This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
Author's personal copy
Use of a video monitoring approach to reduce at-risk driving behaviors
in commercial vehicle operations
Jeffrey S. Hickman
, Richard J. Hanowski
Virginia Tech Transportation Institute, 3500 Transportation Research Plaza, Blacksburg, VA 24061, United States
Received 5 November 2009
Received in revised form 17 November 2010
Accepted 18 November 2010
At-risk driving behaviors
The Federal Motor Carrier Safety Administration (FMCSA) funded this project to provide an
independent evaluation of a commercially available onboard safety monitoring (OBSM)
system. Participating drivers drove a truck instrumented with data collection equipment
(e.g., two video cameras and three accelerometers) for 17 consecutive weeks while they
made their normal, revenue-producing deliveries. During the 4-week Baseline phase, the
OBSM device recorded safety-related events; however, the feedback light on the OBSM
device was disabled and safety managers did not have access to the recorded safety-related
events to provide feedback to drivers. During the 13-week Intervention phase, the feedback
light on the OBSM device was activated and safety managers had access to the recorded
safety-related events and followed the coaching protocol with drivers (when necessary).
Carrier A signiﬁcantly reduced the mean rate of recorded safety-related events/10,000
miles traveled from Baseline to Intervention by 37% (p= 0.046) and Carrier B signiﬁcantly
reduced the mean rate of recorded safety-related events/10,000 miles traveled from Base-
line to Intervention by 52.2% (p= 0.034). The results suggest the combination of video
monitoring and behavioral coaching was responsible for the reduction in the mean rate
of safety-related events/10,000 miles traveled at Carriers A and B.
Ó2010 Elsevier Ltd. All rights reserved.
Many drivers choose to behave in ways that put themselves and others at risk for a vehicle crash and/or serious injuries.
One of the most signiﬁcant studies on the factors that contribute to motor vehicle crashes was the Indiana Tri-Level Study
(Treat et al., 1979). To provide insight into the factors that contribute to trafﬁc crashes, collision data were collected across
three different levels to assess causal factors as being deﬁnite, probable, or possible. The study determined that 90.3% of the
crashes involved some type of human error, such as at-risk driving behavior, inadvertent errors, and impaired states.
Although the vehicles in Treat et al. (1979) were predominantly passenger vehicles, the same relationship can be found
in heavy vehicles. The recently completed Large Truck Crash Causation Study (LTCCS) assessed the causes of, and contribut-
ing factors to, crashes involving commercial motor vehicles (CMV). The LTCCS found that 87.3% of the critical reasons as-
signed to the large-truck driver were driver errors, including decision errors (38%; e.g., driver drove too fast for
conditions), recognition errors (28.4%; e.g., driver did not recognize the situation due to not paying proper attention),
non-performance errors (11.6%; e.g., driver fell asleep), and performance errors (9.2%; e.g., driver exercised poor directional
control) (Federal Motor Carrier Safety Administration (FMCSA), 2006).
1369-8478/$ - see front matter Ó2010 Elsevier Ltd. All rights reserved.
Corresponding author. Tel.: +1 540 231 1442.
E-mail address: email@example.com (J.S. Hickman).
Transportation Research Part F 14 (2011) 189–198
Contents lists available at ScienceDirect
Transportation Research Part F
journal homepage: www.elsevier.com/locate/trf
Author's personal copy
1.1. Behavioral approaches to safety
Behavioral approaches to safety have provided robust positive results when applied in organizations seeking to reduce
employee injuries due to at-risk behaviors. Primary techniques include peer observation and feedback, goal setting, and
training and education sessions (Geller, 2001; Krause, Robin, & Knipling, 1999). Almost all prior behavioral safety research
has been conducted in work settings where employees can systematically observe the safe versus at-risk behavior of their
coworkers. Drawbacks to this approach include: nonobjective, unreliable, or biased observation of behavior; the need for
extensive training of observers; paid employee time needed to make interpersonal behavioral observations; lack of motiva-
tion to make behavioral observations and deliver feedback; and resistance to accept nonobjective and potentially biased
feedback (Geller & Clarke, 1999). These drawbacks are exacerbated in workers who operate heavy trucks as they are typically
solitary workers or workers with little supervision. Because employees who operate a heavy vehicle as part of their job du-
ties work alone, and the large human and economic costs associated with large-truck crashes, there would be great potential
beneﬁt from research developing practical behavioral safety techniques with CMV drivers.
Self-management has been one behavioral approach to safety that has been effective with lone workers. Self-manage-
ment is a behavior improvement process whereby individuals change their own behavior in a goal-directed fashion (Mahon-
ey, 1972, 1976) by: (i) manipulating behavioral antecedents; (ii) observing and recording speciﬁc target behaviors; and (iii)
self-administering rewards for personal achievements (Watson & Tharp, 1993). Hickman and Geller (2003) instructed short-
haul truck drivers at two trucking terminals several self-management strategies, including the identiﬁcation of antecedents
and consequences of at-risk driving behaviors, goal setting strategies, self-rewards, peer support, and how to self-observe
their own safety-related work behaviors using a self-monitoring form. The data suggests using self-management strategies
in increasing safety-related driving behaviors with professional drivers is not only feasible, but resulted in signiﬁcant de-
creases in two critically important at-risk driving behaviors (extreme braking and overspeed). Similarly, self-management
techniques were successfully used to increase the safety practices of bus divers (Olson & Austin, 2001) and commercial mo-
tor vehicle drivers (Krause, 1997). However, self-management has several of the drawbacks noted above in peer observation
and feedback, including: nonobjective, unreliable, or biased observation of behavior; the need for extensive training of
observers, and lack of motivation to make behavioral self-observations.
