Accident Analysis and Prevention 38 (2006) 801–810

Individual difference factors in risky driving: The roles of anger/hostility,

conscientiousness, and sensation-seeking
David C. Schwebel a,∗ , Joan Severson b , Karlene K. Ball a , Matthew Rizzo c
a Department of Psychology, University of Alabama at Birmingham, 1300 University Blvd., CH 415, Birmingham, AL 35294, United States
b Digital Artefacts, LLC, Iowa City, IA, United States
c Department of Neurology, University of Iowa, United States

Received 16 August 2005; received in revised form 1 November 2005; accepted 8 February 2006

Abstract
Motor vehicle crashes claim the lives of more Americans than any other cause of injury. One factor long recognized as relevant to predicting danger-
ous driver behavior is the driver’s personality. This study examines the independent and combined roles of three personality traits – sensation-seeking,
conscientiousness, and anger/hostility – in predicting risky driving behavior. Seventy-three participants completed personality and driving history
questionnaires, and also engaged in a virtual environment (VE) task designed to assess risk-taking driving behavior. Each facet of personality was
correlated to risky driving behavior in independent univariate analyses. In multivariate analyses, sensation-seeking emerged as the best predictor of
self-reported driving violations. Anger/hostility and the interactive effect of anger/hostility by sensation-seeking also emerged in a multivariate anal-
ysis predicting one measure of self-reported driving violations. No personality trait predicted risky driving in the VE in multivariate analyses. Results
are discussed with respect to previous work in the field, challenges involved in measuring the constructs of interest, and implications to prevention.
© 2006 Elsevier Ltd. All rights reserved.

Keywords: Risky driving; Personality; Conscientiousness; Sensation-seeking; Anger and hostility; Virtual environments

1. Introduction ticular interest as predictors of risky driver behavior: sensation-

Motor vehicle crashes claim the lives of over 42,000 Amer-
icans annually, more than any other cause of injury (National
Safety Council, 2004). Although both traffic engineering and
legislation over automobile safety devices have contributed to a
reduction in the rate of death and disability due to motor vehi-
cle crashes over the past several decades, motor vehicle crashes
remain a leading cause of fatality in the United States and a
topic of vital public health concern (National Center for Injury
Prevention and Control, 2002).
Safe management of a motor vehicle is influenced by a
wide range of individual difference variables in the driver. One
set of traits long recognized as relevant to predicting danger-
ous driver behaviors is the driver’s personality (Tillmann and
Hobbs, 1949; Fine, 1963; Arthur et al., 1991). Recent litera-
ture reviews (Beirness, 1993; Jonah, 1997) and empirical studies
(e.g., Dahlen et al., 2005) summarize empirical findings on the
topic and target three individual difference constructs of par-
∗ Corresponding author. Tel.: +1 205 934 8745; fax: +1 205 975 6110.
E-mail address: schwebel@uab.edu (D.C. Schwebel).
seeking, conscientiousness, and anger/hostility.1 This paper con-
siders those three constructs.
1.1. Individual difference factors of interest
Sensation-seeking is defined as the desire for and engage-
ment in varied, novel, complex, and arousing sensations and
1 Semantic issues complicate efforts to conceptualize individual differences
contributing to risky driving. Constructs of interest are frequently labeled with
various names, and are conceptualized slightly differently in different studies,
but they generally fall into the three broad constructs outlined in this paper.
Psychopathological factors such as depression also are known to contribute to
risky driving, but are not considered in the present study. Perhaps the trickiest
construct conceptually is “conscientiousness”. Some researchers label conscien-
tiousness using a construct from the opposite tail, such as “impulsivity”. The term
“impulsivity” is complicated because it is defined and used in widely varying
ways by different researchers. Historically, “impulsivity” was considered part
of the broader factors of “extraversion” (e.g., Eysenck and Eysenck, 1963 [who
later changed their placement of the construct]; McCrae and Costa, 1985). More
recently, theorists have placed impulsivity into the “neuroticism” (e.g. Costa &
McCrae’s NEO-PI-R, Costa and McCrae, 1992) or “conscientiousness” (e.g.,
Rothbart et al., 2001; Zuckerman, 1993) factors. We use the latter schema in the
present paper.

