A Comparison of the Cell Phone Driver and the Drunk Driver

A Comparison of the Cell Phone Driver and the Drunk Driver
David L. Strayer, Frank A. Drews, and Dennis J. Crouch, University of Utah, Salt Lake
City, Utah
Objective: The objective of this research was to determine the relative impairment
associated with conversing on a cellular telephone while driving. Background:
Epidemiological evidence suggests that the relative risk of being in a traf?c accident
while using a cell phone is similar to the hazard associated with driving with a blood
alcohol level at the legal limit. The purpose of this research was to provide a direct
comparison of the driving performance of a cell phone driver and a drunk driver in
a controlled laboratory setting. Method: We used a high-?delity driving simulator
to compare the performance of cell phone drivers with drivers who were intoxicated
from ethanol (i.e., blood alcohol concentration at 0.08% weight/volume). Results:
When drivers were conversing on either a handheld or hands-free cell phone, their
braking reactions were delayed and they were involved in more traf?c accidents than
when they were not conversing on a cell phone. By contrast, when drivers were intox-
icated from ethanol they exhibited a more aggressive driving style, following closer
to the vehicle immediately in front of them and applying more force while braking.
Conclusion: When driving conditions and time on task were controlled for, the im-
pairments associated with using a cell phone while driving can be as profound as
those associated with driving while drunk. Application: This research may help to
provide guidance for regulation addressing driver distraction caused by cell phone
conversations.
INTRODUCTION
al., 2003). However, over the last decade many
new electronic devices have been developed, and
Although they are often reminded to pay full
they are making their way into the vehicle. In
attention to driving, people regularly engage in a
many cases, these new technologies are engag-
wide variety of multitasking activities when they
ing, interactive information delivery systems. For
are behind the wheel. Indeed, data from the 2000
example, drivers can now surf the Internet, send
U.S. census indicates that drivers spend an aver-
and receive E-mail or faxes, communicate via a
age of 25.5 min each day commuting to work,
cellular device, and even watch television. There
and there is a growing interest in trying to make
is good reason to believe that some of these new
the time spent on the roadway more productive
multitasking activities may be substantially more
(Reschovsky, 2004). Unfortunately, because of
distracting than the old standards because they
the inherent limited capacity of human attention
are more cognitively engaging and because they
(e.g., Kahneman, 1973; Navon & Gopher, 1979),
are performed over longer periods of time.
engaging in these multitasking activities often
The current research focuses on a dual-task
comes at a cost of diverting attention away from
activity that is commonly engaged in by more
the primary task of driving. There are a number
than 100 million drivers in the United States: the
of more traditional sources of driver distraction.
concurrent use of cell phones while driving (Cel-
These “old standards” include talking to passen-
lular Telecommunications Industry Association,
gers, eating, drinking, lighting a cigarette, apply-
2006; Goodman et al., 1999). Indeed, the National
ing makeup, and listening to the radio (Stutts et
Highway Transportation Safety Administration
Address correspondence to David L. Strayer, Department of Psychology, 380 South, 1530 East, RM 502, University of Utah,
Salt Lake City, UT 84112-0251; david.strayer@utah.edu. HUMAN FACTORS, Vol. 48, No. 2, Summer 2006, pp. 381–391.
Copyright © 2006, Human Factors and Ergonomics Society. All rights reserved.