The problem, until recently, has been getting quality behavioral data on driving behaviors. Most CMV organizations use
reactive approaches to assess safety outcomes. These include the frequency and severity of crashes and violations. However,
crashes and violations only show a snapshot of driver behavior and it’s too late to intervene on driver behavior once a crash
occurs. A proactive approach focuses on speciﬁc driver behaviors—a leading indicator of driver safety that can address at-risk
driving behaviors as they occur, prior to a crash and/or violation. If behavioral approaches can be integrated with technol-
ogies that monitor driver behavior, ﬂeet safety managers would have an effective tool to improve safety-related behaviors
that occur when there is little or no opportunity for interpersonal observation and feedback. Moreover, these data provide
the safety manager with leading indicators of driver safety; thus, ﬂeet safety managers can address potential safety issues
prior to the occurrence of a crash and/or violation.
1.2. On-board safety monitoring devices
New technologies are currently available that provide objective measures of driver behavior. These in-vehicle technolo-
gies are able to provide continuous measures on a wide variety of driving behaviors previously unavailable to ﬂeet safety
managers. The most efﬁcacious onboard safety monitoring (OBSM) systems use in-vehicle video technology to record driver
behavior. These video recordings can be used by ﬂeet safety managers to provide feedback on safe and at-risk driving behav-
iors. Behavioral approaches to safety are directed at modifying at-risk driving behaviors to reduce crash and injury risk. Thus,
OBSM systems have the potential to be used in conjunction with behavioral safety techniques to greatly reduce a variety of
McGehee, Raby, Carney, Lee, and Reyes (2007) used in-vehicle video technology with newly licensed teen drivers by pair-
ing the video monitoring with parental feedback in the form of a weekly video review and a graphical report card. Each teen
driver had his/her personal vehicle equipped with an event-triggered video device, designed to capture 20-s clips of the for-
ward and cabin views whenever the vehicle exceeded lateral or forward threshold accelerations. Results indicated that the
combination of video feedback and a graphical report card signiﬁcantly decreased the rate of safety-related events in teen
drivers. Note the participants in McGehee et al. (2007) were novice teen drivers; these novice drivers were still acquiring
basic knowledge and skills in regards to driving a motor vehicle, whereas the participants in the current study were expe-
rienced CMV drivers. It’s unclear if professional CMV drivers will experience the same safety beneﬁts as teen drivers with
parental oversight. The current study provided an independent evaluation of a commercially available OBSM system with
1.3. Speciﬁc aims
The speciﬁc aim of this project was to assess the safety beneﬁts of a commercially available OBSM system. The ﬁrst
hypothesis stated there would be a signiﬁcant reduction in the mean rate of safety-related events from the Baseline to Inter-
vention phase (i.e., the mean rate of safety-related events during the Baseline and Intervention phases was compared). The
190 J.S. Hickman, R.J. Hanowski / Transportation Research Part F 14 (2011) 189–198
Author's personal copy
dependent measure was the mean rate of safety-related events/10,000 miles traveled. Thus, for each participant the fre-
quency of safety-related events during the Baseline phase was divided by the number of miles traveled during data collec-
tion in the Baseline phase. This normalized the data and accounted for missing days, dropouts, and/or exposure. The same
procedures were used for data collected during the Intervention phase. A paired sample t-test was used to assess if a there
was a signiﬁcant reduction in the mean rate of safety-related events/10,000 miles during the Baseline to the Intervention
The second hypothesis stated there would be a signiﬁcant reduction in the mean rate of severe safety-related events from
the Baseline to Intervention phase (i.e., the mean rate of severe safety-related events during the Baseline and Intervention
phases were compared). A ‘‘severe’’ event was deﬁned as any safety-related event with an Event Score > 5. For each partic-
ipant the frequency of severe safety-related events during the Baseline phase was divided by the number of miles during the
Baseline phase. The same procedures were used for data collected during the Intervention phase. A paired sample t-test was
used to assess if there was a signiﬁcant reduction in the mean frequency of severe safety-related events/10,000 miles from
Baseline to Intervention (
The quasi-experiment (i.e., no participant randomization) used a simple A
design; where ‘‘A’’ and ‘‘B’’ referred to the
Baseline and Intervention phases, respectively. The superscript referred to the number of weeks in each phase (i.e., ‘‘4’’ re-
ferred to four weeks). During the 4-week Baseline phase, drivers drove an instrumented vehicle during their normal, reve-
nue-producing deliveries. The OBSM device was conﬁgured to record safety-related events as normal; however, the feedback
light (a light on the OBSM device, visible to the driver, ﬂashed each time an event was recorded) was disabled and no driver
coaching occurred. Immediately following the 4-week Baseline, the Intervention phase began. During the 13-week Interven-
tion phase, drivers drove an instrumented vehicle during their normal, revenue-producing deliveries. During this time, the
OBSM device recorded safety-related events as normal and the safety coaching program was enabled (i.e., the feedback light
was activated and safety managers followed the coaching protocol when necessary). As the Virginia Tech Transportation
Institute (VTTI) was the independent evaluator in this research, the procedures described below were limited to those per-
formed by VTTI.
