CHAPTER 1: Light Vehicle - Heavy Vehicle Interactions Collected in the 100-CAR Study

Introduction

Overview of the Light Vehicle-Heavy Vehicle Problem

Truck crashes represent a significant problem on our highways. In 2002, 434,000 large trucks (gross weight > 10,000 lbs) were involved in vehicle crashes; 4,542 of these crashes resulted in a fatality. A total of 4,897 people died and an additional 210,000 were injured. Large trucks accounted for 4% of all registered vehicles in 2002, yet represented 8% of all vehicles involved in fatal crashes (National Highway Traffic Safety Administration, NHTSA, 2003). Truck crashes and their associated injuries and fatalities cost an estimated 24.4 billion in direct and indirect costs in 2002 (FMCSA, 2002).

The disproportionate number of vehicles to fatalities among large trucks is likely to contribute to the perception that truck drivers are irresponsible. However, these data do not signify that truck drivers are necessarily the problem. In fact, truck drivers have lower non-fatal crash rates per million vehicle miles traveled than light vehicles (NHTSA, 2003). However, light vehicles are extremely vulnerable when they interact with trucks because trucks often weigh 20-30 times as much as light vehicles (Insurance Institute for Highway Safety, 2002), and trucks take 20-40% farther to stop than light vehicles (NHTSA, 1987). This is best illustrated by the fact that over three-fourths of multiple vehicle fatal truck crashes resulted in the occupant(s) of the other vehicle being killed (NHTSA, 2004).

To combat this problem, proposals have been made to separate light and heavy vehicles on high-volume roads. For example, STAR Solutions has proposed to separate heavy trucks from passenger traffic on Interstate 81 (http://www.virginiadot.org/infoservice/resources/is-I-81-Star-exec.pdf), thereby reducing the likelihood of light vehicle-heavy vehicle (LV-HV) interactions. However, the enormous cost and logistical difficulties associated with new and modified road construction suggest that, in most cases, these vehicles will have to share the road for the foreseeable future. Thus, a better understanding of LV-HV interactions is needed to develop alternative interventions and countermeasures directed at reducing and/or eliminating the problem.

Prior Research on Light Vehicle-Heavy Vehicle Interactions

When Hanowski, Wiewille, Gellatly, Early, and Dingus (1998) conducted focus groups with local/short-haul truck drivers, they found that participants ranked "problems with light vehicles" as the most important safety issue. In fact, this was the only safety issue that was consistently cited in all 11 focus groups that were conducted. Similarly, Neale et al. (1998) found that LV-HV interactions were a significant safety concern among a sample of long-haul truck drivers. Only recently has empirical evidence supported truck drivers' claims.

Blower (1998) analyzed the University of Michigan Transportation Research Institute's "Trucks Involved in Fatal Accidents" database for all two-vehicle, truck-passenger vehicle fatal crashes in 1994 and 1995 (n = 5,453). He found that truck drivers were cited with a driver-related factor in 26.5% of the fatal crashes, while passenger vehicle drivers were cited in over 80% of the fatal crashes. The passenger vehicle driver was the only driver cited in 70.3% of the fatal crashes, while truck drivers were the only driver cited in 16.2% of the fatal crashes.

Stuster (1999) found similar results when he reviewed the U.S. Department of Transportation's Fatality Analysis Reporting System. He found that truck driver-related factors were cited in 29% of fatal truck crashes involving a passenger vehicle, while 67% of these same interactions were cited as passenger vehicle-related. Moreover, Wang, Knipling, and Blincoe (1999) found that LVs were the initiators in LV-HV fatal crashes by a ratio of approximately 3:1. Thus, it appears the actions of LVs are responsible for a substantial amount of the fatal LV-HV interactions.

Council, Harkey, Nabors, Khattack, and Mohamedshah (2003) took a different approach when they analyzed 16,264 LV-HV interactions from the North Carolina Highway Safety Information System. Rather than examining the police reports from fatal LV-HV interactions, they examined the police reports of both crashes and fatal crashes. While the prior studies assessed the most severe crashes, the Council et al. (2003) study assessed the overall LV-HV crash picture. Contrary to the other studies, however, Council et al. (2003) found that the truck driver was assigned fault in 48% of the crashes, while the passenger vehicle driver was assigned fault in 40.2% of the crashes (8.9% of the crashes were assigned fault to both drivers, while 2.9% were assigned fault to neither driver). The Council et al. (2003) data suggests that HV drivers were responsible for the majority of the LV-HV interactions (for all types of crashes). Thus, there appears to be some inconsistencies in the literature regarding assigned culpability in LV-HV interactions.