0001-4575/$ – see front matter © 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aap.2006.02.004
802 D.C. Schwebel et al. / Accident Analysis and Prevention 38 (2006) 801–810
experiences (Zuckerman, 1984, 1994) and is consistently linked
to risky driving behavior in empirical research (see Jonah,
1997, for a review). Early studies linking sensation-seeking with
risky driving relied on self-report measures and found mod-
erate correlations between self-reported sensation-seeking and
crash involvement among small samples (e.g., Loo, 1979). More
recent studies replicate earlier findings using more sophisticated
methodology. Burns and Wilde (1995), for instance, found links
between sensation-seeking and risky driving on the job among a
sample of almost 80 professional taxi drivers. Others used larger
samples – for example, 279 college students (Jonah et al., 2001),
120 young men in the community (Trimpop and Kirkcaldy,
1997), and over 2500 randomly sampled drivers licensed in
Norway (Iversen and Rundmo, 2002) – and reported correla-
tional links between self-reported sensation-seeking and risky
driving behaviors. Research with adolescent samples suggests
sensation-seeking is related to risky driving practices among
young drivers (e.g., Arnett, 1990, 1996, 1997). A case-control
study comparing drivers convicted and not convicted of offenses
such as speeding or reckless driving also yielded significant dif-
ferences in sensation-seeking measures among the two groups
(Furnham and Saipe, 1993).
Conscientiousness, one of the “Big Five” personality traits, is
defined as the tendency to be disciplined, responsible, and reli-
able (McCrae and Costa, 1987). Some researchers in the driving
literature conceptualize conscientiousness using other labels –
often from the opposite tail using labels such as impulsivity (e.g.,
Beirness, 1993; Dahlen et al., 2005) – but all conclude that higher
levels of conscientiousness are associated with reduced risky
driving behaviors. Arthur and Doverspike (2001), for instance,
found a correlation between conscientiousness and self-reported
crashes over the past 3 years among a sample of 48 college
students. In a larger study, Arthur and Graziano (1996) found
Conscientiousness was the only one of the Big Five personality
traits to predict crash involvement in two samples of over 200
drivers each.
Anger and hostility are constructs measured both as stable
emotional patterns and as transient dispositional states. Whether
considered as states or traits, the tendency toward hostile, frus-
trated, and angry behavior is repeatedly linked to risky driving
(e.g., Deffenbacher et al., 2001, 2002; Iversen and Rundmo,
2002). The link appears robust and has been reported in large
representative community samples (e.g., Iversen and Rundmo,
2002) as well as samples of college students (e.g., Deffenbacher
et al., 2001, 2002). For example, in a study of over 2500 ran-
domly sampled Norwegian drivers, Iversen and Rundmo (2002)
reported a significant correlation between a brief self-report
measure of risky driving and the short form of the driver anger
scale (Deffenbacher et al., 1994).
Together, existing research suggests sensation-seeking, con-
scientiousness, and anger/hostility are consistent and moderate
independent predictors of risky driving. However, most of the
research designed to examine those links has been conducted
independently—that is, one laboratory examines links between
sensation-seeking and risky driving while a different labora-
tory considers the links between anger and risky driving. This
approach has merits, but is somewhat limiting because it fails to
consider how the traits might function together to predict risky
driving. A recent study by Dahlen et al. (2005) demonstrated the
value of considering multiple traits within the same sample. In
that report, 224 young drivers (median age = 19) were recruited
to complete self-report measures of driving anger, sensation-
seeking, impulsiveness, and driving history. A series of mul-
tiple regression equations predicted aggressive driving, risky
driving, and history of poor driving (including moving tickets
and crashes). Anger emerged as the most powerful and consis-
tent predictor of all measures of poor driving, with sensation-
seeking also predicting a large portion of the driving behaviors
measured.
Along with the scarcity of studies examining how sensation-
seeking, conscientiousness, and anger/hostility might work
together in the same sample to predict risky driving, the field
faces methodological challenges. Early research relied primarily
upon self-report measures to assess risky driving. Such methods
suffer from problems with recall bias, purposeful or uninten-
tional misreporting of driving behavior, and shared method vari-
ance with other self-report measures (Boyce and Geller, 2002;
Kirk-Smith, 1998). With recent advances in the technology of
virtual reality, simulation offers a new alternative for researchers
to assess risky driving behavior.
The present study therefore assessed three individual dif-
ference factors of interest – sensation-seeking, conscientious-
ness, and anger/hostility – within the same sample, and also
asked the participants to complete a measure of risky driving
within a computer-simulated virtual environment. The study had
three primary objectives. First, we sought to replicate previous
findings independently linking risky driving with sensation-
seeking, conscientious, and angry/hostile behavior patterns.
Second, building off the results of Dahlen et al. (2005), we
tested whether anger/hostility might be a stronger predictor of
risky driving than sensation-seeking or conscientiousness when
placed together in regression equations. Finally, we consid-
ered how sensation-seeking, conscientious, and angry/hostile
behavior patterns might interact with each other to predict risky
driving. That is, we tested whether an individual who is both high
in sensation-seeking and anger/hostility might have particularly
high levels of risky driving compared to individuals scoring high
on just one of those traits.
To test our hypotheses, we used a combination of self-report
and behavioral measures. All individual difference constructs
were tested through multiple instruments. Measuring constructs
of interest through multiple measures permits more precise mea-
surement of the construct through aggregation of the multiple
measures (Epstein, 1983; Rushton et al., 1983). Risky driv-
ing was assessed both through traditional self-report measures
and through a computerized virtual environment (VE) driving
task.
2. Methods
2.1. Participants
Seventy-three college students from introductory psychology
courses at the University of Alabama at Birmingham volun-
 D.C. Schwebel et al. / Accident Analysis and Prevention 38 (2006) 801–810
teered to participate in the study as one option to complete a
course assignment (41% male, 55% female, 4% unknown sex;
44% Caucasian, 47% African American, 10% other ethnicity;
median age = 24, mean age = 27.82, S.D. = 7.94, range = 21–51
years). One other participant was dropped from analyses due to
noncompliance with the protocol. Participants had been driv-
ing for a mean of 10.52 (S.D. = 7.70) years and reported driving
an average of 198.61 miles (S.D. = 178.66) weekly. All partic-
ipants provided signed informed consent to participate in the
study, and the study protocol was reviewed and approved by the
university’s Institutional Review Board.
2.2. Protocol
Participants were invited to a laboratory setting for a 45 min
experimental session, provided informed consent to participate
in the study, and completed a brief demographic form. Partici-
pants were then taken to the virtual environment setting, which
was situated in a quiet, darkened room. The driving task was run
on a Dell Dimension 8300 Series PC with a Logitech MOMO
steering wheel and accelerator/brake hardware peripherals. The
PC utilized an Nvidia GeForce FX 5200 graphics card and other
consumer-level components. It was connected to a single 20 in.
LCD monitor.
The VE software provided a real-time 3D graphics display of
a virtual driving environment consisting of a straight, flat, two-
lane road intersected by crossroads at periodic intervals (Rizzo
and Severson, 2004; Rizzo et al., 2003). Each crossroad was rep-
resented by a large gate on both sides of the road. On a randomly
determined basis, one of three things happened as participants
approached the crossings: the gates remained fully open, the
gates closed partway (but still permitted passing through) and
then re-opened, or the gates closed entirely and then re-opened.
To create cognitive complexity, attentional distracters,
or “mudsplash” events, occurred randomly as participants
approached several of the gates. When these events occurred,
various animated figures appeared to fly toward the drivers’
visual field, catapulting into the “windshield” and therefore cre-
ating a potential distraction for drivers making decisions. These
distracters were designed to represent real-world distracters in
driving such as turning vehicles, pedestrians, and so on.
To summarize, study participants drove down a straight road
intersected by numerous crossroads. At each crossroad, partic-
ipants faced a decision about risk: they could cross directly,
without slowing, and risk the gate closing upon them; they could
decelerate to some degree, hoping to perceive whether the gate
would open or close before accelerating again; or they could stop
entirely, watching the gate before determining how to proceed.
At random intervals, visual distractions appeared during the opti-
mal decision-making times. The safest and most time-efficient
option in the course was cautious deceleration; continued accel-