2.1. Participants and setting
Carrier A was a long-haul carrier located in the Southeastern US that primarily delivered dry goods. A total of 50 drivers
had an OBSM device installed in their trucks (36 drivers completed data collection). A total of 46 drivers at Carrier A signed
an Informed Consent Form (ICF) that allowed researchers to send questionnaires to participating drivers. The mean age of
these 46 drivers was 44 years old (Range = 23 to 61 years old). Carrier B was a local/short-haul carrier located in the North-
western US that primarily delivered beverage and paper goods. A total of 50 drivers had an OBSM device installed in their
trucks (41 drivers completed data collection). A total of 30 drivers at Carrier B signed an ICF that allowed researchers to send
questionnaires to participating drivers. The mean age of these 30 drivers was 50 years old (Range = 27 to 71 years old).
Prior to the OBSM device being installed in the vehicles, drivers attended an initial project brieﬁng. The project brieﬁng
lasted approximately two hours and included details regarding the project, informed consent, how the OBSM technology
worked (i.e., feedback light, capture of videos, etc.), and the coaching process. Drivers indicated their interest in participating
in the study by signing the ICF. Fifty OBSM devices were installed in 50 trucks at both participating carriers. After the OBSM
systems were installed, drivers were instructed to make their normal, revenue-producing deliveries.
Prior to the start of the 13-week Intervention phase, safety managers attended a training seminar that lasted approxi-
mately three hours. The safety manager training seminar included details regarding the project, informed consent, how
OBSM technology worked, how to use the technology vendor’s software, and how to ‘‘coach’’ drivers by using the video data.
The coaching protocol included the following eight steps: (1) thoroughly review the video event, (2) review the driver’s pre-
vious safety-related events (if necessary), (3) play the video in the meeting with the driver, (4) explain your viewpoint
regarding the video in question, (5) keep the meeting objective and positive, (6) determine follow-up steps, and (7) docu-
ment the meeting. Moreover, safety mangers were instructed to not use the videos as a disciplinary tool (e.g., acknowledge
positive behaviors as well. Safety managers indicated their interest in participating in the study by signing the ICF. After the
safety manager training seminar, they had access to all the data collected by the OBSM device during the 13-week Interven-
tion phase and coached drivers (when necessary).
2.2.1. Data collection process
DriveCam (hereafter referred to as the technology vendor) was responsible for all data collection. The technology vendor’s
OBSM device had two camera views: (1) driver’s face view, and (2) forward-facing view. Fig. 1 displays the two camera views
captured by the OBSM device. The OBSM device had three accelerometers (y-, x-, and z-axis) that triggered an event to be
J.S. Hickman, R.J. Hanowski / Transportation Research Part F 14 (2011) 189–198 191
Author's personal copy
recorded. If the criterion was met or surpassed (e.g., greater than or equal to |0.5g|) the OBSM device saved 12 s of video (i.e.,
8 s prior to the criterion being met or surpassed and 4 s after). The threshold was determined by DriveCam based on their
prior experience with over 60,000 installed event recorders (i.e., minimize the possibility of missing possible safety-related
events; however, not so sensitive that the device records numerous spurious triggers). Note the threshold value remained
constant throughout data collection. The OBSM device would record video data if the truck exceeded the criterion during:
hard cornering, hard braking, hard acceleration, collision, and rough/uneven surface. As the video recording was trigger-
based (and not continuous), at-risk driving behaviors (or safe driving behaviors) would not be recorded unless these behav-
iors were performed 8 s before and/or 4 s after the criterion was exceeded.
These video and quantitative data were automatically sent to the technology vendor’s headquarters in San Diego, CA via
cellular transmission. Once these data were received at the technology vendor’s headquarters, a trained data analyst re-
viewed the event. The data analyst reviewed the event to determine if the event represented a valid safety-related event
or a spurious trigger value (e.g., hit a pothole in the street, driving on a bumpy road, etc.). Spurious events were not reduced;
however, the valid safety-related events were reduced by the trained data analyst. Data reduction involved reviewing the
video and recording the trigger type (e.g., hard brake), outcome (e.g., collision), root cause (distraction, poor awareness), de-
meanor (e.g., aggressive), risky behaviors (unbelted, following too close, trafﬁc violation, failed to keep an out, poor lane
selection, in others blind area), and adverse weather conditions (if necessary). Once the technology vendor received these
data, they were reviewed, reduced (i.e., data analysts marked the presence of speciﬁc variables pertaining to the event),
and uploaded to a server in approximately 24–48 h from the time an event was captured by the event recorder in the instru-
Data analysts went through a month-long training program prior to reviewing videos unsupervised. The ﬁrst two weeks
of training were instructional and involved the following topics: common safe driving practices/standards according to the
, safety ride-a-long in an equipped vehicle, utilizing the vendor’s behavior deﬁnitions, proprietary software/
components, two milestone tests, correctly analyzing videos and writing effective reviewer notes, and the milestone 3 test
(had to score > 80% to move to the next training module). The following two weeks of on-the-job training include the fol-
lowing: 16 h working with a mentor, assisting with analysis of various client videos (these events will be randomly selected
for quality control), and a ﬁnal certiﬁcation test (had to score > 90% to complete training program).
As part of the technology vendor’s standard review process, each data analyst went through statistical quality assurance
sampling. The vendor had a tool that randomly selected video clips from each data analyst to ensure a 97% quality rating
(based on a 95% conﬁdence interval). Quality assurance was completed daily on the safety–critical events reviewed by each
data analyst (compared to an expert data analyst who served as a quality assurance manager). Each day, data analysts re-
ceived a list of the reviewed safety-related events with their associated errors and successes. Data analysts attended weekly
meetings with their manager and quality assurance team to review their overall quality as well as to determine areas where
refresher training might be needed. The technology vendor also tracked the most common mistakes made by all data ana-
lysts; this information was used to improve their training guidelines and behavioral deﬁnitions. Data analysts had an inter-
rating agreement of 97% (<3% error rate) against their standards. Data analysts were blind to the before-after design.
