicle.
they are representative of the type of activities
commonly engaged in while operating a motor
vehicle (Stutts et al., 2003). Method
In each of the experiments, we used a combi- Participants. Thirty-eight participants (20
nation of performance indices to assess mental men and 18 women) from the University of Utah
workload, including RT and accuracy, in participated in the experiment. Participants
response to a peripheral light detection task (the ranged in age from 18 to 30 years, with an aver-
Detection Response Task [DRT]; ISO, 2012), age age of 22.2 years. All had normal neurologi-
subjective workload measures from the NASA cal functioning, normal or corrected-to-normal
Task Load Index (NASA-TLX; Hart & Stave- visual acuity, normal color vision (Ishihara,
land, 1988), and physiological measures associ- 1993), and a valid driver’s license and were flu-
ated with EEG activity and ERPs time locked to ent in English. Participants’ years of driving
the peripheral light detection task. We also experience ranged from 2.5 to 14.5, with an
obtained primary-task measures of driving in average of 6.9 years. All participants owned a
experiments using the driving simulator and cellular phone and reported that they used their
instrumented vehicle. By combining these dif- phone regularly while driving. They were
ferent measures of cognitive workload, we pro- recruited via university-approved flyers posted
vide a more comprehensive assessment than on campus bulletin boards. Interested individu-
what would be afforded with using only one als contacted an e-mail address for further infor-
technique (Gopher & Donchin, 1986; Sirevaag mation and to schedule an appointment. Eligible
et al., 1993). participants reported a clean driving history
After describing the methods and results of (e.g., no at-fault accidents or history of traffic
each study in greater detail, we report a meta- violations).
analysis that integrates the different dependent Materials. Subjective workload ratings were
measures across the three studies to provide an collected using the NASA-TLX survey devel-
overall cognitive distraction metric for each of oped by Hart and Staveland (1988). After com-
the concurrent secondary tasks. In particular, pleting each of the conditions, participants
we used these data to develop a rating system responded to each of the six items on a 21-point
for cognitive distraction whereby nondistracted Likert scale ranging from very low to very high.
single-task driving anchored the low end (Cat- The questions in the NASA-TLX were as
egory 1) and the OSPAN task anchored the high follows:
end (Category 5) of the scale. For each of the
other tasks, its relative position compared to the (a) How mentally demanding was the task?
low and high anchors provides an index of the (b) How physically demanding was the task?
cognitive workload for that activity when con- (c) How hurried or rushed was the pace of the task?
currently paired with the operation of a motor (d) How successful were you in accomplishing what
vehicle. you were asked to do?
1304 December 2015 - Human Factors

(e) How hard did you have to work to accomplish artifact rejection techniques (Semlitsch, Anderer,
your level of performance? Schuster, & Presslich, 1986), and the data were
(f) How insecure, discouraged, irritated, stressed, epoched 200 ms before to 1,200 ms after the
and annoyed were you? onset of the green target light. These epochs
were then filtered with a band pass, zero phase
Equipment. Cellular service was provided shift filter of 0.1 to 10 Hz. Finally, events that
by Sprint. The cellular phone was manufactured exceeded an artifact rejection criterion of 100
by Samsung (Model M360), and the hands-free µV were rejected, and the remaining events were
earpiece was manufactured by Jawbone (Model averaged to obtain one subject’s average wave-
Era). Participants dialed a friend or family mem- form for each condition in the experiment.
ber, and the volume for both the cellular phone Procedure.  Prior to their appointment time,
and the hands-free earpiece was adjusted prior to participants were sent a general demographic
the task. survey. Upon arrival at the lab in the Behavioral
NaturalReader 10.0 software was used to Sciences building, participants read and signed
simulate an interactive messaging service with the institutional review board (IRB)–approved
speech-to-text features. Participants indicated consent document, and the research team fitted
friend names prior to beginning the study. These an EEG cap to each participant’s head. A refer-
names were entered into a template containing ence electrode was placed behind the left ear on
generic e-mail and text messages (e.g., “Text the mastoid bone, and electrode site FP1 served
from [name]: ‘Hey! Let’s meet for lunch some- as the ground. Electrooculogram electrodes
time this week. When are you free?’”). Partici- were placed at the lateral canthi of both eyes
pants were given a short list of commands (i.e., (horizontal) and above and below the left eye
repeat, reply, forward, delete, and next message) (vertical) to track eye movements and record eye
that were used in order for the messaging pro- blinks for later data processing. Participants’
gram to respond. The NaturalReader program field of view and normal range of motion were
was controlled by the experimenter, who reacted not impeded when wearing the EEG cap.
to the participants’ verbal commands, mimick- Participants were asked to complete eight dif-
ing a speech detection system with perfect fidel- ferent 10-minute conditions that were chosen to
ity. If a participant did not use the correct com- provide a range of cognitive workload. The
mand, the NaturalReader program would not order of these conditions were counterbalanced
continue. across participants. The participants were seated
Hosted on a 32-bit research laptop, Neu- an average of 65 cm from a computer screen dis-
roScan 4.5 software was used to collect continu- playing a fixation cross. Participants were asked
ous EEG in the experiment. The EEG was to look at the fixation cross and try to avoid
recorded using a NeuroScan32-electrode making unnecessary movements during the
NuAmp amplifier. The EEG was filtered online completion of each task.
with a low pass filter of 50 Hz and a high pass Described here in hypothesized ascending
filter set to DC with a sample A/D rate of 250 order of cognitive workload (Strayer et al., 2011;
Hz. The DRT software communicated with the Strayer & Drews, 2007b), the single-task condi-
NeuroScan system via a parallel port connection tion was selected to provide a baseline of cogni-
to create event markers associated with the con- tive workload when performing the DRT (see
tinuous EEG. These event markers allowed for details described later). In the second condition,
offline stimulus-locked analysis of the EEG participants were allowed to select a radio sta-
recordings (i.e., the DRT stimuli were used for tion to which they normally listen when driving.
the creating of time-locked ERPs; see details Depending on the participant’s selection, the
described later). The EEG was first visually live radio broadcast was a mix of music and
inspected for artifacts, and any sections with talking. Before the condition began, participants
excessive noise from movement or electronic selected the station and adjusted the volume to a
interference were removed. Next, the influence comfortable level. Once they began the record-
of blinks on the EEG was corrected using ocular ing session, they were not allowed to change the
Cognitive Distraction in the Automobile 1305

station to avoid the influence of manual manipu- were implemented using the “Wizard of Oz”
lations. paradigm (Kelley, 1983; Lee, Caven, Haake, &
In the third condition, participants chose one Brown, 2001) in which the participant’s speech
of three audio book excerpts. They selected from was actually being secretly entered into the com-
excerpts of chapter 1 from The Giver (Lowry, puter by the experimenter with perfect fidelity.
2002), chapter 20 from Water for Elephants Prior to beginning the condition, the participant
(Gruen, 2006), or chapter 11 from Harry Potter was familiarized with the program’s basic com-
and the Sorcerer’s Stone (Rowling, 1998). Once mands, which were repeat, reply, forward, delete,
again, all manual adjustments to volume were and next message. The participant completed a
made before the condition began. Participants simple tutorial to become familiar with how the
were informed that at the end of the audio book, commands function. Next is an example of an
they would take a simple five-item quiz about the interaction using the speech-to-text system:
events in their chosen audio book excerpt. This
quiz was to ensure that participants attended to Participant’s command: “Next message.”
the story. Across the three experiments reported, Message read to participant: “Text message
the accuracy on the quiz averaged 86%. from Mary: ‘When do you want to go to the
Conditions 4 through 6 involved different grocery store?’”
forms of conversation. In each of the conditions, Commands list read to participant: “Repeat,
the interlocutors were asked to speak and listen reply, forward, delete, next message.”
in equal proportions (i.e., 50% speaking and Participant’s response: “Reply: ‘Hi Mary, I can
50% listening). The fourth condition entailed go this evening, let’s say 8 o’clock.’ Send.”
conversation with the experimenter seated next
to the participant. Participants wrote down a few The final condition provided the highest level
conversation topics at the beginning of the study. of cognitive workload: solving simple math
Experimenters would ask the participant to start problems and remembering words. The math
telling an interesting story from the list and then and memory problems were read aloud by the
helped to maintain an engaging conversation by experimenter, and the participant’s verbal
asking questions about the story or by respond- answers were recorded by the experimenter. Par-
ing with a story of their own. ticipants completed a short example of the
The fifth condition required the participant to OSPAN task before beginning the condition.
call a friend or family member and talk with that OSPAN is a standardized task that can be
person on a handheld cellular phone. The call used to create a challenging dual-task condition
was initiated and the volume was adjusted before (Sanbonmatsu, Strayer, Medeiros-Ward, & Wat-
the condition began. Because the microswitch son, 2013). Participants completed an auditory
for the DRT (described later) was attached to the version of the OSPAN task developed by Wat-
left thumb, participants held the phone with their son and Strayer (2010) in which they attempted
right hand. Most participants indicated that this to recall single-syllable words in serial order
was the hand they normally used to hold their while solving mathematical problems. In the
handheld device. Similarly, the sixth condition auditory OSPAN task, participants were asked
was a conversation with either the same or a dif- to remember a series of two to five words that
ferent friend or family member, but it occurred were interspersed with math verification prob-
via the hands-free Bluetooth earpiece. Partici- lems (e.g., Given “[3 / 1] – 1 = 2? Cat. [2 × 2] +
pants indicated in which ear they wished to use 1 = 4? Box. Recall,” the participant should have
the hands-free earpiece. answered “True” and “False” to the math prob-
For Condition 7, the participant interacted lems when they were presented and recalled cat
with a text-to-speech program, NaturalReader and box in the order in which they were pre-
10.0, which simulated speech-based e-mail and sented when given the recall probe). The math
text-messaging services. Participants interacted problems could be repeated at the participant’s
with the program as if it were a fully automated request if the participant did not comprehend
system. Perfect speech recognition capabilities them.
1306 December 2015 - Human Factors