eration or complete stopping represented dangerous choices in
real-world driving.
The VE driving task presented an abstract environment
designed as a tool to test and assess driver behavior, including
risk-taking. It included visual cues to permit veridical assess-
ment of the driving environment (e.g., poles offering three-
803
dimensional perception; speedometer and odometer) and phys-
ical cues to replicate the feeling of a vehicle (e.g., brake and
accelerator pedals; steering wheel). It did not attempt to replicate
vehicle dynamics or environmental scenarios of actual driving
as traditional driving simulators do.
Once situated in the room with the simulated driving task,
participants were permitted to adjust a straight-back chair to
a comfortable position and then given the following instruc-
tions: “Go through the course as if you’re actually driving.
You will see several gates. Avoid hitting closed gates, but try
to finish the course as quickly as possible. You may be dis-
tracted by computerized objects from time to time. You’ll have
5 practice gates and then 75 ‘real’ ones to finish. If you start
to feel sick at any point, just stop and I will come to get you.
Any questions?” Researchers answered any questions and then
departed. All participants were monitored through a one-way
mirror as they completed the course (which took about 12 min,
on average). No participants chose to stop due to sickness. As
described below, several measures of driving behavior were
culled from participants’ driving in the task. Following comple-
tion of the VE course, participants completed several self-report
questionnaires, also detailed below. At the end of the study,
participants were debriefed and presented with course credit
slips.
2.3. VE measures
Five measures were computed based on driving behavior in
the VE task. Because outliers created extreme positive skew in
the distributions of several measures, scores greater than 3 stan-
dard deviations above or below the mean in each measure were
removed from analyses, as listed below. Outliers were removed
only from that particular measure; that is, individuals whose
scores were outliers on one measure were not removed on other
measures in the analyses. Anecdotal evidence suggested the out-
liers were a result of attentional distraction (that is, looking away
from the computer monitor or stopping to stretch one’s arms)
rather than genuine driving behavior.
2.3.1. Bumping curb (two outliers removed)
Number of times the participant steered into the curb on either
side of the road, through the full course. The simulated road was
five times the width of the vehicle, so bumping curbs required
substantial off-center steering.
2.3.2. Hitting closed gates (three outliers removed)
Number of closed or half-open gates “hit” by the participant’s
vehicle. Such errors would have resulted in a crash in real-world
driving.
2.3.3. Speed departing opening gate
The average time in which participants covered the first 147 m
from a closed gate, after the gate opened. The 147 m distance was
chosen because it was the distance in which the vehicle could
reach maximum speed if the gas pedal was “floored” imme-
diately upon a gate opening. This measure was developed to
simulate re-acceleration following a stop.
804 D.C. Schwebel et al. / Accident Analysis and Prevention 38 (2006) 801–810
2.3.4. Slowing at open gates (three outliers removed)
The number of times the participant slowed significantly
(defined as the driver’s speed dropping below 2 m/s) before pass-
ing through an open gate. Such errors would be dangerous in
real-world driving.
2.3.5. Time to complete course (two outliers removed)
Average time, in seconds, in which participants completed
the full course of 75 gates.
2.4. Self-report measures
Participants completed seven self-report questionnaires,
described below in the order of completion.
The Simulator Sickness Questionnaire (SSQ; Kennedy et al.,
1993) is a 16-item questionnaire designed to evaluate nausea,
dizziness, and perceptual difficulties experienced as a result of
being in a virtual environment. The SSQ lists physical symptoms
(e.g., fatigue, headache, sweating) answered using a four-point
Likert scale. The items have good internal consistency in a
three-factor structure (oculomotor, disorientation, and nausea;
Kennedy et al., 1993; internal consistency in this sample was
Cronbach’s alpha = .86, .78, and .70, respectively, for the oculo-
motor, disorientation, and nausea scales).
The Driving Habits Questionnaire (DHQ) is a 12-item instru-
ment developed and adapted by the authors to assess partici-
pants’ driving history (Ball et al., 1998; Sloane et al., 1990). Sev-
eral items were of particular interest to the present report. First,
participants reported how many times they had been stopped by
the police for a traffic violation (whether or not a ticket was actu-
ally issued). Second, participants reported how many times they
had been in a crash that was serious enough to require police
involvement. Third, participants responded to four questions
concerning the average speed they usually drive (e.g., in neigh-
borhood environments, on interstates, and so forth). Questions
about driving speed were averaged to create a single measure of
average self-reported driving speed (Cronbach’s alpha = .85).
The short version of the Driving Behavior Questionnaire
(DBQ; Parker et al., 1995) is a 24-item inventory that yields
three broad factors of driving behavior: violations, errors, and
lapses. The British-developed questionnaire was altered slightly
to make it appropriate for North American drivers.2 Test-retest
and internal reliability are adequate, as is predictive validity of
the three factors (Parker et al., 1995; internal consistency in this
sample was Cronbach’s alpha = .79, .65, and .82 for violations,
errors, and lapses, respectively).
The Big Five Inventory (BFI; Benet-Martinez and John,
1998) is a 44-item instrument that yields scores on each of
the “Big Five” personality traits. The BFI has good internal
consistency and convergent validity with other measures of
2 Two types of changes were made to make the DBQ appropriate for North
American drivers. First, directions were changed since British and American
drivers use the opposite sides of the road—for example, questions about right
turns were altered to be about left turns instead. Second, British terminology
was changed to use terms more familiar to American drivers—for example, the
term ‘zebra crossing’ was changed to ‘crosswalk’.
adult personality (Benet-Martinez and John, 1998; John and
Srivastava, 1999). Of particular interest in the present study was
the Conscientiousness factor, designed to measure discipline,
responsibility, and reliability (McCrae and Costa, 1987; Cron-
bach’s alpha in this sample = .78).
The short form of the Adult Temperament Questionnaire
(ATQ; Derryberry and Rothbart, 1984, 1988; Evans and
Rothbart, 2006) is a 77-item measure yielding scores on 19
scales that fall into four broad factors—extraversion, negative
affect, effortful control, and orienting sensitivity. The ATQ has
adequate internal consistency (alphas range from .60 to .79) and
convergent validity with other measures of temperament and per-
sonality (Derryberry and Rothbart, 1988; Evans and Rothbart,
2006; Rothbart et al., 2000). Several scales from the ATQ were
of interest for the present study: activation control, designed
to assess capacity to perform an action despite a tendency to
avoid it (Cronbach’s alpha in this sample = .66); attentional con-
trol, designed to assess capacity to focus and shift attention as
desired and dictated by external stimuli (Cronbach’s alpha in
this sample = .66); inhibitory control, designed to assess capacity
to suppress approach tendencies when appropriate (Cronbach’s
alpha in this sample = .36); frustration, designed to assess neg-
ative affect related to interruption or blocking of tasks and
goals (Cronbach’s alpha in this sample = .60); and high inten-
sity pleasure, designed to assess pleasure related to high stimulus
intensity, rate, complexity, novelty, and incongruity (Cronbach’s
alpha in this sample = .64; Evans and Rothbart, 2006).
The Driving Anger Scale (DAS; Deffenbacher et al., 1994,
2001) is a 14-item inventory of self-reported anger while driving
(Cronbach’s alpha in this sample = .88). It has good internal reli-
ability and construct validity (Deffenbacher et al., 1994, 2001).
The Sensation-Seeking Scale-Form V (SSS-V; Zuckerman
et al., 1964; Zuckerman, 1994) is a 40-item measure yielding
a single overall score of sensation-seeking as well as four sub-
traits. The overall score (Cronbach’s alpha = .88 in this sample)
plus the disinhibition (alpha = .80) and boredom susceptibility
(alpha = .64) subscales were used in this study because they are
most closely related to risky driving in previous research (e.g.,
Dahlen et al., 2005; Jonah, 1997). The SSS-V has good inter-
nal consistency and convergent validity with related constructs
(Zuckerman, 1984, 1994).
3. Results
Analyses were divided into four steps: (a) examination of
descriptive statistics, including age and sex effects; (b) con-
sideration of the validity of the VE measures; (c) correlations
between individual difference factors and risky driving mea-
sures; (d) stepwise regression analysis predicting risky driving
using independent individual difference factors and then inter-
actions between them.
3.1. Descriptive analyses
Table 1 presents descriptive data for all measures of interest.
To assess for the effects of age and driving experience, cor-
relations were computed between age and all risky driving and
 D.C. Schwebel et al. / Accident Analysis and Prevention 38 (2006) 801–810 805