These prior studies assessed LV-HV interactions by examining vehicle crash databases that rely on police reports and crash reconstruction. These approaches are generally reliable, but they do have limitations, including witnesses and crash participants can be biased and report conflicting stories; police officers, while often experienced, generally do not receive extensive training in crash reconstruction; and witnesses or crash participants that were severely injured or killed in the crash are unlikely or are unable to effectively persuade the police officer about their side of the crash (also referred to as the "surviving driver" hypothesis).

Blower (1998) acknowledged the surviving driver limitation in his discussion and compared fatal LV-HV interactions with respect to driver survivability. When only the truck driver survived the fatal LV-HV interaction, the LV driver was cited with at least one driver-related factor in 81.9% of the fatal crashes, while the HV driver was cited in only 24.1% of the fatal LV-HV crashes. Conversely, when only the LV driver survived the fatal LV-HV interactions, the LV drivers were cited with at least one driver-related factor in 46.7% of the crashes, while HV drivers were cited in 57.7% of the crashes.

It would appear that driver survivability does affect which driver is cited in fatal LV-HV crashes. This makes intuitive sense because the surviving driver is able to report their biased account of the event. However, when both LV and HV drivers survive the crash, the LV driver was cited in 74.1% of the crashes while the HV driver was cited in only 35.5% of the crashes (Blower, 1998).

The crash database approach does not necessarily shed light on the full variety of LV-HV interactions because they rely solely on crashes and fatal crashes. While LV drivers have been shown to be culpable in a significant proportion of LV-HV interactions (Wang, Knipling, and Blincoe, 1999), we do not know why. An alternative approach, and the method used in the current study, is to study the pre-event behaviors of all LV-HV interaction critical incidents, including crashes, near-crashes, and crash-relevant conflicts.

Because the focus of this research is on analyzing critical incidents, it is important to define the three categories that are of most interest: crashes, near-crashes, and crash-relevant conflicts. In the 100-Car Study, Dingus et al. (2004) defined crashes, near-crashes, and crash relevant conflicts as follows:

Crash: Any contact with an object, either moving or fixed, at any speed in which kinetic energy is measurably transferred or dissipated, and includes other vehicles, roadside barriers, objects on or off of the roadway, pedestrians, cyclists, or animals.

Near-Crash: Any circumstance that requires a rapid, evasive maneuver by the subject vehicle, or any other vehicle, pedestrian, cyclist, or animal to avoid a crash. A rapid, evasive maneuver is defined as a steering, braking, accelerating, or any combination of control inputs that approaches the limits of the vehicle capabilities. As a guide: Subject vehicle braking >0.5 g or steering input that results in a lateral acceleration >0.4 g to avoid a crash constitutes a rapid maneuver.

Crash-Relevant Conflict (Incident): Any circumstance that requires a crash avoidance response on the part of the subject vehicle, any other vehicle, pedestrian, cyclist, or animal that is less severe than a rapid evasive maneuver (as defined above), but greater in severity than a "normal maneuver" to avoid a crash. A crash avoidance response can include braking, steering, accelerating, or any combination of control inputs. A "crash avoidance response" for the subject vehicle is defined as a control input that falls outside of the 99% confidence limit for control input as measured for the same subject.

Hanowski, Keisler, and Wierwille (2004) assessed two on-road in situ data collection efforts, one involving local/short-haul (L/SH) trucking and the other long-haul trucking with drivers who used sleeper berths, to examine critical incidents that occurred between LVs and HVs. In this study, critical incidents were defined as crashes and near-crashes. Near-crashes were events resulting in a close call or requiring rapid action by a driver to avoid a crash. Video and non-video data collected during the two studies were used to characterize 210 critical incidents involving LV-HV interactions. Of the 210 critical incidents analyzed in the Hanowski, Keisler, and Wierwille (2004) study, 78% were assessed to have been initiated by the LV driver, while the remaining 22% were initiated by the HV driver. It should be noted that in Hanowski, Keisler, and Wierwille (2004), "initiate" is synonymous with "at-fault." Thus, a vehicle that initiated an incident is meant to reflect the vehicle that was at-fault or responsible for the incident.