Figure 1. Detection Response Task reaction time Figure 2. Detection Response Task A′ across
across Experiments 1 through 3. Error bars indicate Experiments 1 through 3. Error bars indicate standard
standard error of the mean. error of the mean.

In each of these conditions, participants also peripheral detection task. The RT and accuracy
performed a variant of the DRT (cf. the nascent data for the DRT are plotted in Figures 1 and 2,
ISO standard for measuring cognitive distrac- respectively. RT for correct responses (i.e.,
tion; ISO, 2012). In our version of the DRT, red green-light responses) was measured to the
and green lights were presented every 3 to 5 s nearest millisecond, and the accuracy data were
via a head-mounted device. The light was posi- converted to the nonparametric measure of sen-
tioned an average 15° to the left and 7.5° above sitivity, A′, whereby a response to a green light
the participant’s left eye and was held in a fixed was coded as a “hit,” nonresponses to a red light
position on the head with a headband that did were coded as a “correct rejection,” nonre-
not interfere with the EEG data collection or sponses to a green light were coded as a “miss,”
with a clear field of view. Red lights were pre- and responses to a red light were coded as a
sented 80% of the time, and green lights were “false alarm” (Pollack & Norman, 1964). A′
presented 20% of the time. Both the color of the ranges between 0.5 for chance performance and
light and the interval between trials (e.g., 3–5 s) 1.0 for perfect accuracy (Parasuraman & Davies,
were randomized (i.e., this is a 20/80 oddball 1984). A repeated-measures ANOVA revealed a
with stimuli presented in a Bernoulli sequence main effect of condition on RT, F(7, 259) =
with an interstimulus interval of 3 to 5 s; see 33.87, p < .01, partial η2 = .48, and A′, F(7,
Donchin, 1981). Using a go/no-go design, par- 259) = 7.62, p < .01, partial η2 = .17.
ticipants were instructed to respond to the green NASA-TLX.  The data for the six NASA-TLX
light as quickly as they could by depressing a subjective workload ratings are plotted in Figure
microswitch that was placed on the left thumb 3. In each of the figures, the eight conditions are
but to not respond to the red lights. The lights plotted across the abscissa, and the 21-point Lik-
remained illuminated until a response was made ert scale workload rating is represented on the
or 1 s had elapsed. Response RT was recorded ordinate, ranging from very low, 1, to very high,
with millisecond accuracy. 21. The subjective workload ratings increased
systematically across the conditions, with the
Results notable exception of physical workload, which
DRT. The DRT data reflect the manual remained relatively flat (with a noticeable bump
response to the red and green lights in the in the handheld cell phone condition). Given
Cognitive Distraction in the Automobile 1307

Figure 3. The six NASA Task Load Index subjective workload ratings across Experiments 1 through 3.
Error bars indicate standard error of the mean.
1308 December 2015 - Human Factors

-5

0
Amplitude (uV)
5

10
Laboratory
15 Simulator
Vehicle
20

1100
0

1000

1200
-100

100
200
300
400

500
600
700
800
900
Time (msec)

Figure 4. Single-task event-related brain potentials at Pz across

Experiments 1 through 3.

that the conditions were selected to allow the we measured both its peak latency and the
driver to keep their eyes on the road and their amplitude.
hands on wheel, the physical workload ratings In Figure 5, P300 peak latency, measured as
are consistent with what we would expect. the point in time of maximum positivity in a
A series of repeated-measures ANOVAs window between 400 and 700 ms, is plotted for
revealed that NASA-TLX ratings increased for each of the conditions in the experiment. An
mental workload, F(7, 259) = 83.12, p < .01, par- ANOVA revealed that P300 latency systemati-
tial η2 = .69; physical workload, F(7, 259) = cally increased across the conditions, F(7, 259) =
3.33, p < .01, partial η2 = .08; temporal demand, 13.80, p < .01, partial η2 = .27. The P300 ampli-
F(7, 259) = 28.74, p < .01, partial η2 = .44; per- tude was quantified by computing the average
formance, F(7, 259) = 14.92, p < .01, partial η2 = area under the curve between 400 and 700 ms.
.29; effort, F(7, 259) = 64.87, p < .01, partial η2 = Figure 6 plots P300 amplitude as a function of
.64; and frustration, F(7, 259) = 33.79, p < .01, condition. An ANOVA revealed a main effect of
partial η2 = .48. condition, F(7, 259) = 4.02, p < .01, partial η2 =
Physiological measures. Figure 4 presents .10; however, there was no systematic pattern in
the grand average ERP waveforms for single this effect.
task obtained in Experiments 1 through 3 at the
midline parietal electrode site (Pz) that were Discussion
time locked to the onset of green lights in the Experiment 1 was designed to provide a
DRT. In the figure, the amplitude in microvolts baseline assessment of several activities com-
is cross-plotted with time in milliseconds. A close monly engaged in while operating a motor
inspection reveals a well-defined P2-N2-P300 vehicle (Stutts et al., 2003). In this assessment,
ERP component structure in Experiment 1 that participants did not drive but were seated in
becomes noisier (i.e., more ambient artifact) in front of a computer monitor that displayed the
the driving simulator and instrumented vehicle static fixation cross.
studies (to be discussed in Experiments 2 and 3, Taken together, cognitive workload increased
respectively). We focused on the P300 compo- with condition, with single-task driving anchor-
nent of the ERP because of its sensitivity to cog- ing the low end and the OSPAN condition
nitive workload (Strayer & Drews, 2007a), and anchoring the high end. Clearly, not all
Cognitive Distraction in the Automobile 1309

Figure 5. P300 peak latency from Experiment 1. Error bars indicate

standard error of the mean.