Table 1 sure (t(68) = −2.59, p < .05), and higher scores on the SSS-
Descriptive statistics (N = 73) V disinhibition (t(68) = 3.18, p < .01), boredom susceptibility
Variable Mean (S.D.) (t(68) = 2.65, p = .01), and total scores (t(68) = 3.01, p < .01).
Demographics
Although the tests for sex differences involved multiple com-
Age (years) 27.82 (7.94) parisons and therefore an inflated risk of Type I error, the
Years with driver license 10.52 (7.70) conservative option to control sex in subsequent analyses was
Virtual environment: risky driving selected.
Bumping curb (times) 1.30 (1.71) To assess whether participants felt sick while participating in
Hitting closed gates (times) 2.11 (4.25) the VE task (which could create problems with valid interpreta-
Speed departing opening gate (s) 31.44 (7.94) tion of results), participants’ scores on the SSQ were considered.
Slowing at open gates (times) 1.26 (2.02)
Overall, participants reported very minimal levels of sickness
Time to complete course (s) 700.46 (143.54)
(average overall mean was .31, S.D. = .42, with 0 = no sick-
Questionnaires: risky driving ness, 1 = slight sickness, 2 = moderate sickness, and 3 = severe
DBQ errors .65 (.52)
DBQ lapses .88 (.67)
sickness), suggesting behavior in the VE environment was not
DBQ violations 1.12 (.69) influenced by illness and the task was a safe and valid means of
DHQ speed .00 (.84) testing risky driving behavior. All three SSQ subscores were also
DHQ police contact 1.03 (1.64) low: nausea (M = .25, S.D. = .33), oculomotor sickness (M = .39,
DHQ crashes .37 (.64) S.D. = .50), and disorientation (M = .29, S.D. = .41).
Anger and hostility
ATQ frustration 3.71 (.98) 3.2. Validity of the VE measures
DAS total score 3.04 (.79)
Conscientiousness To test the construct validity of the VE measures, the speed
ATQ activation control 4.81 (.97)
and distance composites were correlated to self-reported risky
ATQ attentional control 4.34 (1.08)
ATQ inhibitory control 4.07 (.84) driving on the DBQ and the DHQ, with sex and years of driving
BFI conscientiousness 3.90 (.68) experience partialed.3 Higher rates of self-reported real-world
Sensation-seeking
driving violations on the DBQ were related to four of the
ATQ high intensity pleasure 4.21 (1.13) VE measures—higher rates of hitting closed gates (r(63) = .32,
SSS-V disinhibition 3.83 (2.82) p < .01), lower rates of excessive slowing at open gates
SSS-V boredom susceptibility 2.36 (1.87) (r(63) = −.29, p < .05), less time in the course (r(63) = −.33,
SSS-V total score 16.82 (7.33) p < .01), and quicker speed departing gates (r(63) = .33, p < .01).
Together, the trends suggested that people with more driv-
ing violations in their history drove through the virtual envi-
ronment more quickly. Correlational analyses also suggested
individual difference variables, and between years of driving self-reported use of speed on the DHQ correlated negatively
experience and all risky driving and individual difference vari- with time spent to complete the virtual environment course
ables. Several statistically significant correlations emerged. Age (r(63) = −.22, p < .10) and with instances of excessive slowing
was correlated to DBQ violations (r = −.36, p < .01), DHQ speed at open gates (r(63) = −.29, p < .05). Other correlations were
(r = −.42, p < .01), ATQ frustration (r = −.40, p < .01), DAS nonsignificant.
total (r = −.39, p < .01), ATQ activation control (r = .26, p < .05),
ATQ attentional control (r = .34, p < .01), ATQ inhibitory control 3.3. Correlations between individual difference factors and
(r = .40, p < .01), SSS-V disinhibition (r = −.35, p < .01), SSS-V risky driving
boredom susceptibility (r = −.34, p < .01), and BFI conscien-
tiousness (r = .43, p < .01). Not surprisingly, years of driving Table 2 shows a correlation matrix between the individual dif-
experience was highly correlated to age (r = .90, p < .01) and ference measures of interest and the VE risky driving measures,
was correlated to the same outcome variables (all rs = between with sex and years of driving experience partialed in all correla-
|.23| and |.38|, all ps ≤ .05, except correlation with SSS-V bore- tions. As shown, a number of significant correlations emerged.
dom susceptibility, p = .08). Years of driving experience was also Failing to stop at closed gates – perhaps the most serious driving
correlated to the DBQ error measure (r = −.25, p < .05). Because error in the VE – was most closely related to boredom suscepti-
of these findings, years of driving experience was controlled in bility, but also was significantly related to driver anger and poor
subsequent analyses.
To assess for sex effects, independent samples t-tests were
conducted comparing men and women on the primary vari- 3 In testing construct validity of the VE task by comparing it to self-reported
ables of interest. Several significant effects emerged, with driving history, it is important to recognize that self-reported driving history
men departing gates in the VE more quickly than women is not a “gold standard” measure of risky driving—in fact, agreement between
self-reported driving history and state Department of Transportation records
(t(68) = 2.31, p < .05) and reporting driving more quickly on is only moderate (McGwin et al., 1998). State Department of Transportation
the DHQ speed measure (t(68) = 2.63, p < .05). Men also had records were unavailable from this sample, and future evaluations of the VE
lower scores than women on the BFI conscientiousness mea- task’s construct validity are needed.
806 D.C. Schwebel et al. / Accident Analysis and Prevention 38 (2006) 801–810