---

382
Summer 2006 – Human Factors
estimated that 8% of drivers on the roadway at
use their cell phone while driving may be more
any given daylight moment are using their cell
likely to engage in risky behavior, and this in-
phone (Glassbrenner, 2005). It is now well estab-
crease in risk taking may be the cause of the cor-
lished that cell phone use impairs the driving per-
relation. It may also be the case that being in an
formance of younger adults (Alm & Nilsson,
emotional state may increase one’s likelihood of
1995; Briem & Hedman, 1995; Brookhuis, De
driving erratically and may also increase the like-
Vries, & De Waard, 1991; I. D. Brown, Tickner, &
lihood of talking on a cell phone. Finally, limita-
Simmonds,1969; Goodman et al.,1999; McKnight
tions on establishing an exact time of the accident
& McKnight, 1993; Redelmeier & Tibshirani,
lead to uncertainty regarding the precise rela-
1997; Strayer, Drews, & Johnston, 2003; Strayer
tionship between talking on a cell phone while
& Johnston, 2001). For example, drivers are
driving and increased traf?c accidents.
more likely to miss critical traf?c signals (traf?c
If the relative risk estimates of Redelmeier and
lights, a vehicle braking in front of the driver,
Tibshirani (1997) can be substantiated in a con-
etc.), slower to respond to the signals that they do
trolled laboratory experiment and there is a
detect, and more likely to be involved in rear-end
causal link between cell phone use and impaired
collisions when they are conversing on a cell
driving, then these data would be of immense
phone (Strayer et al., 2003). In addition, even
importance for public safety and legislative bod-
when participants direct their gaze at objects in
ies. Here we report the result of a controlled study
the driving environment, they often fail to “see”
that directly compared the performance of driv-
them when they are talking on a cell phone be-
ers who were conversing on either a handheld
cause attention has been directed away from the
or hands-free cell phone with the performance
external environment and toward an internal,
of drivers with a blood alcohol concentration
cognitive context associated with the phone con-
at 0.08% weight/volume (wt/vol). Alcohol has
versation. However, what is lacking in the litera-
been used as a benchmark for assessing perfor-
ture is a clear benchmark with which to evaluate
mance impairments in a variety of other areas,
the relative risks associated with this dual-task
including aviation (Billings, Demosthenes, White,
activity (e.g., Brookhuis, 2003).
& O’Hara, 1991; Klein, 1972), anesthesiology
In their seminal article, Redelmeier and Tib-
(Thapar, Zacny, Choi,& Apfelbaum,1995; Tiplady,
shirani (1997) reported epidemiological evidence
1991) nonprescription drug use (Burns & Mos-
suggesting that “the relative risk [of being in a
kovitz, 1980), and fatigue (Williamson, Feyer,
traf?c accident while using a cell phone] is sim-
Friswel,& Finlay-Brown,2001). Indeed, the World
ilar to the hazard associated with driving with a
Health Organization recommended that the be-
blood alcohol level at the legal limit” (p. 456).
havioral effects of drugs be compared with those
These estimates were made by evaluating the cel-
of alcohol under the assumption that performance
lular records of 699 individuals involved in motor
on drugs should be no worse than that at the legal
vehicle accidents. It was found that 24% of these
blood alcohol limit (Willette & Walsh, 1983).
individuals were using their cell phone within the
We used a car-following paradigm (see also
10-min period preceding the accident, and this
Alm & Nilsson, 1995; Lee, Vaven, Haake, &
was associated with a fourfold increase in the
Brown, 2001; Strayer et al., 2003) in which par-
likelihood of getting into an accident. Moreover,
ticipants drove on a multilane freeway following
these authors suggested that the interference
a pace car that would brake at random intervals.
associated with cell phone use was attributable to
We measured a number of performance variables
attentional factors rather than to peripheral fac-
(e.g., driving speed, following distance, brake re-
tors such as holding the phone. However, there
action time, time to collision) that have been
are several limitations to this important study.
shown to affect the likelihood and severity of
First, although the study established a strong
rear-end collisions, the most common type of
association between cell phone use and motor
traf?c accident reported to police (T. L. Brown,
vehicle accidents, it did not demonstrate a causal
Lee, & McGehee, 2001; Lee et al., 2001). Three
link between cell phone use and increased accident
counterbalanced conditions were studied using a
rates. For example, there may be self-selection
within-subjects design: single-task driving (base-
factors underlying the association: People who
line condition), driving while conversing on a