Although all safety-related events were uploaded to the server for review, only those safety-related events that exceeded
a certain threshold (or ‘‘Event Score’’) were requested to be reviewed with the driver. Event Scores in the current study ran-
ged from 0 to 11 (e.g., 0 = collision; 3 = driver unbelted; and 11 = driver involved in a near-crash was talking on a cell phone
and unbelted). Although higher Event Scores reﬂect higher severity safety-related events, the technology vendor does not
score collisions due to liability concerns. Note that DriveCam considers the scoring process to be proprietary. Typically,
an Event Score P5 was marked to be reviewed by the safety manager with the driver present; however, it was ultimately
up to the safety manager which safety-related events were reviewed with the driver. Once on the server, VTTI personnel and
safety managers (only during the Intervention phase) had access to the data via the technology vendor’s proprietary software
Fig. 1. Front camera view from DriveCam’s OBSM device (Left) and driver’s face view (Right).
192 J.S. Hickman, R.J. Hanowski / Transportation Research Part F 14 (2011) 189–198
Author's personal copy
(which was accessible via the Internet). VTTI personnel checked the software each day and recorded the frequency of safety-
related events, severity, driving behaviors, date, driver#, and quantitative data.
2.2.2. Coaching during the Intervention phase
Drivers received immediate feedback from the feedback light on the OBSM device (i.e., the light ﬂashed each time the
accelerometer criterion was met and/or exceeded in the Intervention phase). Limited information was collected during
the Intervention phase on the coaching process followed by ﬂeet safety managers. As the aim of the current study was to
assess the safety beneﬁts of a commercially available OBSM system as employed in the ﬁeld, the research team made an
a priori decision to limit their interactions with ﬂeet safety managers. Fleet safety managers were instructed to follow the
coaching protocol, when necessary, as described above. A random sample of drivers that received a coaching session (as indi-
cated in the technology vendor’s software) was sent an In-Study Questionnaire to assess safety-manager’s adherence to the
coaching protocol described above (a total of ten items). A total of 10 and 14 drivers that received a coaching session at Car-
riers A and B, respectively, returned the In-Study Questionnaire. Table 1 shows selected results from the In-Study Question-
naire at Carriers A and B. The ﬁrst four items in Table 1 illustrate the coaching protocol: review the video with the driver,
clearly identify the ‘‘root cause’’ in the safety-related event, identify ways to prevent the safety-related event in the future,
and keep the coaching session positive. As can be seen in Table 1, ﬂeet safety managers at each carrier vastly differed in their
ability to follow the coaching protocol.
The primary analyses described below assessed the safety beneﬁts of the technology vendor’s OBSM program.
3.1. Hypothesis 1 at Carrier A
During the 4-week Baseline phase a total of 58 safety-related events were captured by the OBSM device (2 collisions and
56 risky driving events) from the 36 drivers that completed the study (14 drivers quit, resigned, withdrew, had a malfunc-
tioning OBSM device, and/or did not meet the criteria for inclusion in the analyses). Participating drivers needed at least
3 weeks of Baseline data and four weeks of Intervention data to be included in all analyses. These 36 drivers drove a total
of 291,969 miles during the Baseline phase. A rate was calculated to account for exposure (i.e., frequency of safety-related
events/10,000 miles traveled). The solid line in Fig. 2 displays the mean rate of safety-related events/10,000 miles per week
across the 17 weeks of data collection.
As shown in Fig. 2, the mean rate of safety-related events/10,000 miles during the Baseline phase was 1.98 (SD = 2.67)
safety-related events/10,000 miles. During the 13-week Intervention phase a total of 141 safety-related events were cap-
tured by the OBSM device (2 collisions and 139 risky driving events). These drivers drove a total of 1170,721 miles during
the Intervention phase. As shown in Fig. 2, the mean rate of safety-related events/10,000 miles traveled during the Interven-
tion phase was 1.23 (SD = 1.84) safety-related events/10,000 miles. A paired sample t-test found the 39.9% reduction in the
mean rate of safety-related events/10,000 miles traveled from the Baseline to Intervention phases was signiﬁcant
= 1.81, p= 0.039; Std Error = 0.415).
3.2. Hypothesis 1 at Carrier B
During the 4-week Intervention phase a total of 65 safety-related events were captured by the OBSM device (1 collision
and 64 risky driving events) from the 42 drivers that completed the study (8 drivers quit, resigned, withdrew, had a malfunc-
tioning OBSM device, and/or did not meet the criteria for inclusion in the analyses). These 42 drivers drove a total of 162,492
miles during the Baseline phase. As indicated above, a rate was calculated to account for exposure. The dashed line in Fig. 2
displays the mean rate of safety-related events/10,000 miles per week across the 17 weeks of data collection at Carrier B.
Selected results from the in-study driver questionnaire at Carriers A and B.