Figure 6. P300 amplitude across Experiments 1 through 3. Error bars

indicate standard error of the mean.
1310 December 2015 - Human Factors

in-vehicle activities have the same level of cog- cognitive distraction (Caird et al., 2008; Horrey
nitive workload. Indeed, some of the older in- & Wickens, 2006).
vehicle activities, such as listening to the radio,
were associated with negligible increases in Method
cognitive workload. By contrast, some of the Participants.  Thirty-two participants (22
newer technologies, such as speech-to-text inter- men and 10 women) from the University of Utah
actions with e-mail, were associated with some participated in the experiment. Participants
of the highest levels of workload. It is notewor- ranged in age from 19 to 36, with an average age
thy that the in-vehicle activities that were evalu- of 23.5 years. All participants met the same eli-
ated were “pure” measures of cognitive work- gibility requirements and were recruited in the
load in that these tasks do not require the partici- same manner as described in Experiment 1.
pant to divert his or her eyes from the road or Equipment. In addition to the equipment
hands from the steering wheel. used in Experiment 1, the present study used a
The results from the different measures fixed-base high-fidelity driving simulator (Mfr.
obtained in Experiment 1 had a good correspon- L-3 Communications) with high-resolution dis-
dence and together help to lay the foundation for plays providing a 180° field of view. The dash-
a metric of cognitive workload that increases board instrumentation, steering wheel, gas, and
across the conditions. As the cognitive workload brake pedals are from a Ford Crown Victoria
associated with performing an activity increases, sedan with an automatic transmission. The sim-
the cognitive distraction associated with per- ulator incorporated vehicle dynamics, traffic
forming that activity while operating a motor scenario, and road surface software to provide
vehicle increases. Given the capacity limits of realistic scenes and traffic conditions. In the
attention (Kahneman, 1973), performing an in- driving simulator, the DRT was implemented by
vehicle task that is associated with significant mounting the red/green light on the vehicle
cognitive workload leaves less attention to be dashboard directly in front of the participant. All
allocated to the task of driving. That is, cogni- other equipment was identical to Experiment 1.
tive distraction is the consequence of perform- Procedure.  The procedures used in Experi-
ing an attention-demanding concurrent activity ment 1 were also used in Experiment 2, with the
while driving (i.e., driving performance in following modifications. In Experiment 2, we
Experiments 2 and 3 should be adversely used a car-following paradigm in which partici-
affected by in-vehicle cognitive workload). pants drove on a simulated multilane freeway
with moderate traffic (approximately 1,500
Experiment 2 vehicles per lane per hour). Each condition
The goal of Experiment 2 was to extend lasted approximately 10 min, and the posted
the findings from Experiment 1 to operating speed on the freeway was 65 mph. Participants
a high-fidelity driving simulator. Given the followed a pace car that would apply its brakes
increase in cognitive workload associated with aperiodically (there were total of 20 unpredict-
performing the different in-vehicle activities, able braking events in each condition). Partici-
we expected that measures of driving perfor- pants were not allowed to change lanes to pass
mance would be adversely affected with their the pace car and were asked to maintain a 2-s
concurrent performance. The driving simulator following distance behind the pace car. Partici-
used a car-following paradigm on a multilane pants were given a 3-min practice session to
highway with moderate traffic. Participants fol- familiarize themselves with the driving simula-
lowed a lead vehicle that braked aperiodically tor, scenario design, and prescribed following
throughout the scenario, and in addition to the distance.
measures collected in Experiment 1, we also
collected brake RT and following distance, as Results
these variables associated with the primary task Driving performance measures. We exam-
of driving have been shown to be sensitive to ined two measures of driving performance in
Cognitive Distraction in the Automobile 1311

Figure 7. Brake reaction time from Experiment 2. Error bars indicate

standard error of the mean.

Experiment 2 that prior research has established difference between conditions is even larger at
are sensitive to cognitive distraction (Caird the tail of the RT distribution (e.g., the speech-
et al., 2008; Horrey & Wickens, 2006). Figure 7 to-text/single-task RT ratio was 9% at the
presents the brake RT measured as the time median, whereas this difference was 36% at the
interval between the onset of the pace car’s 90th percentile of the RT distribution; see also
brake lights and the onset of the participant’s Ratcliff & Strayer, 2014).
braking response (i.e., a 1% depression of the DRT.  The RT and accuracy data for the DRT
brake pedal). are plotted in Figures 1 and 2, respectively. A
Figure 8 presents the following distance, repeated-measures ANOVA revealed that RT
measured as the distance between the rear bum- increased across condition, F(7, 217) = 13.51,
per of the pace car and the front bumper of the p < .01, partial η2 = .30, and that A′ decreased
participant’s car at the moment of brake onset. A across condition, F(7, 217) = 21.54, p < .01, par-
repeated-measures ANOVA revealed that both tial η2 = .41.
RT, F(7, 217) = 10.11, p < .01, partial η2 = .25, NASA-TLX. The data for the six NASA-
and following distance, F(7, 217) = 6.26, p < .05, TLX subjective workload ratings are plotted in
partial η2 = .17, increased across condition. A Figure 3. The subjective workload ratings
subsidiary linear mixed-model analysis that held increased systematically across the conditions.
following distance constant showed that brake A series of repeated-measures ANOVAs
RT increased as a function of condition over and revealed that NASA-TLX ratings increased for
above any compensatory effects associated with mental workload, F(7, 217) = 43.55, p < .01,
following distance, F(7, 3972) = 12.77, p < .01. partial η2 = .58; physical workload, F(7, 217) =
These data establish that in-vehicle activities 5.03, p < .01, partial η2 = .14; temporal demand,
that differ in their attentional requirements have F(7, 217) = 26.92, p < .01, partial η2 = .47; per-
differential effects on driving. Moreover, the formance, F(7, 217) = 9.27, p < .01, partial η2 =
1312 December 2015 - Human Factors

Figure 8. Following distance from Experiment 2. Error bars indicate

standard error of the mean.

.23; effort, F(7, 217) = 35.48, p < .01, partial Discussion

η2 = .53; and frustration, F(7, 217) = 23.83, p < Experiment 2 replicated and extended the
.01, partial η2 = .44. pattern obtained in Experiment 1. Most impor-
Physiological measures. EEG was recorded tantly, the increases in cognitive workload
in Experiment 2 using the same protocol as that of resulted in systematic changes in driving per-
Experiment 1. The resulting ERPs from single formance compared to nondistracted driving.
task are plotted in Figure 4 alongside the same In particular, brake RT to imperative events in
conditions from Experiments 1 and 3. As dis- the driving simulator systematically increased
cussed earlier, the ERP was degraded as we moved as a function of the cognitive workload associ-
from the laboratory to the driving simulator due to ated with performing the different in-vehicle
the increased biological noise from eye/head/body activities. Importantly, this pattern held even
movements and electronic noise from the driving when controlling for the increased following
simulator. Because of the added noise, we were distance drivers adopted in these conditions.
not able to reliably measure P300 latency. As in The P300 data also replicate our earlier reports
Experiment 1, the P300 amplitude data, presented of suppressed P300 activity when comparing
in Figure 6, were quantified by computing the single-task and hands-free cell phone conditions
average area under the curve between 400 and 700 (Strayer & Drews, 2007a).
ms. An ANOVA of the P300 amplitude revealed a Despite the fact that the processing require-
main effect of condition, F(7, 217) = 4.38, p < .01, ments of the DRT are minimal, it is possible that
partial η2 = .12. Planned comparisons found that its inclusion could increase the cognitive work-
the single task did not differ from the radio and load of the driver compared to driving condi-
audio book but was significantly different from tions without the DRT. To test for this possibil-
the conversation and OSPAN tasks. ity, we ran another control experiment with 19
Cognitive Distraction in the Automobile 1313

participants using the same protocol as Experi- participated in the experiment. Participants
ment 2 but without the DRT. We focused on the ranged in age from 18 to 33, with an average age
subjective workload ratings from the NASA- of 23.5 years. Participants’ years of driving
TLX because these measures were common to experience ranged from 2 to 17, with an average
both experiments. A 2 (DRT task load: with vs. of 7.1 years. All participants met the same eligi-
without the DRT) × 8 (condition) between-sub- bility requirements and were recruited in the
jects MANOVA was performed to test for differ- same manner as described in Experiment 1. The
ences in subjective workload associated with the Division of Risk Management Department at the
imposition of the DRT. The MANOVA revealed University of Utah ran a motor vehicles record
a significant main effect of condition, F(35, (MVR) report on each prospective participant to
15) = 8.67, p < .01, partial η2 = .95; however, nei- ensure participation eligibility based on a clean
ther the main effect of DRT task load nor the driving history (e.g., no at-fault accidents, sus-
DRT Task Load × Condition interaction were pended licensures, or history of traffic viola-
significant (all ps > .25). This finding establishes tions). In addition, following university policy,
that the imposition of the DRT did not increase each prospective participant was required to
the cognitive workload of the drive (as measured complete a university-devised 20-minute online
using the NASA-TLX). defensive driving course and pass the certifica-
It is worth considering the pattern of data had tion test.
participants protected the driving task at the Equipment. In addition to the equipment
expense of the other in-vehicle activities. In such used in Experiment 1, Experiment 3 used an
a case, we would expect that the primary task instrumented 2010 Subaru Outback. The vehicle
measures would be insensitive to secondary-task was augmented with four 1,080-pixel LifeCam
workload (i.e., Figures 7, 8, and 9, to be discussed USB cameras that captured the driving environ-
later), would be flat, and that there would be no ment and participants’ facial features. All other
main effect of condition for these measures. equipment was identical to Experiment 1.
Instead, we show that the mental resources avail- Procedure.  The procedures used in Experi-
able for driving are inversely related to the cogni- ment 1 were also used in Experiment 3, with the
tive workload of the concurrent secondary task. following modifications. Prior to their appoint-
ment time, participants were sent the IRB-
Experiment 3 approved informed consent document, general
The purpose of Experiment 3 was to establish demographic survey, and instructions for com-
that the patterns obtained in the laboratory and pleting the 20-min online defensive driving
driving simulator generalize to the operation course and the certification test. Prior to the
of an instrumented vehicle on the residential experimental session, we obtained an MVR
roadways. Doing so is important because the report on the driver to ensure a clean driving
consequences of impaired driving in the city are record.
different from that of a driving simulator (e.g., Before beginning the study, the driver was
a crash in the real world can have life-or-death familiarized with the controls of the instrumented
consequences, whereas this is not the case in vehicle, adjusted the mirrors and seat, and was
the driving simulator). Participants drove an informed of the tasks to be completed while driv-
instrumented vehicle in a residential section of ing. The participant drove around a parking lot in
a city while concurrently performing the eight order to become familiar with the handling of the
conditions used in Experiments 1 and 2. If the vehicle. Next, participants drove one circuit on a
findings generalize, then there should be corre- 2.75-mile loop in the Avenues section of Salt
spondence between the results of Experiment 3 Lake City, Utah, in order to become familiar with
and those of Experiments 1 and 2. the route itself. The route provided a suburban
driving environment and contains nine all-way
Method controlled stop signs, one two-way stop sign, and
Participants. 
Thirty-two participants (12 two stoplights. A research assistant and an exper-
men and 20 women) from the University of Utah imenter accompanied the participant in the
1314 December 2015 - Human Factors