---

CELL PHONE DRIVERS AND DRUNK DRIVERS
383
cell phone (cell phone condition), and driving
gender were signi?cant in the current sample. Ad-
with a blood alcohol concentration of 0.08% wt/
ditional analyses comparing the driving perfor-
vol (alcohol condition). The driving tasks were
mance of participants who owned a cell phone with
performed on a high-?delity driving simulator.
that of those who did not own a cell phone failed
to ?nd any signi?cant differences (all ps > .60).
METHOD
Similarly, there was no signi?cant difference in
driving performance between participants who
Participants
reported that they used a cell phone while driv-
Forty adults (25 men, 15 women), recruited
ing and those who did not use a cell phone while
via advertisements in local newspapers, partici-
driving (all ps >.70).
pated in the Institutional Review Board approved
study. Participants ranged in age from 22 to 34
Stimuli and Apparatus
years, with an average age of 25 years. All had
A PatrolSim high-?delity driving simulator,
normal or corrected-to-normal vision and a valid
illustrated in Figure 1 and manufactured by GE-
driver’s license with an average of 8 years of
ISIM, was used in the study. The simulator is com-
driving experience. Of the 40 participants, 78%
posed of five networked microprocessors and
owned a cell phone, and 87% of the cell phone
three high-resolution displays providing a 180°
owners reported that they have used a cell phone
field of view. The dashboard instrumentation,
while driving. A further requirement for inclusion
steering wheel, gas pedal, and brake pedal are
in the study was that participants were social
from a Ford Crown Victoria® sedan with an auto-
drinkers, consuming between three and ?ve alco-
matic transmission. The simulator incorporates
holic drinks per week. The experiment lasted
proprietary vehicle dynamics, traffic scenario,
approximately 10 hr (across the three days of the
and road surface software to provide realistic
study), and participants were remunerated at a
scenes and traf?c conditions.
rate of $10/hr.
A freeway road database simulated a 24-mile
A preliminary comparison of male and female
(38.6-km) multilane interstate with on- and off-
drivers found greater variability in following dis-
ramps, overpasses, and two- or three-lane traf?c
tance for female drivers, F(1, 38) = 10.9, p < .01;
in each direction. Daytime driving conditions with
however, this gender effect was not modulated by
good visibility and dry pavement were used. A
alcohol or cell phone use. No other effects of
pace car, programmed to travel in the right-hand
Figure 1. A participant talking on a cell phone while driving in the GE-ISIM driving simulator.

---

384
Summer 2006 – Human Factors
Figure 2. An example of the sequence of events occurring in the car following paradigm.
lane, braked intermittently throughout the sce-
car, the pace car released its brake and accelerated
nario. Distractor vehicles were programmed to
to normal highway speed. If the participant failed
drive between 5% and 10% faster than the pace
to depress the brake, he or she would eventually
car in the left lane, providing the impression of a
collide with the pace car. That is, as in real high-
steady ?ow of traf?c. Unique driving scenarios,
way stop-and-go traffic, the participant was
counterbalanced across participants, were used
required to react in a timely and appropriate man-
for each condition in the study. Measures of real-
ner to a vehicle slowing in front of them.
time driving performance, including driving
Figure 2 presents a typical sequence of events
speed, distance from other vehicles, and brake
in the car-following paradigm. Initially both the
inputs, were sampled at 30 Hz and stored for later
participant’s car (solid line) and the pace car (long-
analysis. Cellular service was provided by Sprint
dashed line) were driving at about 62 miles/hr
PCS. The cell phone was manufactured by LG
(mph) with a following distance of 40 m (dotted
Electronics Inc. (Model TP1100). For hands-free
line). At some point in the sequence, the pace
conditions, a Plantronics M135 headset (with
car’s brake lights illuminated for 750 ms (short-
earpiece and boom microphone) was attached to
dashed line) and the pace car began to decelerate
the cell phone. Blood alcohol concentration levels
at a steady rate. As the pace car decelerated, fol-
were measured using an Intoxilyzer 5000, man-
lowing distance decreased. At a later point in time,
ufactured by CMI Inc.
the participant responded to the decelerating pace
car by pressing the brake pedal. The time interval
Procedure
between the onset of the pace car’s brake lights
The experiment used a within-subjects design
and the onset of the participant’s brake response
and was conducted in three sessions on different
de?nes the brake onset time. Once the participant
days. The ?rst session familiarized participants
depressed the brake, the pace car began to accel-
with the driving simulator using a standardized
erate, at which point the participant removed his
adaptation sequence. The order of subsequent
or her foot from the brake and applied pressure
alcohol and cell phone sessions was counterbal-
to the gas pedal. Note that in this example, follow-
anced across participants. In these latter sessions,
ing distance decreased by about 50% during the
the participant’s task was to follow the intermit-
braking event.
tently braking pace car driving in the right-hand
In the alcohol session, participants drank a mix-
lane of the highway. When the participant stepped
ture of orange juice and vodka (40% alcohol by
on the brake pedal in response to the braking pace
volume) calculated to achieve a blood alcohol