Item Carrier A Carrier B
(1) Reviewed video during coaching session 8 out of 10 (80%) 1 out of 14 (7%)
(2) How clearly was the ‘‘root cause’’ identiﬁed Mean = 7.1 out of 9.0 (moderately
Mean = 1.0 out of 9.0 (very unclear)
(3) Identiﬁed ways to prevent future events 9 out of 10 (90%) 1 out of 14 (7%)
(4) The coaching session was positive Mean = 6.3 out of 9.0 (positive) Mean = 3.0 out of 9.0 (moderately
(5) How likely are you to use the information learned in coaching
Mean = 7.8 out of 9.0 (moderately
Mean = 2.0 out of 9.0 (very unlikely)
(6) Length of coaching session Mean = 10 min Mean = 10 min
J.S. Hickman, R.J. Hanowski / Transportation Research Part F 14 (2011) 189–198 193
Author's personal copy
As shown in Fig. 2, the mean rate of safety-related events/10,000 miles traveled during the Baseline phase was 4.02
(SD = 9.73) safety-related events/10,000 miles. During the 13-week Intervention phase a total of 117 valid safety-related
events were captured by the OBSM device (2 collisions and 115 risky driving events). These same 41 drivers drove a total
of 615,403 miles. As shown in Fig. 2, the mean rate of safety-related events/10,000 miles traveled during the Intervention
phase was 1.93 (SD = 4.03). A paired sample t-test found the 52.2% reduction in the mean rate of safety-related events/
10,000 miles traveled from the Baseline to Intervention phases was signiﬁcant (t
= 1.88, p= 0.034; Std Error = 1.11).
Fig. 2. Weekly Mean rate of safety-related events/10,000 miles traveled across the baseline and intervention phases at Carriers A (solid line) and B (dashed
Fig. 3. Mean rate of severe safety-related events/10,000 miles traveled across the baseline and intervention phases at Carriers A and B.
194 J.S. Hickman, R.J. Hanowski / Transportation Research Part F 14 (2011) 189–198
Author's personal copy
3.3. Hypothesis 2 at Carrier A
At Carrier A there were a total of 16 severe safety-related events out of a total of 199 safety-related events (8.0%). Fig. 3
displays the mean rate of severe safety-related events/10,000 miles across the 17 weeks of data collection at Carrier A. The
mean rate of severe safety-related events/10,000 miles during the Baseline phase was 0.22 (SD = 0.615) severe safety-related
events/10,000 miles, and the mean rate of severe safety-related events/10,000 miles during the Intervention phase was 0.09
(SD = 0.256) severe safety-related events/10,000 miles. The 59.1% reduction in the mean rate of severe safety-related events/
10,000 miles from the Baseline to the Intervention phase was not signiﬁcant (paired sample t-test
= 1.19, p= 0.121; Std
Error = 0.11). Due to insufﬁcient statistical power the 2nd hypothesis was not supported at Carrier A. Note the power analysis
indicated that 30 drivers would be sufﬁcient to detect a signiﬁcant difference; however, this analysis did not consider severe
safety-related events. Nonetheless, despite the lack of a signiﬁcant reduction, a 59.1% decrease in the mean rate of severe
safety-related events/10,000 miles is noteworthy.
3.4. Hypothesis 2 at Carrier B
At Carrier B there were a total of 28 severe safety-related events out of a total of 182 safety-related events (15.4%). Fig. 3
also displays the mean rate of severe safety-related events/10,000 miles across the 17 weeks of data collection at Carrier B.
The mean rate of severe safety-related events/10,000 miles during the Baseline phase was 0.36 (SD = 1.11) severe safety-re-
lated events/10,000 miles, and the mean rate of severe safety-related events/10,000 miles during the Intervention phase was
0.2 (SD = 0.417) severe safety-related events/10,000 miles. The 44.4% reduction in the mean rate of severe safety-related
events/10,000 miles from the Baseline to the Intervention phase was not signiﬁcant (paired sample t-test
p= 0.16; Std Error = 0.16). As with the Carrier A results, the 2nd hypothesis was not supported at Carrier B. As indicated
above, this was due to limited statistical power. However, as with the Carrier A ﬁndings, a substantial reduction of 44.4%
in the mean rate of severe safety-related events/10,000 miles was observed at Carrier B. Although not signiﬁcant, due to
the small number of severe events, the percentage reduction in severe safety-related events at Carriers A and B have practical
In this quasi-experiment (i.e., lack of random assignment of drivers) the effectiveness of an OBSM system to decrease the
risky driving behaviors of local/short-haul and long-haul truck drivers was evaluated. Almost all prior behavioral safety re-
search has targeted work behaviors in settings where employees can systematically observe the safe versus at-risk behavior
of their coworkers (Geller, 1998; Krause, Hidley, & Hodson, 1996). However, employees who work in relative isolation or
have little oversight from a supervisor or peer require a process where objective data can be obtained to provide feedback
The current study addressed this limitation with a commercially available OBSM system. The technology vendor was
DriveCam. The technology vendor used in-vehicle video technology to record driver behavior. These video recordings were
used by ﬂeet safety managers to provide feedback on the safe and at-risk driving behaviors of participating drivers. During
the 4-week Baseline phase, the OBSM device recorded safety-related events. However, the feedback light on the OBSM device
was disabled and safety managers did not have access to the recorded safety-related events to provide feedback to drivers.
During the 13-week Intervention phase, the feedback light on the OBSM device was activated and safety managers had ac-
cess to the recorded safety-related events and followed the coaching protocol with drivers (when necessary).
Based on a total of 1462,590 and 777,895 miles driven at Carriers A and B, respectively, the OBSM system was effective in
decreasing the mean rate of safety-related events/10,000 miles traveled. At Carrier A, the mean rate of safety-related events/
10,000 miles traveled during the Intervention phase was signiﬁcantly reduced by 37% compared to the Baseline phase. Sim-
ilarly, drivers at Carrier B signiﬁcantly reduced the mean rate of safety-related events/10,000 miles traveled during the Inter-
vention phase by 52.2% compared to the Baseline phase. Thus, the ﬁrst hypothesis was supported.