Figure 9. Glances at hazard locations in Experiment 3. Error bars

indicate standard error of the mean.

vehicle at all times. The research assistant sat in critical locations were identified for analysis.
the rear, and the experimenter sat in the front pas- These locations included all four-way and two-
senger seat and had ready access to a redundant way stops as well as pedestrian crosswalks. At
braking system and notified the driver of any each of these critical locations, eye movement
potential roadway hazards. data were coded frame by frame to record
The driver’s task was to follow the route glances to the left and to the right of the forward
defined previously while complying with all roadway. A measure of glance probability
local traffic rules, including a 25-mph speed reflects the proportion of complete scans (out of
restriction. If drivers began to exceed 25 mph, the 24 critical locations) in each of the condi-
they were reminded of this restriction by the tions. Scans were recorded as complete if driv-
research team. Throughout each condition, the ers looked to both the left and right. Partially
driver completed the DRT. Each condition lasted complete scans were recorded where the drivers
approximately 10 min, which was the average looked to either the left or right, and incomplete
time required to make one loop around the track. scans were recorded when drivers failed to scan
for hazards. Each drive was analyzed by at least
Results two trained coders, and any discrepancies in the
Driving performance. Because participants coding were flagged and reviewed for consis-
did not follow a lead vehicle as they did in tency by a third coder. In general, coders were
Experiment 2, following distance or brake very accurate and only a small number of events
response data were not available for analysis. needed to be double-checked.
However, because high-definition cameras were The eye glance data for each condition are
used to record the driving environment, manual plotted in Figure 9. A repeated-measures
coding of eye movement data was possible. ANOVA indicated that at critical locations, driv-
Prior to analyzing the eye movement data, 24 ers made progressively fewer scans to the right
Cognitive Distraction in the Automobile 1315

and left of the forward roadway as cognitive effect of condition, but planned comparisons
workload increased, F(7, 168) = 5.92, p < .01, found that the single-task condition significantly
partial η2 = .20. These data replicate and extend differed from speech-to-text and OSPAN condi-
the important work of Taylor et al. (2013) by tions (p < .05).
establishing that the same failures to scan for
roadway hazards observed in a driving simulator Discussion
are found in an instrumented vehicle. Moreover,
Experiment 3 replicated and extended the
these results establish that that there is a system-
findings from the prior experiments in sev-
atic decrease in scanning for hazards as cogni-
eral important ways. Most importantly, they
tive workload increases.
document that the patterns observed in the con-
DRT.  The RT and accuracy data for the DRT
trolled laboratory setting of Experiment 1 and
are plotted in Figures 1 and 2, respectively. A
in the driving simulator setting of Experiment
repeated-measures ANOVA revealed that RT
2 generalize to what was observed in the instru-
increased across condition, F(7, 217) = 27.21,
mented vehicle. There was a systematic increase
p < .01, partial η2 = .47, and that A′ decreased
in cognitive workload across condition, and
across condition, F(7, 217) = 19.17, p < .01, par-
importantly, driving performance as measured
tial η2 = .38.
by scanning for potential hazards decreased as a
NASA-TLX.  The data for the six NASA-TLX
function of in-vehicle secondary task. This latter
subjective workload ratings are plotted in Figure
finding replicates prior studies (Harbluk & Noy,
3. The subjective workload ratings increased sys-
2002; Harbluk, Noy, Trbovich, & Eizenman,
tematically with condition. A series of repeated-
2007; McCarley et al., 2004; Recarte & Nunes,
measures ANOVAs showed that NASA-TLX
2000, 2003; Reimer, 2009; Reimer, Mehler,
ratings increased for mental workload, F(7,
Wang, & Coughlin, 2012; Sodhi, Reimer, &
217) = 52.46, p < .01, partial η2 = .63; physical
Llamazares, 2002) that have shown that visual
workload, F(7, 217) = 10.01, p < .01, partial η2 =
scanning behavior is impaired with increases
.24; temporal demand, F(7, 217) = 37.81, p < .01,
in cognitive workload. As such, it suggests
partial η2 = .55; performance, F(7, 217) = 19.66,
that the diversion of attention from the task of
p < .01, partial η2 = .39; effort, F(7, 217) = 47.99,
driving results in a degraded representation of
p < .01, partial η2 = .61; and frustration, F(7,
the driving environment (i.e., impaired situa-
217) = 26.06, p < .01, partial η2 = .40.
tion awareness of the driving context; Fisher
Physiological measures.  EEG was recorded
& Strayer, 2014; Gugerty, 1997; Kass, Cole, &
in Experiment 3 using the same protocol as that
Stanny, 2007). The data provide clear evidence
of the prior studies. The resulting ERPs from
that scanning the driving environment for poten-
single task are plotted in Figure 4 alongside the
tial hazards is an active process that is disrupted
same conditions from Experiments 1 and 2. The
by the diversion of attention to subsidiary in-
P300 component of the ERP was considerably
vehicle activities.
degraded by the added noise of the instrumented
vehicle, the added head and eye movements of
the driver as he or she scanned the driving envi- General Discussion
ronment, and the increased cognitive load of the The patterns observed in the three experi-
driving task (i.e., driving complexity increased ments reported are remarkably consistent, estab-
from the laboratory to the driving simulator to lishing that lessons learned in the laboratory and
the instrumented vehicle). Moreover, the P300 driving simulator are in good agreement with
became even less distinct at higher workloads studies of cognitive distraction on the road-
(e.g., while concurrently performing the OSPAN way. In each case, they document a systematic
task). As in Experiment 2, we were not able to increase in cognitive workload as participants
reliably measure P300 latency. P300 amplitude, performed different in-vehicle activities. The
plotted in Figure 6, was quantified by computing data for the three studies were entered into a
the average area under the curve between 400 MANOVA to determine how cognitive work-
and 700 ms. An ANOVA failed to show a main load changed across condition for the three
1316 December 2015 - Human Factors