---

CELL PHONE DRIVERS AND DRUNK DRIVERS
385
concentration of 0.08% wt/vol. Blood alcohol
extracting 10-s epochs of driving performance that
concentrations were veri?ed using infrared spec-
were time locked to the onset of the pace car’s
trometry breath analysis immediately before and
brake lights. That is, each time that the pace car’s
after the alcohol driving condition. Participants
brake lights were illuminated, the data for the en-
drove in the 15-min car-following scenario while
suing 10 s were extracted and entered into a 32 ×
legally intoxicated. Average blood alcohol con-
300 data matrix (i.e., on the jth occasion that the
centration before driving was 0.081% wt/vol and
pace car brake lights were illuminated, data from
after driving was 0.078% wt/vol.
the 1st, 2nd, 3rd, …, and 300th observations fol-
In the cell phone session, three counterbal-
lowing the onset of the pace car’s brake lights were
anced conditions, each 15 min in duration, were in-
entered into the matrix X[j,1], X[j,2], X[j,3],...X[j,300] ,
cluded: single-task baseline driving, driving while
in which j ranges from 1 to 32 re?ecting the 32
conversing on a handheld cell phone, and driving
occasions in which the participant reacted to the
while conversing on a hands-free cell phone. In
braking pace car). Each driving pro?le was creat-
both cell phone conditions, the participant and a
ed by averaging across j for each of the 300 time
research assistant engaged in naturalistic conver-
points. We created profiles of the participant’s
sations on topics that were identi?ed on the ?rst
braking response, driving speed, and following
day as being of interest to the participant. As would
distance.
be expected with any naturalistic conversation,
Figure 3 presents the braking pro?les. In the
they were unique to each participant. The task of
baseline condition, participants began braking
the research assistant in our study was to main-
within 1 s of pace car deceleration. Similar brak-
tain a dialog in which the participant listened and
ing pro?les were obtained for both the cell phone
spoke in approximately equal proportions. How-
and alcohol conditions. However, compared with
ever, given that our cell phone conversations were
baseline, when participants were intoxicated they
casual, they probably underestimate the impact
tended to brake with greater force, whereas par-
of intense business negotiations or other emo-
ticipants’ reactions were slower when they were
tional conversations conducted over the phone.
conversing on a cell phone.
To minimize interference from manual compo-
Figure 4 presents the driving speed pro?les. In
nents of cell phone use, the call was initiated
the baseline condition, participants began decel-
before participants began driving.
erating within 1 s of the onset of the pace car’s
brake lights, reaching minimum speed 2 s after
RESULTS
the pace car began to decelerate, whereupon par-
ticipants began a gradual return to prebraking
In order to better understand the differences
driving speed. When participants were intoxicat-
between conditions, we created driving pro?les by
ed they drove slower, but the shape of the speed
Braking Pro?le
Speed Pro?le
Figure 3. The braking pro?le.
Figure 4. The speed pro?le.

---

386
Summer 2006 – Human Factors
pro?le did not differ from baseline. By contrast,
remaining until a collision between the partici-
when participants were conversing on a cell
pant’s vehicle and the pace car if the course and
phone it took them longer to recover their speed
speed were maintained (i.e., had the participant
following braking.
failed to brake). Also reported are the frequency
Figure 5 presents the following distance
of trials with TTC values below 4 s, a level found
profiles. In the baseline condition participants
to discriminate between cases in which the drivers
followed approximately 28 m behind the pace
?nd themselves in dangerous situations and those
car, and as the pace car decelerated the following
in which the driver remains in control of the vehi-
distance decreased, reaching nadir approximately
cle (e.g., Hirst & Graham, 1997). Half recovery
2 s after the onset of the pace car’s brake lights.
time is the time for participants to recover 50%
When participants were intoxicated, they followed
of the speed that was lost during braking (e.g., if
closer to the pace car, whereas participants in-
the participant’s car was traveling at 60 mph [96.5
creased their following distance when they were
km/hr] before braking and decelerated to 40 mph
conversing on a cell phone.
[64.4 km/hr] after braking, then half recovery
Table 1 presents the nine performance vari-
time would be the time taken for the participant’s
ables that were measured to determine how par-
vehicle to return to 50 mph [80.4 km/hr]). Also
ticipants reacted to the vehicle braking in front of
shown in the table is the total number of collisions
them. Brake reaction time is the time interval be-
in each phase of the study. We used a multivariate
tween the onset of the pace car’s brake lights and
analysis of variance (MANOVA) followed by
the onset of the participant’s braking response
planned contrasts (shown in Table 2) to provide
(i.e., de?ned as a minimum of 1% depression of
an overall assessment of driver performance in
the participant’s brake pedal). Maximum braking
each of the experimental conditions.
force is the maximum force that the participant
We performed an initial comparison of partic-
applied to the brake pedal in response to the brak-
ipants driving while using a handheld cell phone
ing pace car (expressed as a percentage of maxi-
versus a hands-free cell phone. Both handheld
mum). Speed is the average driving speed of the
and hands-free cell phone conversations impaired
participant’s vehicle (expressed in miles per
driving. However, there were no signi?cant dif-
hour). Mean following distance is the distance
ferences in the impairments caused by these two
prior to braking between the rear bumper of the
modes of cellular communication (all ps > .25).
pace car and the front bumper of the participant’s
Therefore, we collapsed across the handheld and
car. SD following distance is the standard devia-
hands-free conditions for all subsequent analyses
tion of following distance.
reported in this article. The observed similarity be-
Time to collision (TTC), measured at the onset
tween handheld and hands-free cell phone conver-
of the participant’s braking response, is the time
sations is consistent with earlier work (e.g., Patten,
Kircher, Ostlund, & Nilsson, 2004; Redelmeier
& Tibshirani, 1997; Strayer & Johnston, 2001)
Following Distance
and calls into question driving regulations that
prohibit handheld cell phones and permit hands-
free cell phones.
MANOVAs indicated that both cell phone and
alcohol conditions differed significantly from
baseline, F(8, 32) = 6.26, p < .01, and F(8, 32) =
2.73, p < .05, respectively. When drivers were
conversing on a cell phone, they were involved
in more rear-end collisions, their initial reaction
to vehicles braking in front of them was slowed
by 9%, and the variability in following distance
increased by 24%, relative to baseline. In addition,
compared with baseline, participants who were
talking on a cell phone took 19% longer to recov-
Figure 5. The following distance pro?le.
er the speed that was lost during braking.