The benefits of the naturalistic data collection approach used in Hanowski, Keisler, and Wierwille (2004) are three-fold: (1) video and other supporting data are collected before, during, and after the event occurs, thereby providing a complete picture of the incident as it unfolds; (2) various types of non-crash LV-HV interactions can be analyzed; and (3) the use of video and non-video data allowed one to make objective assessments on the critical reason(s) for the incident (rather than incomplete, subjective police reports).

However, one limitation of this approach was that the video cameras were only installed in the HVs and not the LVs. Therefore, Hanowski, Keisler, and Wierwille (2004) were only able to assess LV-HV interactions from the HV driver's perspective. Thus, it is possible they missed critical incidents that were only apparent from the LV driver's perspective. Furthermore, the lack of instrumentation in LVs limits the understanding of the LV driver's behavior during the incident.

These limitations were addressed in the 100-Car Study by installing video cameras on LVs (Dingus et al., 2004). All identified LV-HV interactions from the 100-Car data set were included in the current analyses. Together, results from the current study and the Hanowski, Keisler, and Wierwille (2004) study may provide a more complete picture of the LV-HV interaction problem.

The current study used two classification methodologies to assess all LV-HV interactions: the classification methodology used in Hanowski, Keisler, and Wierwille (2004) (originally developed by Wierwille et al., 2001), and the methodology and terminology from the Large Truck Crash Causation Study (LTCCS) (Thieriez, Radja, and Toth, 2002). Thus, all LV-HV interactions in the current data set were coded with two similar, yet distinct, classification approaches. The primary advantage of this method is that reliable and valid comparisons can be made with both prior and future research studies using either approach.

Research Goals

The data from the 100-Car Study (Dingus et al., 2004) were used in the current project to assess the LV-HV interaction problem from the LV drivers' perspective. There were four primary aims in the current effort:

• Gain a better understanding of LV-HV interactions on our nation's roadways.
• Continue to develop the classification scheme and corresponding Contributing Factors list for LV-HV interactions used in Hanowski, Keisler, and Wierwille (2004) and use the terminology and methodology described in the LTCCS (Thieriez, Radja, and Toth, 2002).
• Compare the current data to the data obtained in the Hanowski, Keisler, and Wierwille (2004) study for a more complete picture of the LV-HV interaction problem.
• Provide background information that would serve as a necessary prerequisite to the development of countermeasures for LV-HV interactions.

Method

Participants and Setting

One hundred participants who commute to and from the Washington, DC metro area were initially recruited as drivers in the 100-Car Study (Dingus et al., 2004). As some participants had to be replaced for various reasons (e.g., dropped out of the study because they moved from the area), the final number of participants was 109. Age and number of miles driven annually were used to select a subject population that would increase the probability that rear-end collisions would occur in 13 months of data collection. High mileage drivers were selected to increase the number of vehicle miles traveled per year (increase exposure). Greater number of younger drivers (ages 18 through 25) were recruited as they are overly represented in rear-end collisions as compared to other age groups. Also, more males than females were recruited since males are overly represented in rear-end crashes (Knipling, Wang, and Yin, 1993). It should be noted that participants were recruited from all age groups and that the target average annual mileage per year was approximately 27,000 miles/year. However, the actual mileage driven by participants in the 100-Car Study did not match their self-reported annual mileage prior to the study. The actual mileage of participants in the 100-Car Study can be found in Dingus et al. (2004). Table 1, shown below, displays the age and gender distribution of participants.

Table 1. Participant Age and Gender Distributions.

Light Vehicle Types

The data that were collected in the 100-Car Study came from six makes/models/years of LVs, including Toyota Camry (1997-2001), Toyota Corolla (1993-2002), Ford Explorer (1995-2000), Ford Taurus (2000-2002), Chevrolet Malibu (2002), and Chevrolet Cavalier (2002). The Toyota and Ford models were chosen based on recent sales figures and on the number of vehicles available in the Washington, DC area. Figures 1 and 2 show examples of the LVs used in the 100-Car Study.

Figure 1 . Toyota Camry and Toyota Corolla Used in 100-Car Study.

Figure 2 . Ford Explorer and Ford Taurus Used in 100-Car Study.

A total of 20 Chevrolet vehicles (10 Malibus and 10 Cavaliers) were leased from the Virginia Tech Motor Pool and were instrumented with data collection equipment. Twenty participants were given leased vehicles to drive for one year. The additional 80 vehicles (comprised of the aforementioned Toyota and Ford models) were the participants' personal vehicles. These vehicles were instrumented with the identical data collection systems as the leased vehicles.