experiments. For the sake of clarity, we focused NASA-TLX measures also increased from
our analyses based upon secondary, subjec- Experiment 1 to Experiment 3 for mental work-
tive, and physiological assessments because load, F(2, 99) = 5.50, p < .01, partial η2 = .10;
these measures were identical across the three physical workload, F(2, 99) = 8.34, p < .01, par-
experiments. tial η2 = .14; temporal demand, F(2, 99) = 8.38,
By using a head-mounted version of the DRT, p < .01, partial η2 = .14; effort, F(2, 99) = 5.04, p
the impact of head and eye movements on detec- < .01, partial η2 = .09; and frustration, F(2, 99) =
tion was negated. As drivers move their head, 7.98, p < .01, partial η2 = .13, but not for perfor-
the DRT device moved with them and remained mance (p > .14). On the whole, the subjective
in a constant location relative to the eyes. The workload measures were in agreement across
resulting RT and accuracy data provide a much six subscales, eight conditions, and three experi-
more finely calibrated metric than the more tra- ments. In particular, there was a consistent
ditional measures of brake RT and following increase in subjective workload ratings from
distance (which often covary, making unambig- Condition 1 to Condition 8 and also a systematic
uous interpretation difficult). A MANOVA per- increase in subjective workload ratings from
formed on the secondary-task DRT data revealed Experiment 1 to Experiment 3.
a significant effect of condition, F(14, 86) = A MANOVA performed on P300 amplitude
19.58, p < .01, partial η2 = .76, and experiment, revealed a main effect of condition, F(7, 93) =
F(4, 198) = 26.84, p < .01, partial η2 = .35, and a 6.67, p < .01, partial η2 = .33; experiment, F(2,
Condition × Experiment interaction, F(28, 99) = 13.3, p < .01, partial η2 = .21; and a Condi-
174) = 45.01, p < .01, partial η2 = .45. Further tion × Experiment interaction, F(14, 88) = 1.88,
univariate analysis showed a main effect of con- p < .05, partial η2 = .12. Overall, P300 amplitude
dition such that RT increased, F(7, 693) = 65.02, was similar in magnitude for single task, radio,
p < .01, partial η2 = .40, and A′ decreased, F(7, and audio book; smaller in magnitude for the
693) = 26.71, p < .01, partial η2 = .42, across conversation conditions (i.e., Conditions 4
condition. In addition, RT increased, F(2, 99) = through 7); and smallest for the OSPAN task.
85.14 p < .01, partial η2 = .63, and A′ decreased, P300 amplitude was largest in Experiment 1,
F(2, 99) = 35.78, p < .01, partial η2 = .42, from intermediate in magnitude for Experiment 2,
Experiment 1 to Experiment 3. On the whole, and smallest in Experiment 3, and this result
there is good agreement with the DRT measures undoubtedly reflects the degraded quality of the
across experiments; however, the laboratory- ERP signal as the experiments progressed from
and simulator-based studies would appear to the laboratory to the driving simulator to the
provide a more conservative estimate of the instrumented vehicle. In fact, the P300 ampli-
impairments to driving associated with in-­ tude was the noisiest of all the measures we
vehicle technology use. recorded, with contamination from movements
A MANOVA performed on the subjective of the mouth, jaw, eyes, head, and body accom-
workload ratings revealed a significant effect of panied with environmental noise from the simu-
condition, F(42, 58) = 26.48, p < .01, partial η2 = lator and instrumented vehicle. Consequently,
.95, and of experiment, F(12, 190) = 2.86, p < the P300 measures were the least sensitive of
.01, partial η2 = .15; however, the Condition × our measures to changes in cognitive workload,
Experiment interaction was not significant. and this limitation was most apparent in the
Across experiments, main effects of condition instrumented vehicle.
were obtained for mental workload, F(7, 693) = In the main, moving from the laboratory to
170.79, p < .01, partial η2 = .63; physical work- the driving simulator to the instrumented vehicle
load, F(7, 693) = 16.08, p < .01, partial η2 = .14; increased the intercept of the cognitive work-
temporal demand, F(7, 693) = 90.04, p < .01, load curves, and similar condition effects were
partial η2 = .48; performance, F(7, 693) = 44.99, obtained for the different dependent measures.
p < .01, partial η2 = .31; effort, F(7, 693) = This experimental cross-validation is important
140.92, p < .01, partial η2 = .59; and frustration, in its own right, establishing that the effects
F(7, 693) = 81.16, p < .01, partial η2 = .45. The obtained in the simulator generalize to on-road
Cognitive Distraction in the Automobile 1317

Effect Size Point Estimate and 95% Confidence Intervals

1.5

1
Effect Size

0.5

−0.5
Radio Book Passenger Hand−Held Hands−Free Speech−to−Text OSPAN

Figure 10. Effect size point estimates and 95% confidence intervals across all experiments. Effect size estimates
were computed as the difference from the single-task condition normalized by the pairwise standard deviation.
Note that the confidence intervals, plotted as “cat eyes,” are normally distributed around the respective point
estimates (which provides an estimate of the accuracy of the point estimate; cf. Cumming, 2014).

driving. In fact, our measures in Experiment 1 systematically with secondary-task condition. It

were remarkably consistent with those obtained is noteworthy that these secondary tasks allowed
in Experiment 3, suggesting that there are occa- the driver to have his or her eyes on the road and
sions when the added complexity, expense, and hands on the wheel (except when he or she was
risk of on-road study may be unnecessary. More- holding the cell phone). That is, these in-vehicle
over, the similarity of the primary, secondary, activities are cognitively distracting to different
subjective, and physiological measures provides degrees. Whereas the procedural demands of the
convergence in our workload assessments. tasks themselves did not require the driver to
However, these different measures were differ- divert his or her eyes from the roadway or to
entially sensitive to the manipulation of cogni- otherwise alter his or her scanning pattern,
­
tive load. For example, in the MANOVAs performing these cognitively demanding in-
­
reported earlier that correct for any collinearities vehicle activities clearly altered the visual scan-
in the subscales within each measure, the partial ning behavior of the driver in Experiment 3.
η2 values for the main effect of condition were The data from our studies also speak to the
greater for the NASA-TLX (partial η2 = .95) fidelity of the driving simulator in studying cog-
than for the DRT (partial η2 = .76) and the physi- nitive distraction. There have been some sugges-
ological measures (partial η2 = .33). tions that the patterns obtained in the driving
Figure 10 combines the standardized data simulator are not representative of the real world,
across the dependent measures and collapsed perhaps because the consequences of a crash in
across the experiments (see Steps 1 and 2 in the the simulator are different from the consequences
next section) to further document how the differ- of a crash on the roadway. Although we made no
ence from the single-task baseline, expressed as attempt to match the driving scenarios in the sim-
a pairwise effect size (i.e., Cohen’s d), grew ulator and instrumented vehicle, the patterns
1318 December 2015 - Human Factors

obtained in Experiments 2 and 3 were remark- and subjects. Table 1 lists the 13 dependent
ably similar. If anything, the data obtained in the measures that were used in the standardized
simulator may underestimate the impairments condition averages separated in gray by the
associated with using different in-vehicle activi- metric of which they belong.
ties on the road. It is noteworthy that Experiment Step 3: The standardized condition averages
1 provided a low-cost alternative to the driving across dependent measures were computed
simulator and instrumented vehicle, and the data with an equal weighting for physical, sec-
provided in this study were very predictive of ondary, subjective, and physiological met-
driving performance on the roadway (Lee, 2004). rics. The measures within each metric were
also equally weighted. For example, the
secondary-task workload metric was com-
Toward a Standardized Scale
posed of an equal weighting of the measures
of Cognitive Distraction
DRT-RT and DRT-A′ so that the two each
The primary goal of the current research was contributed one eighth to the overall work-
to develop a metric of cognitive distraction asso- load scale (i.e., 25% of the workload scale
ciated with performing different activities while is made up of secondary-task measures,
operating a motor vehicle. Because the differ- each of which is equally weighted). Simi-
ent dependent measures are on different scales larly, the six subjective workload measures
(e.g., milliseconds, meters, amplitude), each were equally weighted so that each of the
was transformed to a standardized score. This measures contributed one 24th to the overall
involved Z-transforming each of the dependent workload scale. Note that eye glances, A′,
measures to have a mean of 0 and a standard and P300 amplitude were inversely coded in
deviation of 1 (across the experiments and con- the summed condition averages.
ditions), and the average for each condition was Step 4: The standardized mean differences
then obtained. The standardized scores for each were range-corrected so that the nondis-
condition were then summed across the differ- tracted single-task condition had a rating of
ent dependent measures to provide an aggregate 1.0 and the OSPAN task had a rating of 5.0.
measure of cognitive distraction. Finally, the
aggregated standardized scores were linearly
X i = {[ X i − min ) / ( max − min )] * 4.0} + 1.
scaled such that the nondistracted single-task
driving condition anchored the low end (Cat-
egory 1) and the OSPAN task anchored the high The cognitive distraction scale presented in
end (Category 5) of the cognitive distraction Figure 11 ranges from 1.0 for the single-task
scale. For each of the other tasks, the relative condition and 5.0 for the OSPAN task. In-­vehicle
position compared to the low and high anchors activities, such as listening to the radio (1.21) or
provided an index of the cognitive workload an audio book (1.75), were associated with a
for that activity when concurrently performed small increase in cognitive distraction. The con-
while operating a motor vehicle. The four-step versation activities of talking to a passenger in
protocol for developing the cognitive distraction the vehicle (2.33) and conversing with a friend
scale is listed next. on a handheld (2.45) or hands-free cell phone
(2.27) were associated with a moderate increase
Step 1: For each dependent measure, the stan- in cognitive distraction, and the speech-to-text
dardized scores across experiments, condi- condition (3.06) had a large cognitive distraction
tions, and subjects were computed using rating.
Zi = (xi – X) / SD, where X refers to the over- The goal of this research was to be compre-
all mean and SD refers to the pooled stan- hensive, using a variety of driving environments,
dard deviation. a wide array of secondary task conditions, and
Step 2: For each dependent measure, the stan- an inclusive set of dependent measures. With the
dardized condition averages were com- standardized values for each dependent measure
puted by collapsing across experiments provided in Table 1, it is possible to use Steps 3
Cognitive Distraction in the Automobile 1319