---

CELL PHONE DRIVERS AND DRUNK DRIVERS
387
TABLE 1: Means and Standard Errors (in Parentheses) for the Alcohol, Baseline,
and Cell Phone Conditions
Alcohol
Baseline
Cell Phone
Total accidents
0
0
3
Brake reaction time (ms)
779 (33)
777 (33)
849 (36)
Maximum braking force
69.8 (3.7)
56.7 (2.6)
55.5 (3.0)
Speed (mph)
52.8 (2.0)
55.5 (0.7)
53.8 (1.3)
Mean following distance (m)
26.0 (1.7)
27.4 (1.3)
28.4 (1.7)
SD following distance (m)
10.3 (0.6)
9.5 (0.5)
11.8 (0.8)
Time to collision (s)
8.0 (0.4)
8.5 (0.3)
8.1 (0.4)
Time to collision < 4 s
3.0 (0.7)
1.5 (0.3)
1.9 (0.5)
Half recovery time (s)
5.4 (0.3)
5.3 (0.3)
6.3 (0.4)
By contrast, when participants were intoxicated,
rates over the long run (e.g., T. L. Brown et al.,
neither accident rates, nor reaction time to vehicles
2001; Hirst & Graham, 1997).
braking in front of the participant, nor recovery
The MANOVA also indicated that the cell
of lost speed following braking differed signi?-
phone and alcohol conditions differed signi?cant-
cantly from baseline. Overall, drivers in the alco-
ly from each other, F(8, 32) = 4.06, p < .01. When
hol condition exhibited a more aggressive driving
drivers were conversing on a cell phone, they
style. They followed closer to the pace vehicle,
were involved in more rear-end collisions and
had twice as many trials with TTC values below
took longer to recover the speed that they had lost
4 s, and braked with 23% more force than in base-
during braking than when they were intoxicated.
line conditions. Most importantly, our study found
Drivers in the alcohol condition also applied
that accident rates in the alcohol condition did not
greater braking pressure than did drivers in the
differ from baseline; however, the increase in hard
cell phone condition.
braking and the increased frequency of TTC values
To sharpen our understanding of the differences
below 4 s are predictive of increased accident
between the cell phone and alcohol conditions, we
TABLE 2: T Test Values for the Pair-Wise Comparisons
Alcohol
Baseline
Brake reaction time (ms)
Alcohol
0.34
Cell phone
1.74*
5.46***
Maximum braking force
Alcohol
4.40***
Cell phone
4.13***
0.67
Speed (mph)
Alcohol
1.41
Cell phone
0.47
1.69*
Mean following distance (m)
Alcohol
0.87
Cell phone
1.11
1.06
SD following distance (m)
Alcohol
1.25
Cell phone
1.59
4.18***
Time to collision (s)
Alcohol
1.18
Cell phone
0.16
1.76*
Time to collision < 4 s
Alcohol
2.06**
Cell phone
1.44
1.10
Half recovery time (s)
Alcohol
0.32
Cell phone
1.96*
3.68***
Note. All comparisons have a df of 39 and are evaluated with a two-tailed signi?cance level.
*p < .10. **p < .05. ***p < .01.