Table 2
Partial correlations: individual differences and risky driving in the virtual environment (N = 73)
Individual difference Slowing at open gates Hitting closed gates Course time Speed to depart closed gates Bumping curbs

Anger and hostility aggregate −.21+ .16 −.25* .26* −.13

ATQ frustration −.26* .03 −.28* .29* −.22+
DAS total −.07 .25* −.15 .15 .00
Conscientiousness aggregate .28* −.22+ .25* −.24+ −.03
ATQ activation control .24+ −.26* .18 −.12 −.01
ATQ attentional control .13 −.22+ .26* −.27* −.05
ATQ inhibitory control .30* −.11 .11 −.21+ −.06
BFI conscientiousness .23+ −.07 .25* −.15 .02
Sensation-seeking aggregate −.37** .19 −.04 .11 .00
ATQ high intensity pleasure −.34** .01 −.12 .22+ −.12
SSS-V disinhibition −.27* .10 −.04 .09 −.10
SSS-V boredom susceptibility −.26* .34** .04 −.03 .27*
SSS-V total −.29* .15 −.00 .06 −.05

Note: years licensed and sex partialed from all correlations. Two-tailed tests. d.f. = 63.
** p < .01.
* p < .05.
+ p < .10.

activation control. The other type of error in the VE task – slow-
ing excessively while approaching gates that did not close – was
most closely related to low sensation-seeking, but was related
to almost all individual difference factors.
The time drivers spent to complete the course and the speed
with which they departed gates were both correlated to frustra-
tion and attentional control. Individuals high in frustration and
low in attentional control are likely to pursue activities quickly,
sometimes placing themselves at risk for error. Finally, the mea-
sure of participants bumping curbs in the VE was predicted
by just one individual difference factor, boredom susceptibility.
Individuals more prone to boredom bumped curbs more fre-
quently.
Table 3 displays the correlations between self-reported risky
driving and the individual difference factors, with sex and years
of driving experience partialed in all correlations. As shown,
errors and lapses reported on the DBQ were most closely related
to the conscientiousness measures, and violations on the DBQ
were related to all individual difference factors. Speed reports on
the DHQ were related to all individual difference factors except
attentional control, but self-reported driving violations on the
DHQ were related only to the total sensation-seeking score and
self-reported motor vehicle crashes were related only to poor
attentional control, low conscientiousness, and high disinhibi-
tion.
Together, the data in Tables 2 and 3 suggest broad relations
between individual difference factors and risky driving. The data
replicate previous work and offer some indication that there may
be differential predictability of particular types of individual
difference traits for particular types of risky driving.

Table 3
Partial correlations: self-reported risky driving and individual differences (N = 73)
Individual difference DBQ errors DBQ lapses DBQ violations DHQ speed DHQ violations DHQ crashes

Anger and hostility aggregate .25* .21+ .55** .44** .09 .21+
ATQ frustration .24+ .21+ .53** .46** .40 .18
DAS total .20 .14 .41** .35** .04 .18
Conscientiousness aggregate −.38** −.45** −.51** −.41** .03 −.33**
ATQ activation control −.28* −.28* −.46** −.34** .02 −.17
ATQ attentional control −.35** −.44** −.35** −.10 .00 −.35**
ATQ inhibitory control −.15 −.20 −.29* −.41** .19 −.13
BFI conscientiousness −.40** −.49** −.49** −.41** −.13 −.39**
Sensation-seeking aggregate .21+ .19 .53** .50** .24* .25*
ATQ high intensity pleasure −.00 −.04 .30* .35** .16 .16
SSS-V disinhibition .24+ .21+ .54** .45** .21+ .30*
SSS-V boredom susceptibility .29* .26* .43** .36** .12 .12
SSS-V total .15 .17 .40** .33** .27* .22+

Note: years licensed and sex partialed from all correlations. Two-tailed tests. d.f. = 63.
** p < .01.
* p ≤ .05.
+ p < .10.
 D.C. Schwebel et al. / Accident Analysis and Prevention 38 (2006) 801–810 807

3.4. Stepwise regression analysis predicting risky driving
The next step of the analysis was construction of stepwise
linear regression equations to predict risky driving. In each
analysis, demographic characteristics (sex and years of driv-
ing experience) were entered first, followed by measures of
anger/hostility, conscientiousness, and sensation-seeking, and
then interactions between the individual difference factors.
Because of the number of measures used in this study, the
principle of aggregation was invoked to reduce predictors in the
regression equations. Aggregation of multiple measures of the
same construct offers greater specificity in the assessment of
a construct and is recommended to achieve greater predictive
power in regression equations and other models (Epstein, 1983;
Rushton et al., 1983).
Individual difference factors were standardized and aggre-
gated using the clusters shown in Tables 2 and 3. Each of the
three individual difference aggregates – anger/hostility, consci-
entiousness, and sensation-seeking – had strong average inter-
item correlations (.54, .51, and .54, respectively). Correlations
between the aggregate measures and the risky driving measures
appear in Tables 2 and 3, with age and gender partialed.
Principal components analysis was used to educate the deci-
sion on how the dependent driving measures might be aggre-
gated. All VE measures were entered into one principal com-
ponents analysis, and two theoretically coherent components
(eigenvalues > 1) emerged. The first included measures of driv-
ing speed (time in course, slowing excessively at open gates, and
[reversed] acceleration after departing opening gates; average
intercorrelation = .57). The second included measures of bump-
ing obstacles in the course (hitting closed gates and bumping
curbs; r = .35). Thus, two aggregates were created from the VE
task: risky use of time and risky control of vehicle. With gender
and years licensed partialed, the risky use of time measure cor-
related to the anger/hostility, conscientiousness, and sensation-
seeking aggregates at r(63) = −.31 (p < .05), .31 (p < .05), and
−.17, respectively, and the risky control of the vehicle aggre-
gate correlated to the same three traits at r(63) = .04, −.16, and
.10, respectively.
DBQ and DHQ measures were also entered into principal
components analyses. Again, two components emerged (eigen-
values > 1), but in this case there were variables that split between
components and the components were not theoretically mean-
ingful. Because intercorrelations between several of the self-
report driving measures were poor, we decided aggregation was
inappropriate and instead focused on the two measures of vio-
lations as the self-report dependent variables in the regression
models.
Tables 4 and 5 outline the stepwise linear regression mod-
els (note that conscientiousness was reversed before computing
interactions). Multicollinearity diagnostics suggested adequate
independence of predictors (all tolerance levels < .80). As shown
in Table 4, sex predicted risky use of time in the VE but personal-
ity and temperament traits were not strong predictors of behavior
in the VE. This finding supports the correlations reported in
Table 2 and suggests that personality was only a modest predic-
tor of risky behavior in the simulated driving task.
When self-reported driving violations were predicted instead
of behavior in the VE (see Table 5), sensation-seeking emerged
as a statistically significant predictor of violations reported on
both measures. Anger/hostility was also a significant predictor
of violations reported on the DBQ. Just one significant interac-

Table 4
Stepwise linear regression predicting risky driving in the virtual environment (N = 73)
Variable Risky use of time Risky vehicle control