Table 1: Standardized Scores for Each Dependent Measure

Single Audio Passenger Handheld Hands-Free Speech

Measure Task Radio Book Conversation Cell Phone Cell Phone to Text OSPAN

Brake reaction −.280 −.381 −.186 −.050 −.073 .015 .355 0.600
time (RT)
Following distance −.310 −.263 −.017 −.028 −.126 −.042 .243 0.544
Glances .527 .380 .036 −.108 −.092 −.036 −.239 −0.468
DRT-RT −.387 −.318 −.305 .161 .106 .071 .131 0.541
DRT-A′ .248 .195 .223 −.004 −.005 .024 −.072 −0.608
Mental −.640 −.558 −.145 −.289 −.162 −.171 .410 1.554
Physical −.321 −.289 −.136 −.138 .400 −.018 .158 0.345
Temporal −.495 −.486 −.220 −.277 −.070 −.172 .435 1.285
Performance −.454 −.419 −.061 −.101 −.070 −.056 .113 1.049
Effort −.599 −.488 −.184 −.326 −.112 −.184 .379 1.513
Frustration −.413 −.428 −.218 −.386 −.129 −.040 .278 1.337
P300 latency −.728 −.391 −.262 −.045 .237 .170 .233 0.788
P300 area −.011 .130 .139 −.254 .078 .258 .092 −0.431

Note. The primary task measures of brake reaction time and following distance were collected in Experiment 2,
glances at hazards were collected in Experiment 3, the secondary-task measures of DRT-RT and A′ were collected
in Experiments 1 through 3, the NASA Task Load Index subjective workload measures of mental workload, physical
workload, temporal demand, performance, effort, and frustration were collected in Experiments 1 through 3, and
the physiological measures of P300 latency and P300 area were collected in Experiment 1 and in Experiments 1
through 3, respectively. OSPAN = Operation Span; DRT = Detection Response Task.

and 4 to alter the contribution of dependent mea- instrument that can be used to evaluate different
sures and/or secondary-task conditions from the in-vehicle activities. Measuring cognitive distrac-
current study. For example, given the fact that tion has proven to be the most difficult of the three
the NASA-TLX and DRT measures are substan- sources of distraction to assess because of the
tially cheaper to collect and that they are very problems associated with observing what a driv-
sensitive to cognitive workload, future research- er’s brain (as opposed to hands or eyes) is doing.
ers may consider the utility of collecting only Here we used a combination of primary, second-
these measures and omitting the collection of ary, subjective, and physiological measures to
physiological measures. It is straightforward to assess cognitive distraction across a variety of in-
modify Steps 3 and 4 to exclude physiological vehicle activities and provide the most compre-
measures to see impact on the cognitive distrac- hensive analysis that has been performed to date.
tion scale. Moreover, provided with two com- The method used herein maps the different
mon anchor points (e.g., single-task driving and dependent measures into a unidimensional cog-
OSPAN), other investigators could easily extend nitive workload metric. Although multidimen-
the workload scale to an entirely different set of sional models of mental workload exist (e.g.,
driving conditions, secondary tasks, and depen- Wickens, 1984, 2002), there is nothing in the
dent measures. However, on a cautionary note, it current analysis that requires the more compli-
is inappropriate to post hoc “cherry-pick” depen- cated model. Indeed, the convergence of the dif-
dent measures to create a desired outcome (cf. ferent measures in the different driving environ-
Cumming, 2014) that is not representative of the ments suggests that the unidimensional model of
overall pattern in the data. cognitive workload is more parsimonious. We
The current research establishes an experi- also think it is useful to consider the current
mentally cross-validated cognitive distraction results in the context of contemporary models of
1320 December 2015 - Human Factors

Workload Point Estimate and 95% Confidence Intervals

6

5
Cognitive Distraction Scale

0
Single Radio Book Passenger Hand−Held Hands−Free Speech−to−Text OSPAN

Figure 11. Cognitive distraction scale. Circles indicate the point estimate for each condition, and the cat eyes
indicate the 95% confidence intervals normally distributed around the respective point estimates (cf. Cumming,
2014). The scale reflects the standardized averages for each condition, equally weighting physical, secondary,
subjective, and physiological measures across the three experiments (see Steps 1 through 4 for additional
detail).

cognitive neuroscience and multitasking (e.g., enacted laws prohibiting a driver’s use of a
Medeiros-Ward, Watson, & Strayer, in press). handheld cell phone, but all states allow the use
These models focus on the role that the prefron- of a hands-free device (Governors Highway
tal cortex–mediated executive attentional sys- Safety Association, 2014) even though the
tem plays in multitasking and how differences in research reported herein and elsewhere indicates
neural efficiency underlie individual differences that there is no difference in cognitive distrac-
in the ability to multitask while driving (e.g., tion between the two modes of cellular commu-
Watson & Strayer, 2010). nication (Ishigami & Klein, 2009). Finally, there
These findings can be used to help inform are in-vehicle activities, such as using a speech-
scientifically based policies on driver distrac- to-text system to send and receive text or e-mail
tion, particularly as it relates to cognitive dis- messages, which produced a relatively high
traction stemming from the diversion of atten- level of cognitive distraction.
tion to other concurrent activities in the vehicle. The speech-to-text-based system that we
For example, Grier et al. (2008) discussed how a evaluated in the current research used a perfect-
“red line” of workload could be empirically fidelity speech recognition system, and there
established and serve as a guide for policy mak- was no requirement to review, edit, or correct
ers. Here we found that some activities, such as garbled speech-to-text translations. Such is not
listening to the radio or an audio book, are not the case with current technology, but it is
very distracting. Other activities, such as con- improving and may someday achieve perfect
versing with a passenger or talking on a hand- fidelity. Given the current trends toward more
held or hands-free cell phone, are associated voice commands in the vehicle, this level of
with moderate/significant increases in cognitive cognitive distraction is troubling. The assump-
distraction. It is interesting that 14 states have tion that if the eyes are on the road and the hands
Cognitive Distraction in the Automobile 1321

are on the steering wheel, then voice-based (e.g., text messaging) is also likely to require cog-
interactions are safe appears to be unwarranted. nitive resources to read and process the message.
In the current research, conversation with a Even when there are no demands for visual pro-
passenger in the vehicle or with a friend over a cessing, interacting with cognitively demanding
cell phone was associated with similar levels of in-vehicle devices can alter how drivers scan and
cognitive distraction. In an earlier study compar- process information in the driving environment.
ing passenger and cell phone conversation In this article, we provide evidence for a rela-
(Drews, Pasupathi, & Strayer, 2008), the passen- tionship between secondary-task workload and
ger was allowed to spontaneously help the driver distracted driving. Given the capacity limits of
with the task of driving (e.g., helping to navigate, attention (e.g., Kahneman, 1973), an increase in
pointing out hazards, or regulating conversation cognitive workload associated with performing a
based upon the real-time demands of driving), concurrent secondary task reduces the available
and significant differences in route navigation attention that can be allocated to driving. How-
were observed. Compared to single-task driving, ever, as of yet, there has not been a comprehen-
when the participant was conversing with a sive mapping of cognitive distraction to on-road
friend seated next to him or her in the vehicle, crash risk. In experimental research that controls
there was no decline in navigation accuracy. By for exposure, it is reasonable to assume that there
contrast, a conversation with a friend on a hands- is a monotonic relationship between cognitive
free cell phone resulted in a 50% decline in navi- distraction and crash risk (the alternative would
gation accuracy (i.e., half the participants missed posit that increasing levels of distraction would
their exit). What is the basis for the discrepancy somehow make drivers safer). In studies that
between the current study and the study reported lack such experimental control, the potential for
by Drews and colleagues (2008)? One important self-regulatory factors to alter exposure patterns
difference is that the DRT device that was used to may complicate the relationship. For example,
measure RT and accuracy and served as a trigger drivers may proactively decide not to use a cell
for the ERP recordings was designed so that the phone when they are operating a motor vehicle.
driver could easily see the device; however, the Drivers may also reactively self-regulate by
passenger could not see the DRT lights and there- moderating their usage in real time based upon
fore could not adjust his or her conversation to driving difficulty or perception of driving errors.
aid the driver as he or she did when navigating to Reactive self-regulation may also involve trad-
a roadway exit. When the passenger cannot help ing off different aspects of driving performance
with the task of driving, as in the current study, when multitasking. Future research is needed to
then any differences between conversation types establish when a behavior is a form of adaptive
will be minimal. self-regulation or, instead, a by-product of the
diversion of attention from driving.
Caveats
Acknowledgments
The cognitive distraction scale provides a
comprehensive analysis of several of the cogni- This research was supported by the AAA Founda-
tive sources of driver distraction. The scale does tion for Traffic Safety.
not directly measure visual/manual sources of
distraction, although changes in visual scanning Key Points
associated with cognitive workload are certainly •• The objective of this research was to establish a
included in the metric. Although driver distrac- systematic framework for measuring and under-
tion can theoretically be separated into visual, standing cognitive distraction in the automobile.
manual, and cognitive sources, this sort of bal- •• The relationship between mental workload, cogni-
kanization may prove overly simplistic in the real tive distraction, and impaired driving was devel-
world. In Experiment 3, we demonstrated that oped.
there is cross-talk between cognitive and visual •• Converging measures provided a stable estimate
processing of potential hazards on the roadway. of driver distraction associated with different in-
Moreover, a task that has a high visual demand vehicle activities.
1322 December 2015 - Human Factors