Data Collection Methodology for the 100-Car Study

A full description of the research methodology used in the on-road portion of the 100-Car Study can be found in Dingus et al. (2004). Because the data used in the current effort consisted of the video recordings of critical incidents, the primary methodological considerations to be described in this report are those related to the video systems.

Video Camera Systems

As shown in Figure 3, five video cameras were used in the video recording system: (1) a forward-looking camera that captured the forward roadway scene, traffic situation, and possible incidents; (2) a driver's face camera that was used to record facial expressions, eyelid closure, glance position, and head turns; (3) a right-side camera was mounted on the A-pillar of the passenger side and faced outward; (4) a dome camera was mounted from inside the vehicle and faced over the driver's shoulder towards the steering wheel, hands, and feet; and (5) a rear camera that was intended to capture the situation behind the vehicle. Infrared lighting was used to illuminate the vehicle cab so that the driver's face as well as their hands could be viewed by the camera during nighttime driving.

Figure 3 . The Five Camera Views Recorded in the Instrumented Vehicle.

The video camera arrangement shown in Figure 3 had several advantages. First, it provided good coverage around the vehicle so that incidents could be captured as they developed. Second, the driver's facial expression, approximate glance direction, and approximate level of eye closure were also captured. Third, the arrangement provided appropriate views, whether moving forward or backward.

The five camera images were multiplexed into a single image as shown in Figures 4 and 5. Note that the right side camera and the rear camera were presented in the lower left quadrant in a split arrangement.

Figure 4 . Diagram of the Multiplexed Camera Views.

Figure 5 . Split-Screen Presentation of the Five Camera Views.

Video Recording Operation

Video recording was tied to the booting/powering system and it began to operate 2 minutes after the ignition was on. It also shut down in an orderly manner when the ignition was turned off. It was desired for the recording system to record for as long as possible without requiring technician/researcher attention. Therefore, multiple recorders designed to operate in sequence were used. The video continuously recorded while the ignition was on, thereby allowing laboratory review and selection of the video without losses of any kind.

The videotaped episodes/incidents were selected and keyed to digitally recorded data. In some cases, the videotape timestamp was used to access the corresponding digital data. In other cases, the incident flags (described later) in the digital data were used to access the corresponding video. Therefore, there was a straightforward keying procedure that allowed both kinds of access to take place efficiently.

Data Collection and Storage

"Chase" vehicles drove to pre-determined locations (e.g., parking lots) and downloaded the data from the experimental vehicles via a data transfer cable that connected to an outlet located near the rear license plate. Each chase vehicle had a laptop computer with a large hard drive to store all vehicle data. After each download from the experimental vehicles, the success of the duplication procedure was verified. Once 2.3 GB of data were downloaded from experimental vehicles, the data were copied to a DVD and verified. This DVD was duplicated; one copy was sent to VTTI and the other copy was kept in Northern Virginia.

As the data arrived at VTTI, they were downloaded to VTTI's network attached storage (NAS) and saved. Once the data was safely copied to the networked attached storage at VTTI and quality checks were performed, the data were then remotely deleted from the experimental vehicle hard drive.

Incident Flag

A critical incident involves an unexpected event resulting in a close call or requiring rapid action (evasive maneuver) on the part of a driver to avoid a crash. Critical incidents were detected by one of three methods. The first method involved flagging events where the car sensors exceeded a specified value. An example of this is a braking response of >0.6 g would be recorded as a potential incident where the driver may have braked in a panic.

Table 2 lists all of the triggers and levels that were used in this first method. The second incident flagging method occurred when the driver pressed an incident pushbutton located on the dashboard. Drivers were instructed to depress a button on the dashboard (after the event, not during the event) if they witnessed an incident or were involved in an incident. The third method of detecting incidents was through analysts' judgments when reviewing the video. Note that the video systems were operational as long as the ignition was turned on. In identifying incidents, analysts looked through epochs flagged from either of the first two methods and could flag additional events within the epoch (termed "user triggered") if an incident was detected visually. Only those events that involved a LV-HV interaction are described in the current analyses. The results of other project analyses can be found in Dingus et al. (2004).

Table 2. Triggers and Their Levels Used to Identify Critical Incidents in the 100-Car Study Database.

The incident flags (associated with the first and second triggering methods) were computed and detected on-line (as well as stored) with the flag appearing in the video. Since the entire video recording was reviewed, the presence of flags served as an indicator to the analyst of the high likelihood, but not certainty, of an incident occurrence. However, the analyst was also mindful of the possibility of incidents without flags and reviewed the tapes accordingly. The data analysts watched 90 s epochs (1 min prior and 30 s post incident) of each driving incident and recorded the information shown in Table 3.