B S.E. B β B S.E. B β

Step 1—demographic variables

Sex (male = 1, female = 2) .48 .23 .26* .04 .22 .03
Years of driving experience −.00 .02 −.02 −.02 .01 −.14
Step 2—individual difference variables
Sex (male = 1, female = 2) .49 .24 .26* .09 .24 .05
Years of driving experience −.02 .02 −.17 −.01 .02 −.09
Anger/hostility aggregate −.21 .16 −.20 −.09 .16 −.09
Conscientiousness aggregate .23 .19 .19 −.22 .19 −.20
Sensation-seeking aggregate −.03 .17 −.03 .05 .17 .05
Step 3—interactions
Sex (male = 1, female = 2) .50 .25 .27+ .15 .26 .08
Years of driving experience −.03 .02 −.21 −.01 .02 −.07
Anger/hostility aggregate −.23 .16 −.22 −.09 .16 −.09
Conscientiousness aggregate .24 .19 .20 −.24 .20 −.22
Sensation-seeking aggregate .00 .17 .00 .06 .17 .06
Anger/hostility by conscientiousness .28 .19 .21 .07 .21 .05
Anger/hostility by sensation-seeking −.32 .21 −.21 −.09 .20 −.07
Conscientiousness by sensation-seeking −.16 .24 −.10 −.15 .25 −.10

Note: predicting risky use of time: R2 = .07 (model p < .10) on Step 1; R2 = .11 (model p < .05) on Step 2; R2 = .06 (model p < .05) on Step 3. Predicting risky
control of vehicle: R2 = .02 on Step 1; R2 = .03 on Step 2; R2 = .02 on Step 3.
* p ≤ .05.
+ p < .10.
808 D.C. Schwebel et al. / Accident Analysis and Prevention 38 (2006) 801–810

Table 5
Stepwise linear regression predicting risky self-reported driving (N = 73)
Variable DBQ violations DHQ violations

B SE B β B SE B β

Step 1—demographic variables

Sex (male = 1, female = 2) −.22 .18 −.16 −.65 .43 −.19
Years of driving experience −.02 .01 −.20 −.01 .03 −.04
Step 2—individual difference variables
Sex (male = 1, female = 2) −.02 .15 −.02 −.24 .46 −.07
Years of driving experience .00 .01 .05 −.01 .03 −.03
Anger/hostility aggregate .26 .10 .32** .27 .30 .14
Conscientiousness aggregate −.15 .11 −.17 .62 .35 .30+
Sensation-seeking aggregate .37 .10 .42** .81 .31 .39*
Step 3—interactions
Sex (male = 1, female = 2) −.05 .15 −.03 −.08 .48 −.02
Years of driving experience .01 .01 .08 .01 .03 .03
Anger/hostility aggregate .27 .10 .34** .29 .30 .15
Conscientiousness aggregate −.16 .11 −.19 .57 .35 .27
Sensation-seeking aggregate .33 .10 .38** .81 .32 .39*
Anger/hostility by conscientiousness −.05 .11 −.05 −.70 .35 −.29+
Anger/hostility by sensation-seeking .26 .12 .22* .58 .38 .21
Conscientiousness by sensation-seeking −.09 .14 −.08 .10 .43 .03

Note: DBQ violations: R2 = .08 (model p < .10) on Step 1; R2 = .45 (model p < .01) on Step 2; R2 = .04 (model p < .01) on Step 3. DHQ violations: R2 = .04 on
Step 1; R2 = .12 (model p < .10) on Step 2; R2 = .07 (model p < .05) on Step 3.
** p < .01.
* p < .05.
+ p < .10.