•• The patterns observed in the three experiment set- Grier, R., Wickens, C. D., Kaber, D., Strayer, D. L., Boehm-Davis,
D., Trafton, G. J., & St. John, M. (2008). The red-line of
tings (laboratory control, driving simulator, and
workload: Theory, research, and design. In Proceedings of the
instrumented vehicle) were strikingly consistent, Human Factors and Ergonomics Society 52nd Annual Meet-
establishing that lessons learned in the laboratory ing (pp. 1204–1208). Santa Monica, CA: Human Factors and
and driving simulator are in good agreement with Ergonomics Society.
Gruen, S. (2006). Water for elephants: A novel. Chapel Hill, NC:
studies of cognitive distraction on the roadway.
Algonquin Books.
•• There are significant impairments to driving that Gugerty, L. (1997). Situation awareness during driving: Explicit
stem from the diversion of attention from the and implicit knowledge in dynamic spatial memory. Jour-
task of operating a motor vehicle that are directly nal of Experimental Psychology: Applied, 3, 42–66.
doi:10.1037/1076-898X.3.1.42
related to the cognitive workload of these in-vehi-
Harbluk, L., & Noy, I. (2002). The impact of cognitive distraction
cle activities. on driver visual behaviour and vehicle control. Ottawa, Can-
•• The data suggest that the use of voice-based tech- ada: Ergonomics Division, Road Safety Directorate and Motor
nology in the vehicle may have unintended conse- Vehicle Regulation Directorate. Retrieved from http://sunburst
.usd.edu/~schieber/materials/trans-canada-13889.pdf
quences that adversely affect traffic safety. Harbluk, J. L., Noy, Y. I., Trbovich, P. L., & Eizenman, M. (2007). An
on-road assessment of cognitive distraction: Impacts on drivers’
References visual behavior and braking performance. Accident Analysis &
Angell, L., Auflick, J., Austria, P. A., Kochhar, D., Tijerina, L., Prevention, 39, 372–379. doi:10.1016/j.aap.2006.08.013
Biever, W., &  . . . Kiger, S. (2006). Driver workload metrics Hart, S. G., & Staveland, L. E. (1988). Development of NASA-
task 2 final report (DOT HS 810 635). Washington, DC: TLX (Task Load Index): Results of empirical and theoretical
National Highway Traffic Safety Administration. research. In P. A. Hancock & N. Meshkati (Eds.), Human men-
Caird, J. K., Willness, C. R., Steel, P., & Scialfa, C. (2008). A meta- tal workload (pp. 139–183). Oxford, UK: North-Holland.
analysis of the effects of cell phones on driver performance. Horrey, W. J., & Wickens, C. D. (2006). Examining the
Accident Analysis & Prevention, 40, 1282–1293. doi:10.1016/j. impact of cell phone conversations on driving using
aap.2008.01.009 meta-analytic techniques. Human Factors, 48, 196–205.
Cooper, J. M., Medeiros-Ward, N., & Strayer, D. L. (2013). The doi:10.1518/001872006776412135
impact of eye movements and cognitive workload on lateral Ishigami, Y., & Klein, R. (2009). Is a hands-free phone safer than
position variability. Human Factors, 55, 1001–1014. a handheld phone? Journal of Safety Research, 40, 157–164.
Cumming, G. (2014). The new statistic: Why and how. Psychologi- doi:10.1016/j.jsr.2009.02.006
cal Science, 25, 7–29. Ishihara, S. (1993). Ishihara’s test for color-blindness. Tokyo,
Dingus, T. A., Klauer, S. G., Neale, V. L., Petersen, A., Lee, S. E., Japan: Kanehara.
Sudweeks, J., & Knipling, R. R. (2006). The 100-Car Naturalistic International Standards Organization. (2012). Road vehicles:
Driving Study: Phase II. Results of the 100-Car Field Experiment Transport information and control systems. Detection-
(DOT HS 810 593). Washington, DC: National Highway Traf- Response Task (DRT) for assessing selective attention in
fic Safety Administration. Retrieved from http://www.nhtsa.gov/ driving (ISO TC 22 SC 13 N17488). Working draft under
Research/Human+Factors/Naturalistic+driving+studies: development by Working Group 8 of ISO TC22, SC 13.
Donchin, E. (1981). Presidential address, 1980: Surprise! . . . Sur- Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ:
prise? Psychophysiology, 18, 493–513. Prentice Hall.
Drews, F. A., Pasupathi, M., & Strayer, D. L. (2008). Passenger and Kass, S. J., Cole, K. S., & Stanny, C. J. (2007). Effects of distraction
cell-phone conversation during simulated driving. Journal of and experience on situation awareness and simulated driving.
Experimental Psychology: Applied, 14, 392–400. doi:10.1037/ Transportation Research Part F, 10, 321–329. doi:10.1016/j.
a0013119 trf.2006.12.002
Engström, J., Johansson, E., & Östlund, J. (2005). Effects of visual Kelley, J. F. (1983). An empirical methodology for writ-
and cognitive load in real and simulated motorway driving. ing user-friendly natural language computer applications.
Transportation Research Part F, 8, 97–120. doi:10.1016/j. In Proceedings of ACM SIG-CHI ’83 Human Factors in
trf.2005.04.012 Computing Systems (pp. 193–196). New York, NY: ACM.
Engström, J., Monk, C. A., Hanowski, R. J., Horrey, W. J., Lee, J. doi:10.1145/800045.801609
D., McGehee, D. V., &  . . . Yang, C.Y.D. (2013). A conceptual Klauer, S. G., Guo, F., Simons-Morton, B. G., Ouiment, M. C.,
framework and taxonomy for understanding and categorizing Lee, S. E., & Dingus, T. A. (2014). Distracted driving and risk
driver inattention. Brussels, Belgium: European Commission. of road crashes among novice and experienced drivers. New
Fisher, D. L., & Strayer, D. L. (2014). Modeling situation aware- England Journal of Medicine, 370, 54–59.
ness and crash risk. Annals of Advances in Automotive Medi- Kramer, A. F., Sirevaag, E. J., & Braun, R. (1987). A psycho-
cine, 5, 33–39. physiological assessment of operator workload during
Gopher, D., & Donchin, E. (1986). Workload: An experimentation simulated flight missions. Human Factors, 29, 145–160.
of the concept. In K. Boff, L. Kaufmann, & J. Thomas (Eds.), doi:10.1177/001872088702900203
Handbook of perception and performance: Vol. 2. Cognitive Lee, J. D. (2004, September). Simulator fidelity: How low can you
processes and performance (pp. 1–49). New York, NY: Wiley. go? Paper presented at the Human Factors and Ergonomics
Governors Highway Safety Association. (2014). Distracted driving Society 48th Annual Meeting, New Orleans, LA.
laws. Retrieved from http://www.ghsa.org/html/stateinfo/laws/ Lee, J. D., Caven, B., Haake, S., & Brown, T. L. (2001). Speech-
cellphone_laws.html based interactions with in-vehicle computers: The effect of
Cognitive Distraction in the Automobile 1323