---

388
Summer 2006 – Human Factors
entered the driving performance measures obtained
shorter following distances, and had more trials
for each participant into a discriminant function
with TTC values less than 4 s. On the other hand,
analysis. The discriminant analysis determines
we found that cell phones drivers had slower
which combination of variables maximally dis-
reactions, had longer following distances, took
criminates between the groups. The larger the stan-
longer to recover speed lost following a braking
dardized coef?cient, the greater the contribution
episode, and were involved in more accidents. In
of that variable to the discrimination between the
the case of the cell phone driver, the impairments
groups. Three of the obtained coef?cients were
appear to be attributable, in large part, to the
negative, affected primarily by alcohol consump-
diversion of attention from the processing of
tion: maximum braking force (–0.674), mean fol-
information necessary for the safe operation of a
lowing distance (–0.409), and TTC less than 4 s
motor vehicle (Strayer et al., 2003; Strayer &
(–0.311). Four of the obtained coef?cients were
Johnston, 2001). These attention-related de?cits
positive, affected primarily by cell phone conver-
are relatively transient (i.e., occurring while the
sations: speed (0.722), SD of following distance
driver is on the cell phone and dissipating rela-
(0.468), half recovery time (0.438), and brake reac-
tively quickly after attention is returned to driv-
tion time (0.296). Average TTC did not differen-
ing). By contrast, the effects of alcohol persist for
tiate between groups (coef?cient = 0.055). Taken
prolonged periods of time, are systemic, and lead
together, the discriminant analysis indicates that the
to chronic impairment.
pattern of impairment associated with the alcohol
Also noteworthy was the fact that the driving
and cell phone conditions is qualitatively different.
impairments associated with handheld and hands-
Finally, the accident data were analyzed using
free cell phone conversations were not signi?-
a nonparametric chi-square statistical test. The
cantly different. This observation is consistent with
chi-square analysis indicated that there were sig-
earlier reports (e.g., Patten et al., 2004; Redel-
ni?cantly more accidents when participants were
meier & Tibshirani, 1997; Strayer & Johnston,
conversing on a cell phone than in the baseline or
2001) and suggests that legislative initiatives that
alcohol conditions, ?2(2) = 6.15, p < .05.
restrict handheld devices but permit hands-free
devices are not likely to eliminate the problems
DISCUSSION
associated with using cell phones while driving.
This follows because the interference can be
Taken together, we found that both intoxicat-
attributed in large part to the distracting effects
ed drivers and cell phone drivers performed
of the phone conversations themselves, effects
differently from baseline and that the driving pro-
that appear to be attributable to the diversion of
files of these two conditions differed. Drivers
attention away from driving. It should be pointed
using a cell phone exhibited a delay in their
out that our study did not examine the effects of
response to events in the driving scenario and
dialing or answering the phone on driving per-
were more likely to be involved in a traf?c acci-
formance; however, Mazzae, Ranney, Watson,
dent. Drivers in the alcohol condition exhibited a
and Wightman (2004) compared handheld with
more aggressive driving style, following closer
hands-free devices and found the former to be
to the vehicle immediately in front of them, neces-
answered more quickly, dialed faster, and associ-
sitating braking with greater force. With respect
ated with fewer dialing errors than the latter.
to traf?c safety, the data suggest that the impair-
Our study also sheds light on the role that
ments associated with cell phone drivers may be
experience plays in moderating cell-phone-
as great as those commonly observed with intox-
induced dual-task interference. Participants’ self-
icated drivers.
reported estimates of the amount of time spent
However, the mechanisms underlying the im-
driving while using a cell phone averaged 14.3%
paired driving in the alcohol and cell phone con-
with a range from 0% to 60%. When real-world
ditions clearly differ. Indeed, the discriminant
usage was entered as a covariate into analyses
function analysis indicates that the driving pat-
comparing baseline and cell phone conditions,
terns of the cell phone driver and the drunk driver
there was no evidence that practice altered the
diverge qualitatively. On the one hand, we found
pattern of dual-task interference (i.e., all main
that intoxicated drivers hit the brakes harder, had
effects and interactions associated with real-world