In interpreting these results, eight issues are noteworthy. First, although it appears Carrier B had superior decreases to
Carrier A in the mean rate of safety-related events/10,000 miles traveled (based on percentage reduction), concluding differ-
ential intervention impact is risky because Carrier A drove more safely than Carrier B during the Baseline phase (1.9 versus
4.0 safety-related events/10,000 miles traveled). For example, Carrier A and B likely experienced different safety-related
environmental conditions due to the predominant roads driven. A naturalistic study by Hanowski, Keisler, and Wierwille
(2004) reported that long-haul drivers typically drive on rural divided roads (e.g., highways), and local/short-haul drivers
typically driver on urban divided roads.
Second, drivers were aware the instrumented vehicles were recording their driving behaviors; thus, they may have al-
tered their performance accordingly (i.e., subject reactivity). However, it is unlikely this awareness inﬂuenced intervention
impact as any reactivity to the OBSM devices or the OBSM safety process was constant across both phases, and any effect of
J.S. Hickman, R.J. Hanowski / Transportation Research Part F 14 (2011) 189–198 195
Author's personal copy
reactivity to being observed is likely to be most prominent at the beginning of such procedures (Campbell, 1957). In fact, the
data obtained during the Baseline phase may have been understated, resulting in a less robust effect during the Intervention
phase. If this was the case, note that OBSM devices were installed in-vehicles at Carriers A and B several weeks prior to the
start of data collection. Thus, drivers would have become familiar to the presence of the OBSM devices by the time data col-
Third, technical difﬁculties with the OBSM devices at Carrier A likely had an adverse effect on the impact of the interven-
tion. A total of thirteen OBSM devices at Carrier A experienced technical difﬁculties. Initially, it was believed the technical
issues prevented the OBSM device from transmitting stored events to the technology vendor (note these are referred to
as events as they had not been reduced). Thus, it was assumed the stored events could be retrieved once they were returned
to the technology vendor (i.e., there would be no loss of data). The study continued and new OBSM devices were installed as
soon as they were received. Some of the driver’s truck received an OBSM device during the Baseline phase; other driver’s
trucks received a new OBSM device during the Intervention phase.
Unfortunately, the technical issue was more signiﬁcant than initially suspected; almost all of the stored events in the mal-
functioning OBSM devices were lost. Those driver’s trucks that did not receive an OBSM device until the Intervention phase
were excluded from the analyses as they did not have any baseline driving data. Those driver’s trucks that received a OBSM
device during the Baseline phase were included in the analyses; however, these driver’s trucks were missing baseline driving
data. These missing events would have likely increased the mean rate of safety-related events/10,000 miles traveled during
the Baseline phase, thereby increasing the impact of the OBSM intervention at Carrier A.
Fourth, although evidence suggests that safety managers at Carrier B were not strictly adhering to the coaching protocols,
drivers were able to signiﬁcantly reduce the mean rate of safety-related events/10,000 miles traveled from the Baseline
phase to the Intervention phase. Why? In fact, all drivers at Carrier B received some feedback, though informally, through
messages via a dispatching device (e.g., please obey the company’s safety belt policy). Apparently, this was sufﬁcient to alter
the driving behaviors at Carrier B. However, the high rate of sabotage casts some doubt on the results at Carrier B (a total of
278 events at Carrier B; 2 and 276 events in the Baseline and Intervention phases, respectively). As the driver face camera
was blocked, data reductionists could not discern driver behaviors; thus, in-cab driver behaviors were unknown during these
events. Although out-of-cab driving behaviors, such as following too close, could be seen in these situations (and were re-
duced as such), drivers’ in-cab behaviors could not be seen (e.g., cell phone use, asleep at the wheel, driver unbelted, etc.). It
is likely that some of these in-cab, at-risk driving behaviors were ongoing during these events; however, data reductionists
were blind to their occurrence as the driver face camera was obstructed.
Fifth, the results replicate those found by McGehee et al. (2007), who also used a DriveCam system with novice teen driv-
ers. However, there were some critical differences between the current study and McGehee et al. (2007). The participants in
McGehee et al. (2007) were novice teen drivers. As indicated above, these novice drivers were still acquiring basic knowledge
and skills in regards to driving a motor vehicle, whereas the participants in the current study were experienced professional
drivers (the mean rate of safety-related events/10,000 miles was tenfold greater with the novice teen drivers compared to
the experienced professional drivers in the current study). McGehee et al. (2007) also included social comparison feedback
during the intervention (i.e., feedback on the performance of the other teen drivers in the study); this was not part of the
coaching protocol in the current study. Williams and Geller (2000) found that the addition of social comparison feedback
regarding workers’ safety performance was sufﬁcient at improving safety-related behaviors beyond levels attained with
individual behavioral feedback alone. The authors explained their unexpected ﬁnding by assuming that social comparison
feedback added a motivational element to select the safe alternatives. Although the individual feedback provided the teen
drivers with speciﬁc information on safe versus at-risk driving behaviors, the public accountability provided by general com-
parisons with similar teen drivers provided an effective motivational consequence.