tion effect emerged: the interaction between anger/hostility and
sensation-seeking significantly predicted violations reported on
the DBQ. This finding offers minimal evidence that individu-
als who are both angry/hostile and high sensation-seekers might
have particularly elevated risk of driving violations.
4. Discussion
Taken together, results replicate previous work suggest-
ing personality is related to risky driving behavior: we found
sensation-seeking, conscientious, and angry/hostile behavior
patterns each predicted risky driving. Many of these relations
were strong, even after partialing the effects of sex and years
licensed, and they reinforce the long-held belief that drivers’
personalities effect how they drive (Beirness, 1993; Tillmann
and Hobbs, 1949).
We tested two other primary hypotheses. We sought to repli-
cate the finding by Dahlen et al. (2005) that anger and hostil-
ity might play a stronger role in explaining risky driving than
sensation-seeking or conscientiousness. Like Dahlen et al., we
found that each of the three traits contributes independently and
incrementally to explain risky driving. Unlike Dahlen et al.,
however, we did not consistently find anger/hostility to be the
strongest predictor of risky driving when entered together with
the other traits in linear regression models. We can only speculate
why our results may have converged from theirs. One possibility
is measurement. The two studies used different self-report ques-
tionnaires to assess individual difference constructs of interest
and the error variance present in measuring the constructs could
explain the disparity. Measurement of risky driving could also
partially explain the divergence: both studies used self-report
measures, but our inclusion of the VE measure could have tapped
somewhat different aspects of risky driving than self-report mea-
sures. Another possible explanation for the differing results is
peculiarities of the samples. Both studies recruited ethnically
diverse college student samples, but Dahlen et al. had a larger
sample and therefore greater statistical power. Further research
on the topic is recommended.
The third primary hypothesis of our study was to test
whether individual difference factors might interact to increase
or decrease risky driving. We wondered, for example, whether
an individual who was low in conscientiousness and high in
sensation-seeking may be at particularly elevated risk for dan-
gerous driving—perhaps at much greater risk than an individual
scoring high on just one trait. Just one statistically significant
interaction effect emerged in our four regression equations,
offering a bit of preliminary evidence that there may be some
multiplicative effect of individual differences but suggesting
more strongly that the effect of individual differences on risky
driving is not strongly multiplicative. In other words, an indi-
vidual who is high on sensation-seeking may not drive in an
particularly more risky manner if he or she also scores high on
anger/hostility.
4.1. Measurement issues
The literature linking personality and driving depends greatly
on self-report measures (Jonah, 1997; Boyce and Geller, 2002).
Although self-report measures have their merits, theorists agree
that observational measures of behavior offer good supplements
 D.C. Schwebel et al. / Accident Analysis and Prevention 38 (2006) 801–810
to self-report measures (e.g., Mount et al., 1994; Rothbart and
Bates, 1998). In this study, self-reported personality was more
closely related to self-reported risky driving history than it was to
risky driving in the VE. Perhaps an artifact of measurement bias
(Campbell and Fiske, 1959), this finding highlights the impor-
tance of obtaining multiple methods of behavior in order to best
examine the relations between constructs of interest.
The use of virtual environments to measure risky driving
behavior is rare in the field of predicting risky driving based on
individual differences (see Deery and Fildes, 1999; Rosenbloom
and Wolf, 2002, for exceptions). Critics might question the valid-
ity of VE as a measure of real-world driving. In particular, one
might question whether the task tested in this study – negoti-
ating gates that close and open in unpredictable manners – is
representative of the types of decisions drivers must make while
negotiating actual roadways.
We respond in two ways. First, there was statistical evidence
of the validity of the measure. Behavior in the VE correlated
moderately with self-reported risky driving behaviors. This find-
ing matches results studying real and simulated driving perfor-
mance among older adults (Freund et al., 2002; Lee et al., 2003)
and suggests use of VE as a means to test risky driving is a valid
and potentially useful alternative to self-report measures.
Second, we paralleled the process of actual driving in the
virtual environment task. Negotiating unpredictable gates in a
virtual environment represents the type of decisions automo-
bile drivers make daily: how does one respond to unpredictable,
rapidly changing stimuli while manipulating a steering wheel, a
gas pedal, and a brake pedal? How does one balance the goals of
safe driving with a desire to reach one’s destination in an efficient
manner? There is no optimal way to negotiate the unpredictabil-
ity found in our virtual course, just as there is no optimal way to
negotiate real-world driving. A driver must reach his or her goal
efficiently, but without crashing. As examples, does a driver dart
through a traffic light as it turns yellow, or instead brake quickly?
Does a driver change lanes to avoid a slow-moving vehicle, or
brake quickly to stop in front of it? Does a driver stop for a
pedestrian approaching the intersection, or traverse the intersec-
tion quickly before she reaches the crosswalk? Each of these
decisions parallel the sort of cognitive and motor processes we
tested in the virtual environment.
4.2. Implications for prevention
Most current interventions to reduce motor vehicle crashes
target engineering of the road environment or the motor vehi-
cle. Prevention programs designed to change driver behavior are
less common; when they are implemented, they tend to focus
broadly on all drivers. Results from this study indicate interven-
tions targeting sensation-seeking, unconscientious, and angry
drivers might be theoretically and economically sensible.
4.3. Limitations
This study had limitations that warrant mention in closing.
First, the sample was limited by including only college students
enrolled in psychology courses. Future work should include a
809
wider age span and more socioeconomically diverse sample.
Second, the study relied entirely on self-report measures of
personality. Future work should include behavioral measures
of personality as well as behavioral measures of risky driving.
Third, statistical analyses were limited by moderate power. The
sample of 73 participants yielded power = .55, .85, and .57 to
detect a medium effect size for t-tests, correlations, and multi-
ple regression with eight predictors, respectively; power was
.91, 1.00, and .95, respectively, to detect a large effect size.
Small effect sizes may not have been detected with the sample
used. Fourth, the issue of exposure is skirted in these analyses.
Some individual drive more miles than others, and therefore
have greater opportunity to violate laws and crash; although
we controlled for years licensed, exposure by miles driven was
not considered. Finally, the study overlooked cognitive aspects
of driving that may play an important role in risk-taking while
driving. Issues of attention and distraction, concentration, and
simultaneous processing of multiple stimuli are potentially rel-
evant but were not assessed.
Acknowledgements
Thanks to Justin Alexander, Ben Barton, and Zach Osborn for
assistance with data collection and coding and to Kerri Mead,
Joseph Altmeir, and Jim Cremer for assistance with software
development. This work was funded by grants from the US
Department of Transportation (Schwebel, Ball) and National
Institutes of Health (Severson, Rizzo).
References
Arnett, J.J., 1990. Drunk driving, sensation seeking, and egocentrism among
adolescents. Pers. Indiv. Differ. 11, 541–546.
Arnett, J.J., 1996. Sensation seeking, aggressiveness, and adolescent reckless
behavior. Pers. Indiv. Differ. 20, 693–702.