---

CELL PHONE DRIVERS AND DRUNK DRIVERS
389
usage had ps > .40). That is, practice in this dual-
and depressed the brake with more vigor when the
task combination did not result in improved per-
lead vehicle began to decelerate. However, the dif-
formance. Given the attentional requirements of
ference in brake onset time between the alcohol
these two activities, it is not surprising that prac-
and baseline conditions was not signi?cant in the
tice failed to moderate the dual-task interference.
current study. The precise reason for the lack of
Because both naturalistic conversation and driv-
an effect on reaction time is unclear; although the
ing (at least reaction to unpredictable or unexpect-
literature on the effects of alcohol on reaction time
ed events) have task components that are variably
has produced mixed results (see Moskovitz &
mapped, there are likely to be few bene?ts from
Fiorentino, 2000). One possibility is that drivers
practicing these two tasks in combination. Indeed,
in the alcohol condition may have reacted with
there is overwhelming evidence in the literature
alacrity out of necessity; given their shorter fol-
that performance on components of a task with a
lowing distance, they may have been pressed into
variable mapping do not benefit from practice
action sooner than in the other conditions. In-
(e.g., Shiffrin & Schneider, 1977).
deed, an examination of the relationship between
Furthermore, the lack of differences in dual-
reaction time and following distance yielded sig-
task interference as a function of real-world usage
ni?cant correlations for the baseline (r = .47, p <
suggests that drivers may not be aware of their
.01) and cell phone (r = .56, p < .01) conditions,
own impaired driving. Indeed, when we debriefed
but not for the alcohol condition, (r = .07, ns). That
participants at the end of the experiment, many of
is, for both the baseline and cell phone conditions,
the drivers with higher levels of real-world cell
reaction time tended to increase with following
phone usage while driving indicated that they
distance, but this pattern was not observed in the
found it no more dif?cult to drive while using a
alcohol condition.
cell phone than to drive without using a cell phone.
No accidents were observed in the alcohol ses-
Thus, there appears to be a disconnect between
sions of our study. Nevertheless, alcohol clearly
participants’ self-perception of driving perfor-
increases the risk of accidents in real-world
mance and objective measures of their driving
settings. For example, the U.S. Department of
performance. Elsewhere, we have suggested that
Transportation (2002) estimated that alcohol was
one consequence of using a cell phone is that it
involved in 41% of all fatal accidents in 2002;
may make drivers insensitive to their own im-
however, it is important to note that in 81% of
paired driving behavior (Strayer et al., 2003).
these cases the blood alcohol level was higher
One factor that is often overlooked when con-
than 0.08% wt/vol and that the average blood
sidering the overall impact of cell phone driving
alcohol level of drivers involved in a fatal crash
is the effect these drivers have on traf?c ?ow. In
was twice the legal limit (i.e., 0.16% wt/vol). For
our study, we found that drivers using a cell phone
cases in which the blood alcohol level was at or
took 19% longer (than baseline) to recover the
below the legal limit, the total number of fatalities
speed that was lost following a braking episode.
in 2002 was 2818.
In situations where traffic density is high, this
Another way to determine the effect of alcohol
pattern of driving behavior is likely to decrease
on driving is to estimate the risk of an accident
the overall traf?c ?ow, and as the proportion of
when driving with a speci?c blood alcohol con-
cell phone drivers increases, these effects are
centration as compared with baseline conditions
likely to be multiplicative. That is, the impaired
when the driver is not under the in?uence of alco-
reactions of a cell phone driver make them less
hol. Using odds ratios, Zandor, Krawchuk, and
likely to travel with the ?ow of traf?c, potentially
Voas (2000) estimated the relative risk of a pas-
increasing overall traf?c congestion.
senger vehicle accident for drivers 21 to 34 years
In the current study, the performance of driv-
old. At blood alcohol concentrations between
ers with a blood alcohol level at 0.08% differed
0.05% and 0.79%, the odds ratio was estimated
signi?cantly from their performance in both the
to be 3.76, and at blood alcohol concentrations
cell phone and baseline conditions. In particular,
between 0.08% and 0.99%, the odds ratio was
when participants were in the alcohol condition,
estimated to be 6.25. Unfortunately, the precise
they followed the pace car more closely, had a
odds ratio for a blood alcohol concentration of
greater frequency of trials with TTC less than 4 s,
0.08% is not readily discernable from the tabular