Sixth, the current study relied on the power of feedback to alter drivers’ at-risk driving behaviors. The delivery of feedback
to increase safety-related work behaviors has received great attention in the organizational behavior management literature
(Fellner & Sulzer-Azaroff, 1984; Ludwig & Geller, 1991; Ludwig & Geller, 1997; Reber & Wallin, 1984; Williams & Geller,
2000). However, the motivational effects of feedback have been questioned. Bandura (1986) suggested that dissatisfaction
with one’s prior attainments can motivate increased effort and vigilance. Without goals people do not have a standard to
compare prior behavior; thus, self-evaluative reactions are not engaged. Without feedback people do not have information
allowing them to gauge progress toward the goal. Therefore, goals or feedback alone do not activate self-regulatory processes
(Bandura & Cervone, 1983; Cervone & Wood, 1995).
A major weakness in studies using feedback as an intervention component is their failure to assess goal setting. A goal is
an object, aim, or endpoint of action that describes what people are trying to accomplish. Several reviews and meta-analyses
have supported the basic tenets of goal-setting theory (Locke & Latham, 1990; Tubbs, 1986; Wood, Mento, & Locke, 1987).
The basic theory proposes that goals and performance have a linear relationship (i.e., higher goals lead to higher perfor-
mance). Locke and Bryan (1969) had participants drive an instrumented car over a prescribed course for three trips. During
trips two and three, goals were assigned to different driving behaviors. However, participants received feedback on all ﬁve
driving-related behaviors. Participants only improved on those driving behaviors for which the experimenter assigned goals.
Several reviews of the goal-setting literature have supported the interdependent relationship of goals and feedback (Locke &
Latham, 1990; Tubbs, 1986; Wood et al., 1987). Locke and Latham (1990) hypothesized that participants in feedback-only
interventions reporting beneﬁcial behavior change were spontaneously setting goals. They further state, ‘‘The unmistakable
message of such interventions [feedback] must be, here is something which you should improve!’’ (p. 196).
196 J.S. Hickman, R.J. Hanowski / Transportation Research Part F 14 (2011) 189–198
Author's personal copy
The current study made no attempt to assess drivers’ goals and/or provide goal-setting training. As VTTI was the indepen-
dent evaluator, and the purpose of the study was to assess a commercially available OBSM system, no attempt was made to
revise the existing safety program. Moreover, assessment of drivers’ goal-setting behavior in the current study would have
introduced an element not present in the existing program, thereby providing an inaccurate assessment of the technology
vendor’s program. This is something that should be addressed in future studies assessing the efﬁcacy of OBSM systems.
Seventh, the ‘‘safety culture’’ at each carrier was different and these differences likely inﬂuenced drivers’ perceptions of
the OBSM program. Many safety experts have suggested that an organization’s safety culture is critical for institutionaliza-
tion of safety processes and maintenance of intervention effect (Geller, 2001; Krause, 1997; McSween, 1995). Safety culture
has remained a popular topic in the safety literature since the Chernobyl disaster in 1986 (Mearns, Flin, Gordon, & Fleming,
1998). The exact deﬁnition of safety culture has been widely debated (Clarke, 1999); however, a broad deﬁnition of an orga-
nization’s safety culture includes both: (i) the attitudes, values, norms and beliefs with respect to risk and safety and (ii) the
visible practices and procedures and rewarded behaviors that characterize an organization (Geller, 2001). Fleets considering
an OBSM program should judge their safety culture prior to implementing any safety intervention, especially an intervention
that monitors driver behavior via video cameras. Drivers may view this as an invasion of their privacy and/or a ‘‘blame the
Eighth, only a portion of the trucks at each location were instrumented with OBSM devices (as the current study was a
pilot). Safety managers decided a priori which trucks were instrumented with OBSM devices. This (and the limited time
frame) limited the ability to assess the potential ‘‘ﬂeet-wide’’ safety beneﬁts of the OBSM system.
As the goal of the current study was to assess the safety beneﬁts of a commercially available OBSM program under nor-
mal, revenue-producing conditions; thus, no attempt was made to signiﬁcantly deviate from the technology vendors existing
OBSM program. Prior research has found the combination of goal setting and feedback to be the optimal approach, future
studies that use an OBSM system should consider the addition of goal-setting training and directly assessing participants’
goals (i.e., asking participants to indicate their speciﬁc improvement goal, ‘‘I will increase my safety score by 10%’’). The cur-
rent study did not assess implicit goal setting; thus, variations in goal setting among drivers could have been the reason for
differential behavior change among drivers.
The current OBSM program was successful in signiﬁcantly reducing the mean rate of safety-related events/10,000 miles
(by 37% and 52.2% at Carriers A and B, respectively); however, management may be more inclined to adopt an OBSM pro-
gram that shows an advantageous return-on-investment. Such an assessment should include the costs associated with
implementing and maintaining the OBSM program, as well as the direct (damage, health care, etc.) and indirect (legal fees,
etc.) costs associated with reduced crashes and violations. Future studies should also assess the potential ﬂeet-wide safety
beneﬁts of an OBSM program. Data collection should last at least one year (possibly longer) and all trucks at the terminal
location should be instrumented with data collection equipment. This will limit selection bias and allow an appropriate
assessment of the safety beneﬁts of the OBSM program.
This research was funded by the FMCSA under Contract # DTMC75-07-D-00006 (Task Order #1). Mr. Olu Ajayi was the
Task Order Manager and Dr. Martin Walker was the Contracting Ofﬁcer Technical Representative. They provided much
appreciated advice and support throughout the project. The authors thank Dr. Walker and Mr. Ajayi for their insight, careful
reviews, and prompt attention to administrative matters. Additionally, this project would not have been possible without the
participation of the trucking ﬂeets and drivers that participated in the current study.
Bandura, A. E. (1986). Social foundations of thought and action: A social cognitive theory. Upper Saddle River, NJ: Prentice Hall.