Arnett, J.J., 1997. Reckless driving in adolescence: “state” and “trait” factors.
Accident Anal. Prev. 29, 57–63.
Arthur Jr., W., Barrett, G.V., Alexander, R.A., 1991. Prediction of vehicular
accident involvement: a meta-analysis. Human Perform. 4, 89–105.
Arthur Jr., W., Graziano, W.G., 1996. The five-factor model, conscientious-
ness, and driving accident involvement. J. Pers. 64, 593–618.
Arthur Jr., W., Doverspike, D., 2001. Predicting motor vehicle crash involve-
ment from a personality measure and a driving knowledge test. J. Prev.
Interv. Commun. 22, 35–42.
Ball, K., Owsley, C., Stalvey, B., Roenker, D.L., Graves, M., 1998. Driving
avoidance, functional impairment, and crash risk in older drivers. Accident
Anal. Prev. 30, 313–322.
Beirness, D.J., 1993. Do we really drive as we live? The role of personality
factors in road crashes. Alcohol, Drugs, and Driving 9, 129–143.
Benet-Martinez, V., John, O., 1998. Los Cinco Grandes across cultures and
ethnic groups: multitrait-multimethod analyses of the Big Five in Spanish
and English. J. Pers. Soc. Psychol. 75, 729–750.
Boyce, T.E., Geller, E.S., 2002. An instrumented vehicle assessment of prob-
lem behavior and driving style: do younger males really take more risks?
Accident Anal. Prev. 34, 51–64.
Burns, P.C., Wilde, G.J.S., 1995. Risk taking in male taxi drivers; relation-
ships among personality, observational data and driver records. Pers. Indiv.
Differ. 18, 267–278.
Campbell, D.T., Fiske, D.W., 1959. Convergent and discriminant validation
by the multitrait-multimethod matrix. Psychol. Bull. 56, 81–105.
Costa Jr., P.T., McCrae, R.R., 1992. The NEO-PI-R: Professional Manual.
Psychological Assessment Resources, Odessa, FL.
810 D.C. Schwebel et al. / Accident Analysis and Prevention 38 (2006) 801–810
Dahlen, E.R., Martin, R.C., Ragan, K., Kuhlman, M.M., 2005. Driving anger,
sensation seeking, impulsiveness, and boredom proneness in the prediction
of unsafe driving. Accident Anal. Prev. 37, 341–348.
Deery, H.A., Fildes, B.N., 1999. Young novice drivers subtypes: relation-
ship to high-risk behavior, traffic accident record, and simulator driving
performance. Human Factors 41, 628–643.
Deffenbacher, J.L., Lynch, R.S., Oetting, E.R., Swaim, R.C., 2002. The driv-
ing anger expression inventory: a measure of how people express their
anger on the road. Behav. Res. Therapy 40, 717–737.
Deffenbacher, J.L., Lynch, R.S., Oetting, E.R., Yingling, D.A., 2001. Driving
anger: correlates and a test of state-trait theory. Pers. Indiv. Differ. 31,
1321–1331.
Deffenbacher, J.L., Oetting, E.R., Lynch, R.S., 1994. Development of a driv-
ing anger scale. Psychol. Rep. 74, 83–91.
Derryberry, D., Rothbart, M.K., 1984. Emotion, attention and temperament.
In: Izard, C.E., Kagan, J., Zajonc, R.B. (Eds.), Emotions, Cognition, and
Behavior. Cambridge University Press, New York, pp. 132–166.
Derryberry, D., Rothbart, M.K., 1988. Arousal, affect, and attention as com-
ponents of temperament. J. Pers. Soc. Psychol. 55, 958–966.
Epstein, S., 1983. Aggregation and beyond: some basic issues on the predic-
tion of behavior. J. Pers. 51, 360–392.
Evans, D.E., Rothbart, M.K., 2006. A Hierarchical Model of Temperament
and the Big Five. University of Oregon, Eugene, in preparation.
Eysenck, S.B.G., Eysenck, H.J., 1963. On the dual nature of extraversion.
Br. J. Soc. Clin. Psychol. 2, 46–55.
Fine, B.J., 1963. Introversion–extraversion and motor vehicle driver behavior.
Percept. Motor Skills 12, 95–100.
Freund, B., Gravenstein, S., Ferris, R., Shaheen, E., 2002. Evaluating driving
performance of cognitively impaired and healthy older adults: a pilot
study comparing on-road testing and driving simulation. J. Am. Geriatr.
Soc. 50, 1309–1310.
Furnham, A., Saipe, J., 1993. Personality correlates of convicted drivers. Pers.
Indiv. Differ. 14, 329–336.
Iversen, H., Rundmo, T., 2002. Personality, risky driving and accident
involvement among Norwegian drivers. Pers. Indiv. Differ. 33, 1251–
1263.
John, O.P., Srivastava, S., 1999. The big five trait taxonomy: History, mea-
surement, and theoretical perspectives. In: Pervin, L.A., John, O.P. (Eds.),
Handbook of Personality: Theory and Research. Guilford, New York, pp.
102–138.
Jonah, B.A., 1997. Sensation seeking and risky driving: a review and syn-
thesis of the literature. Accident Anal. Prev. 29, 651–665.
Jonah, B.A., Thiessen, R., Au-Yeung, E., 2001. Sensation seeking, risky
driving and behavioral adaptation. Accident Anal. Prev. 33, 679–684.
Kennedy, R.S., Lane, N.E., Berbaum, K.S., Lilienthal, M.G., 1993. Simulator
sickness questionnaire: an enhanced method for quantifying simulator
sickness. Int. J. Aviat. Psychol. 3, 203–220.
Kirk-Smith, M., 1998. Psychological issues in questionnaire-based research.
J. Market Res. Soc. 40, 223–236.
Lee, H.C., Cameron, D., Lee, A.H., 2003. Assessing the driving performance
of older adult drivers: on-road versus simulated driving. Accident Anal.
Prev. 35, 797–803.
Loo, R., 1979. Role of primary personality factors in the perception of traf-
fic signs and driver violations and accidents. Accident Anal. Prev. 11,
125–127.
McCrae, R.R., Costa Jr., P.T., 1985. Comparison of EPI and psychoticism
scales with measures of the five-factor model of personality. Pers. Indiv.
Differ. 6, 587–597.
McCrae, R.R., Costa Jr., P.T., 1987. Validation of the five-factor model of
personality across instruments and observers. J. Pers. Soc. Psychol. 52,
81–90.
McGwin Jr., G., Owsley, C., Ball, K., 1998. Identifying crash involvement
among older drivers: agreement between self-report and state records.
Accident Anal. Prev. 30, 781–791.
Mount, M.K., Barrick, M.R., Strauss, J.P., 1994. Validity of observer ratings
of the big five personality factors. J. Appl. Psychol. 79, 272–280.
National Center for Injury Prevention and Control, 2002. CDC Injury
Research Agenda. Centers for Disease Control and Prevention, Atlanta.
National Safety Council, 2004. Injury Facts: 2004 Edition. National Safety
Council, Itasca, IL.
Parker, D., Reason, J.T., Manstead, A.S.R., Stradling, S.G., 1995. Driv-
ing errors, driving violations and accident involvement. Ergonomics 38,
1036–1048.
Rizzo, M., Severson, J., 2004. Abstract virtual environments for assessing
cognitive abilities. In: Proceedings of the 31st International Conference
(SIGGRAPH 2004), Los Angeles, August 2004.
Rizzo, M., Severson, J., Cremer, J., Price, K., 2003. An abstract virtual
environment tool to assess decision-making in impaired drivers. In: Pro-
ceedings of the Second International Driving Symposium on Human
Factors in Driver Assessment, Training, and Vehicle Design, Park City,
UT, pp. 40–47.
Rosenbloom, T., Wolf, Y., 2002. Sensation seeking and detection of risky road
signals: a developmental perspective. Accident Anal. Prev. 34, 569–580.
Rothbart, M.K., Ahadi, S.A., Evans, D.E., 2000. Temperament and person-
ality: origins and outcomes. J. Pers. Soc. Psychol. 78, 122–135.
Rothbart, M.K., Ahadi, S.A., Hershey, K.L., Fisher, P., 2001. Investigations
of temperament at three to seven years: the Children’s Behavior Ques-
tionnaire. Child Dev. 72, 1394–1408.
Rothbart, M.K., Bates, J.E., 1998. Temperament. In: Damon, W. (Series Ed.),
Eisenberg, N. (Vol. Ed.), Handbook of Child Psychology: vol. 3. Social,
Emotional and Personality Development, 5th ed. Wiley, New York, pp.
105–176.
Rushton, J.P., Brainerd, C.J., Pressley, M., 1983. Behavioral development and
construct validity: the principle of aggregation. Psychol. Bull. 94, 18–38.
Sloane, M.E., Ball, K., Owsley, C., Roenker, D.L., Bruni, J.R., 1990. The
driving habits questionnaire: an evaluation of the validity of self-reported
driving habits. Unpublished raw data.
Tillmann, W.A., Hobbs, G.E., 1949. The accident-prone automobile driver:
a study of the psychiatric and social background. Am. J. Psychiatry 106,
321–331.
Trimpop, R., Kirkcaldy, B., 1997. Personality predictors of driving accidents.
Pers. Indiv. Differ. 23, 147–152.
Zuckerman, M., 1984. Experience and desire: a new format for sensation
seeking scales. J. Behav. Assess. 6, 101–114.
Zuckerman, M., 1993. P-impulsive sensation seeking and its behavioral, psy-
chophysiological, biochemical correlates. Neuropsychobiology 28, 30–36.
Zuckerman, M., 1994. Behavioral Expressions and Biosocial Bases of Sen-
sation Seeking. Cambridge University Press, New York.
Zuckerman, M., Kolin, E.A., Price, L., Zoob, I., 1964. Development of a
sensation-seeking scale. J. Consult. Clin. Psychol. 28, 477–482.