Table 3. Information Recorded During 90 s Epoch Analysis.

Data Reduction Reliability

Given that data analysts were asked to perform subjective judgments on the video and driving data, training procedures were implemented to improve both inter- and intra-rater reliability. Reliability testing was then conducted to measure the resultant inter- and intra-rater reliability. First, data analyst managers performed spot checks of the data analysts' work, monitoring both event validity judgments as well as recording all database variables. All data analysts also performed 30 mins of spot-checking of their own or other data analysts' work per week.

To determine how successful these techniques were, an inter- and intra-rater reliability test was conducted during the last three months of data reduction. Three reliability tests were developed (each containing 20 events) for which the data analyst was required to make validity judgments. In each of the three reliability tests, three of the 20 events were also fully reduced by the data analysts. Three of the test events on Test 1 were repeated on Test 2 and three other events were duplicated between Tests 2 and 3, to obtain a measure of intra-rater reliability.

The Kappa statistic was also used to calculate inter-rater reliability (Cicchetti and Feinstein, 1990). The Kappa coefficient (K = 0.65, p < 0.0001) indicated that the association among raters was significant. The average of the pair-wise correlation coefficients for the inter-rater analysis was 0.86. The coefficients for the intra-rater analysis were extremely high with nine raters achieving a correlation of 1.0 among the three reliability tests and five raters achieving a correlation of 0.99. Given these three methods of calculating inter-rater reliability, it appears that the data analysts training coupled with spot-checking and weekly meetings proved to be an effective method for achieving high inter-rater and intra-rater reliability.

Strengths and Limitations of the Methodology Used

All research approaches have strengths and limitations. Listed below are the strengths and limitations of the approach used in the current study.

Strengths

The primary strength of the approach used in this study was that all driver behaviors, visible by way of the video camera, were recorded whenever the vehicle was on and in motion. This information is vital in developing an understanding of the incident, the events leading up to the incident, and the aftermath of the incident from the LV driver's viewpoint. The video camera arrangement described allowed researchers to watch the critical incidents unfold from multiple camera views. The video camera system that was used not only afforded an opportunity to understand what happened, but in many cases why it happened. A second advantage of this approach was that multiple cameras views helped ensure that any critical incidents involving the LV driver were captured and available for analysis.

Possible Limitations

There were two possible limitations of the approach used in the current research. First, because the video cameras were installed in the LVs and not the HVs, critical incidents could only be captured from the LV driver's perspective. It was possible that LV-HV interactions, which may have only been apparent from the HV driver's viewpoint, were not recorded. However, because there was fairly complete video recording coverage around the entire LV, it was likely that most LV-HV interactions that occurred were recorded.

Second, because there were no cameras mounted in any HV, it is difficult to have a complete understanding of the HV driver's behavior during the incident. The video camera that was directed at the LV driver's face, along with the verbal utterances of the driver, provided the researchers with a fairly complete understanding of the LV driver's behavior before, during, and after each incident. However, this was not the case regarding the behavior of the HV driver. The absence of video footage of the HV driver's face meant that the HV driver's behavior had to be surmised based on the video of the HV collected from the LVs and the comments and facial expressions made by the LV driver.

These limitations were also raised in Hanowski, Keisler, and Wierwille (2004) for the LV-HV analyses that were conducted with both L/SH and sleeper berth (SB) trucks. Considering the previous Hanowski, Keisler, and Wierwille (2004) work in conjunction with the current research, a more complete assessment of LV-HV interactions from both the HV and LV driver's perspective was expected. Assumptions regarding driver behavior were required for each of these research efforts, particularly for the driver of the non-instrumented vehicle.

Results

Light Vehicle-Heavy Vehicle Interaction Data Set

The 100-Car Study (Dingus et al., 2004) captured 9,125 incidents. These 9,125 incidents were divided into four categories: LV-LV Interactions, LV-HV Interactions; Single Vehicle Conflicts, and Other. Table 4 provides a description of the different vehicle types in each category.

Table 4. Vehicle Types Captured in the 100-Car Study.

The data set used in the current effort was comprised of a subset of incidents from the 9,125 incidents described above. The 9,125 incidents were reviewed and only those that involved a LV-HV interaction were included in the present analysis.