Bandura, A. E., & Cervone, D. (1983). Self-evaluative and self-efﬁcacy mechanisms governing the motivations effects of goal systems. Journal of Personality
and Social Psychology, 45(5), 1017–1028.
Campbell, D. T. (1957). Factors relevant to the validity of experiments in social settings. Psychological Bulletin, 54, 279–312.
Cervone, D., & Wood, R. (1995). Goals, feedback, and the differential inﬂuence of self-regulatory processes on cognitively complex performance. Cognitive
Therapy & Research, 19(5), 519–545.
Clarke, S. (1999). Perceptions of organizational safety: Implications for the development of safety culture. Journal of Organizational Behavior, 20, 185–198.
Federal Motor Carrier Safety Administration. (2006). Report to congress on the large truck crash causation study (LTCCS). MC-R/MC-RRA, [No Publication
Fellner, D. J., & Sulzer-Azaroff, B. (1984). Increasing industrial safety practices and conditions through posted feedback. Journal of Safety Research, 15, 7–21.
Geller, E. S. (1998). Understanding behavior-based safety: Step-by-step methods to improve your workplace (2nd ed.). Neenah, WI: J. J. Keller & Associates, Inc..
Geller, E. S. (2001). The psychology of safety handbook. Boca Raton, FL: CRC Press.
Geller, E. S., & Clarke, S. W. (1999). Safety self-management: A key behavior-based process for injury prevention. Professional Safety, 44(7), 29–33.
Hanowski, R. J., Keisler, A. S., Wierwille, W. W. (2004). Light vehicle-heavy vehicle interactions: A preliminary assessment using critical incident analysis. Report
No. FMCSA-RT-04-004. Washington, DC: Federal Motor Carrier Safety Administration.
Hickman, J. S., & Geller, E. S. (2003). A self-management for safety intervention for short-haul truck drivers. Journal of Applied Behavior Analysis, 23(4), 1–20.
Krause, T. R. (1997). The behavior-based safety process: Managing involvement. New York, NY: Van Nostrand-Reinhold.
J.S. Hickman, R.J. Hanowski / Transportation Research Part F 14 (2011) 189–198 197
Author's personal copy
Krause, T. R., Hidley, J. H., & Hodson, S. J. (1996). The behavior-based safety process: Managing involvement for an injury-free culture (2nd ed.). New York: Van
Krause, T. R., Robin, J. L., Knipling, R. R. (1999). The Potential Application of Behavior-Based Safety in the Trucking Industry. Federal Highway Administration,
Locke, E. A., & Bryan, J. F. (1969). The directing function of goals in task performance. Organizational Behavior and Human Performance, 4, 35–42.
Locke, E., & Latham, G. (1990). A theory of goal setting and task performance. New Jersey: Prentice-Hall.
Ludwig, T. D., & Geller, E. S. (1991). Improving the driving practices of pizza deliverers: Response generalization and moderating effects of driving history.
Journal of Applied Behavior Analysis, 24, 31–34.
Ludwig, T. D., & Geller, E. S. (1997). Managing injury control among professional pizza deliverers: Effects of goal-setting and response generalization. Journal
of Applied Psychology, 82(2), 253–261.
Mahoney, M. J. (1996). The self-management of covert behavior: A case study. Behavior Therapy, 2, 75–78.
Mahoney, M. J. (1972). Research issues in self-management. Behavior Therapy, 3, 45–63.
McGehee, D. V., Raby, M., Carney, C., Lee, J. D., & Reyes, M. L. (2007). Extending parental monitoring using an event-triggered video intervention in rural teen
drivers. Journal of Safety Research, 38, 215–227.
McSween, T. E. (1995). The values-based safety process: Improving your safety culture with a behavioral approach. New York, NY: John Wiley & Sons, Inc..
Mearns, K., Flin, R., Gordon, R., & Fleming, M. (1998). Measuring safety climate on offshore installations. Work and Stress, 12(3), 238–254.
Olson, R., & Austin, J. (2001). Behavior-based safety and working alone: The effects of a self-monitoring package on the safe performance of bus operators.
Journal of Organizational Behavior Management, 21(3), 5–43.
Treat, J. R., Tumbas, N. S., McDonald, S. T., Shinar, D., Hume, R. D., Mayer, R. E., et al. (1979). Tri-level study of the causes of trafﬁc accidents: Final report volume
I: Causal factor tabulations and assessments. Institute for Research in Public Safety, Indiana University, DOT HS-805 085.
Reber, R. A., & Wallin, J. A. (1984). The effects of training, goal-setting and knowledge of results on safe behavior: A component analysis. Academy of
Management Journal, 27, 544–560.
Tubbs, M. E. (1986). Goal setting: A meta-analytic examination of the empirical evidence. Journal of Applied Psychology, 71(3), 474–483.
Watson, D. L., & Tharp, R. G. (1993). Self-directed behavior: Self-modiﬁcation for personal adjustment (6th ed.). Paciﬁc Grove, CA: Brooks/Cole Publishing
Williams, J. H., & Geller, E. S. (2000). Behavior-based intervention for occupational safety: Critical impact of social comparison feedback. Journal of Safety
Research, 31(3), 135–142.
Wood, R. E., Mento, A. J., & Locke, E. A. (1987). Task complexity as a moderator of goal effects: A meta-analysis. Journal of Applied Behavior, 72(3), 416–425.
198 J.S. Hickman, R.J. Hanowski / Transportation Research Part F 14 (2011) 189–198