---

390
Summer 2006 – Human Factors
information in the Zandor et al. (2000) study,
(1997) suggested that “the relative risk [of being
but presumably it falls somewhere between 3.76
in a traf?c accident while using a cell phone] is
and 6.25.
similar to the hazard associated with driving with
By comparison, this is the third in a series of
a blood alcohol level at the legal limit” (p. 456).
studies that we have conducted evaluating the
The data presented in this article are consistent
effects of cell phone use on driving using the car-
with this estimate and indicate that when driving
following procedure (see also Strayer & Drews,
conditions and time on task are controlled for, the
2004; and Strayer et al., 2003). Across these three
impairments associated with using a cell phone
studies, 120 participants performed in both base-
while driving can be as profound as those asso-
line and cell phone conditions. Two of the par-
ciated with driving with a blood alcohol level at
ticipants in our studies were involved in an
0.08%. With respect to cell phone use, clearly the
accident in baseline conditions, whereas 10 par-
safest course of action is to not use a cell phone
ticipants were involved in an accident when they
while driving. However, regulatory issues are best
were conversing on a cell phone. A logistic
left to legislators who are provided with the latest
regression analysis indicated that the difference
scientific evidence. In the long run, skillfully
in accident rates for baseline and cell phone con-
crafted regulation and better driver education
ditions was signi?cant, ?2(1) = 6.1, p = .013, and
addressing driver distraction will be essential to
the estimated odds ratio of an accident for cell
keep the roadways safe.
phone drivers was 5.36, a relative risk similar to
the estimates obtained by Zandor et al. (2000) for
ACKNOWLEDGMENTS
drivers with a blood alcohol level of 0.08% wt/vol.
A preliminary version of this research was
One factor that may have contributed to the ab-
presented at Driving Assessment 2003: Inter-
sence of accidents in the alcohol condition of our
national Symposium on Human Factors in Driver
study is that the alcohol and driving portion of the
Assessment, Training, and Vehicle Design in
study was conducted during the daytime (between
Park City, Utah. Support for this study was pro-
9:00 a.m. and noon). Data from the National High-
vided through a grant from the Federal Aviation
way Transportation Safety Administration (Na-
Administration. We wish to thank the Utah High-
tional Highway Traffic Safety Administration,
way Patrol for providing the breath analyzer and
2001) indicates that only 3% of fatal accidents on
GE-ISIM for providing access to the driving sim-
U.S. highways occur during this time interval. In
ulator. Danica Nelson, Amy Alleman, and Joel
fact, in the real world there is a natural confound-
Cooper assisted in the data collection. Jonathan
ing of alcohol consumption and fatigue such that
Butner provided statistical consultation. Repre-
nearly 80% of all fatal alcohol-related accidents
sentatives Ralph Becker and Kory Holdaway
on U.S. highways occur between 6:00 p.m. and
from the Utah State Legislature provided guid-
6:00 a.m. In the current study, participants were
ance on legislative issues.
well rested prior to the consumption of alcohol,
potentially lowering the relative risk factors.
REFERENCES
The objective of the present research was to
help to establish a clear benchmark for assessing
Alm, H., & Nilsson, L. (1995). The effects of a mobile telephone task
on driver behaviour in a car following situation. Accident Analysis
the relative risks associated with using a cell
and Prevention, 27, 707–715.
phone while driving. We compared the cell phone
Billings, C. E., Demosthenes, T., White, T. R., & O’Hara, D. B. (1991).
Effects of alcohol on pilot performance in simulated flight.
driver with the drunk driver for two reasons. First,
Aviation, Space, and Environmental Medicine, 62, 233–235.
there are now clear societal norms associated
Briem, V., & Hedman, L. R. (1995). Behavioural effects of mobile tele-
phone use during simulated driving. Ergonomics, 38, 2536–2562.
with intoxicated driving, and laws in the United
Brookhuis, K. A. (2003 July). On the Assessment of Criteria for Driver
States expressly prohibit driving with a blood
Impairment: In search of the Golden Yardstick for Driving Perfor-
alcohol level at or above 0.08%. Logical consis-
mance. Presented at Driving Assessment 2003: International Sympo-
sium on Human Factors in Driver Assessment, Training, and Vehicle
tency would seem to dictate that any activity that
Design, Park City, UT.
leads to impairments in driving equal to or greater
Brookhuis, K. A., De Vries, G., & De Waard, D. (1991). The effects of
mobile telephoning on driving performance. Accident Analysis and
than the drunk driving standard should be avoid-
Prevention, 23, 309–316.
ed (Willette & Walsh, 1983). Second, the epi-
Brown, I. D., Tickner, A. H., & Simmonds, D. C. V. (1969). Interference
between concurrent tasks of driving and telephoning. Journal of
demiological study by Redelmeier and Tibshirani
Applied Psychology, 53, 419–424.

---