speech-based e-mail on drivers’ attention and roadway. Human across three age groups. Human Factors, 54, 454–468.
Factors, 43, 631–640. doi:10.1518/001872001775870340 doi:10.1177/0018720812437274
Lowry, L. (2002). The giver. New York, NY: Laurel Leaf Books. Rowling, J. K. (1998). Harry Potter and the sorcerer’s stone. New
McCarley, J. S., Vais, M., Pringle, H., Kramer, A. F., Irwin, D. E., York, NY: Arthur A. Levine Books.
& Strayer, D. L. (2004). Conversation disrupts scanning and Sanbonmatsu, D. M., Strayer, D. L., Medeiros-Ward, N., &
change detection in complex visual scenes. Human Factors, Watson, J. M. (2013). Who multi-tasks and why? Multi-­
­
46, 424–436. doi:10.1518/hfes.46.3.424.50394 tasking ability, perceived multi-tasking ability, impulsiv-
McEvoy, S. P., Stevenson, M. R., McCartt, A. T., Woodward, M., ity, and sensation seeking. PLOS ONE, 8, 1–8. doi:10.1371/
Haworth, C., Palamara, P., & Cercarelli, R. (2005). Role of mobile journal.pone.0054402
phones in motor vehicle crashes resulting in hospital attendance: Semlitsch, H. V., Anderer, P., Schuster, P., & Presslich, O. (1986).
A case-crossover study. British Medical Journal, 331, 428–433. A solution for reliable and valid reduction of ocular artifacts,
Medeiros-Ward, N., Cooper, J. M., & Strayer, D. L. (2013). Hierar- applied to the P300 ERP. Psychophysiology, 23, 695–703.
chical control and driving. Journal of Experimental Psychology: doi:10.1111/j.1469-8986.1986.tb00696.x
General. Advance online publication. doi:10.1037/a0035097 Sirevaag, E. J., Kramer, A. F., Wickens, C. D., Reisweber, M.,
Medeiros-Ward, N., Watson, J. M., & Strayer, D. L. (in press). On Strayer, D. L., & Grenell, J. H. (1993). Assessment of pilot
supertaskers and the neural basis of extraordinary multitasking. performance and mental workload in rotary wing aircraft.
Psychonomic Bulletin and Review. Ergonomics, 9, 1121–1140. doi:10.1080/00140139308967983
Mehler, B., Reimer, B., & Coughlin, J. F. (2012). Sensitivity of Sodhi, M., Reimer, B., & Llamazares, I. (2002). Glance analysis
physiological measures for detecting systematic variations in of driver eye movements to evaluate distraction. Behavior
cognitive demand from a working memory task: An on-road Research Methods, Instruments & Computers, 34, 529–538.
study across three age groups. Human Factors, 54, 396–412. doi:10.3758/BF03195482
doi:10.1177/0018720812442086 Strayer, D. L., Cooper, J. M., & Drews, F. A. (2004). What do driv-
National Highway Traffic Safety Administration. (2012). Visual- ers fail to see when conversing on a cell phone? In Proceedings
manual NHTSA driver distraction guidelines for in-vehicle of the Human Factors and Ergonomics Society 48th Annual
electronic devices: Department of Transportation (Docket Meeting (pp. 2213–2217). Santa Monica, CA: Human Factors
No. NHTSA-2010-0053). Retrieved from https://www. and Ergonomics Society.
federalregister.gov/articles/2012/02/24/2012-4017/visual- Strayer, D. L., & Drews, F. A. (2007a). Cell-phone-induced driver
manual-nhtsa-driver-distraction-guidelines-for-in-vehicle- distraction. Current Directions in Psychological Science, 16,
electronic-devices 128–131. doi:10.1111/j.1467-8721.2007.00489.
Parasuraman, R., & Davies, D. R. (1984). Varieties of attention. Strayer, D. L., & Drews, F. A. (2007b). Multi-tasking in the auto-
San Francisco, CA: Academic Press. mobile. In A. Kramer, D. Wiegmann, & A. Kirlik (Eds.),
Pollack, I., & Norman, D. A. (1964). Non-parametric analysis of Applied attention: From theory to practice (pp. 121–133).
recognition experiments. Psychonomic Science, 1, 125–126. Oxford, UK: Oxford University Press.
Ranney, T., Mazzae, E., Garrott, R., & Goodman, M. (2000). Strayer, D. L., Drews, F. A., & Johnston, W. A. (2003). Cell phone
NHTSA driver distraction research: Past, present and future. induced failures of visual attention during simulated driv-
Retrieved from http://ntl.bts.gov/ref/biblio/DistractedDriving/ ing. Journal of Experimental Psychology: Applied, 9, 23–32.
Distracted.htm doi:10.1037/1076-898X.9.1.23
Ratcliff, R., & Strayer, D. L. (2014). Modeling simulated driving Strayer, D. L., & Johnston, W. A. (2001). Driven to distraction:
with a one-boundary diffusion model. Psychological Bulletin Dual-task studies of simulated driving and conversing on a cel-
and Review, 21, 577–589. lular phone. Psychological Science, 12, 462–466.
Recarte, M. A., & Nunes, L. M. (2000). Effects of verbal Strayer, D. L., Watson, J. M., & Drews, F. A. (2011). Cognitive
and spatial-imagery tasks on eye fixations while driving. distraction while multitasking in the automobile. In B. Ross
Journal of Experimental Psychology: Applied, 6, 31–43. (Ed.), The psychology of learning and motivation: Advances
doi:10.1037/1076-898X.6.1.31 in research and theory (Vol. 54, pp. 29–58). San Diego, CA:
Recarte, M. A., & Nunes, L. M. (2003). Mental workload while Elsevier Academic Press.
driving: Effects on visual search, discrimination, and decision Stutts, J., Feaganes, J., Rodman, E., Hamlet, C., Meadows, T.,
making. Journal of Experimental Psychology: Applied, 9, 119– Rinfurt, D.,  . . . Staplin, L. (2003). Distractions in everyday
137. doi:10.1037/1076-898X.9.2.119 driving. Washington, DC: AAA Foundation for Traffic Safety.
Redelmeier, D. A., & Tibshirani, R. J. (1997). Association between Retrieved from https://www.aaafoundation.org/distractions-
cellular-telephone calls and motor vehicle collisions. New Eng- everyday-driving-0
land Journal of Medicine, 336, 453–458. Sussman, E. D., Bishop, H., Madnick, B., & Walker, R. (1985).
Reimer, B. (2009). Impact of cognitive task complexity on driv- Driver inattention and highway safety. Transportation
ers’ visual tunneling. Transportation Research Record: Jour- Research Record, 1047, 40–48.
nal of the Transportation Research Board, 2138, 13–19. Taylor, T., Pradhan, A. K., Divekar, G., Romoser, M., Muttart, J.,
doi:10.3141/2138-03 Gomes, R.,  . . . Fisher, D. L. (2013). The view from the road:
Reimer, B., Mehler, B. L., Pohlmeyer, A. E., Coughlin, J. F., & The contribution of on-road glance-monitoring technologies to
Dusek, J. A. (2006). The use of heart rate in a driving simulator understanding driver behavior. Accident Analysis & Preven-
as an indicator of age-related differences in driver workload. tion, 58, 175–186. doi:10.1016/j.aap.2013.02.008
Advances in Transportation Studies an International Journal, Turner, M. L., & Engle, R. W. (1989). Is working memory capac-
Special Issue, 9–20. ity task dependent? Journal of Memory & Language, 28, 127–
Reimer, B., Mehler, B. L., Wang, Y., & Coughlin, J. F. (2012). A 154. doi:10.1016/0749-596X(89)90040-5
field study on the impact of variations in short term memory Wang, J. S., Knipling, R. R., & Goodman, M. J. (1996). The role
demands on drivers’ visual attention and driving performance of driver inattention in crashes: New statistics from the 1995
1324 December 2015 - Human Factors

Crashworthiness Data System. In 40th Annual Proceedings of Joel M. Cooper is the owner/CEO at Precision Driv-
the Association for the Advancement of Automotive M ­ edicine
ing Research. He received a PhD in cognition and
(pp. 377–392). Retrieved from http://www-nrd.nhtsa.dot.
gov/departments/Human%20Factors/driver-distraction/PDF/ neural science in 2010 from the University of Utah.
Wang.PDF
Watson, J. M., & Strayer, D. L. (2010). Profiles in extraordinary James R. Coleman is a PhD candidate at the Univer-
multitasking ability. Psychonomic Bulletin and Review, 17, sity of Utah. He received an MS in cognition and
479–485. doi:10.3758/PBR.17.4.479
Wickens, C. D. (1984). Processing resources in attention. In R. neural science from the University of Utah in 2014.
Parasuraman, J. Beatty, & J. Davies (Eds.), Varieties of atten-
tion (pp. 63–102). New York, NY: Academic Press. Nathan Medeiros-Ward is a post-doctoral fellow at
Wickens, C. D. (2002) Multiple resources and performance the Beckman Institute for Advanced Science and
prediction. Theoretical Issues in Ergonomics Science, 3,
159–177.
Technology, University of Illinois. He received a
PhD in cognition and neural science in 2013 from
David L. Strayer is a professor at the University of the University of Utah.
Utah. He received a PhD in experimental psychol-
Francesco Biondi is a post-doctoral fellow at the
ogy in 1989 from the University of Illinois at
University of Utah. He received a PhD in experi-
Urbana-Champaign.
mental psychology and cognitive sciences in 2015
from the University of Padova, Italy.
Jonna Turrill is a PhD candidate at the University of
Utah. She received an MS in cognition and neural Date received: September 9, 2013
science from the University of Utah in 2014. Date accepted: January 25, 2015