Figure 6 shows a pie chart of the 9,125 events as a function of the vehicles involved and whether or not the incident was an interaction between vehicles. As can be seen, of the 9,125 events, 246 (2.7%) involved a LV-HV interaction. In 2003, there were a total of 6,328,000 crashes in the U.S. (NHTSA, 2004). Of these crashes, 313,663 (5%) were classified as a LV-HV interaction. Thus, the present data set has fewer LV-HV interactions than the national crash statistics.

Of the 246 LV-HV recorded incidents, 219 (89%) were crash-relevant conflicts, 25 (10.2%) were near crashes, 1 (.4%) was a crash, and 1 (.4%) was undetermined. For the 79 incidents where the HV driver was judged to have been at-fault, 66 (83.5%) were crash-relevant conflicts and 13 (16.5%) were near crashes. For the 138 incidents where the LV driver was judged to have been at-fault, 128 (92.8%) were crash-relevant conflicts, 8 (5.8%) were near crashes, 1 (.7%) was a crash, and 1 (.7%) was undetermined. For the 29 incidents where it was unknown if the LV or HV driver was at-fault, 29 (100%) were crash-relevant conflicts.

Figure 6 . Distribution of the 9,125 Incidents Captured in 100-Car Study.

Incident Types

Given that the 100-Car data set was comprised of 246 LV-HV interactions, the next step in the analysis was to determine the vehicles' actions for each incident. To this end, the video and relevant data for each incident were carefully reviewed and then classified as an "Incident Type." Twenty-seven different Incident Types were identified (a detailed description of each is presented in Table 5). It should be noted that the 27 Incident Types listed do not necessarily comprise the entire universe of all types of LV-HV interaction incidents. Rather, the 27 Incident Types listed comprise those that were identified in this data set (N = 246). The Incident Types are written in such a way as to be interchangeable regarding LVs and HVs. Note that this is the same classification strategy outlined in the Hanowski, Keisler, and Wierwille (2004) study. However, in the Hanowski, Keisler, and Wierwille (2004) study, only 20 Incident Types were identified in their data set.

This was a unique event that was not identified from any of the triggering methods. This event involved the driver reporting that he had received a ticket for illegally passing a stopped school bus. This self-report was then confirmed by reviewing the video.

Table 5. Description of the Incident Types that were Identified in the Current Research.

Table 6 shows the frequency, percentage, and rank ordering for the Incident Types across the entire 100-Car data set. The rank ordering highlights the frequency of Incident Types from most frequently occurring (ranked as a low number, "1") to least frequently occurring (ranked as a high number, "23.5"). Incident Types that had an equal number of occurrences were ranked as a "tie" and the mean of the rankings was assigned. For example, "Approaches Traffic Quickly," and "Roadway Entrance Without Clearance" occurred equally with a frequency of "6." Because their order in the ranking would consist of the ninth and tenth positions, a mean ranking of "9.5" was assigned to both Incident Types.

As can be seen from the data presented in Table 6, the most frequent Incident Type involving a LV-HV interaction was Late Braking for Stopped/Stopping Traffic. Across all 246 incidents, this particular Incident Type occurred 66 times and accounted for 26.8% of the incidents captured. The bar graph shown in Figure 7 illustrates the frequency and percentage of each Incident Type across the entire data set. As can be seen from Figure 7, the majority of the incidents (48.8%) involved one of two different Incident Types: Late Braking for Stopped/Stopping Traffic, and Lane Change Without Sufficient Gap.

Table 6. Frequency, Percentage, and Rank Ordering of the Incident Types Across all LV-HV Incidents (N Total = 246).

Figure 7 . Frequency of Incident Types Across all LV-HV Incidents (N Total = 246).

Descriptive statistics for the Incident Types were also calculated for incidents as a function of the at-fault driver. The at-fault driver is the driver that was assessed, by the analyst, to have been responsible for causing the event. Of the 246 LV-HV interaction incidents recorded, 138 (56%) were judged to have been the fault of the LV driver, while 79 (32%) were attributed to the HV driver. For the remaining 29 incidents (12%), it was unclear which vehicle driver was at-fault. By removing the "unknown" cases from the LV-HV driver at-fault analyses, it was found that the LV driver was at-fault in 64% (138/217) of the LV-HV interaction incidents while the HV driver was at-fault in 36% (79/217) of the incidents.

Table 7 shows the frequency, percentage, and rank ordering for the Incident Types where the HV driver was judged to be at-fault. As can be seen, the most frequent Incident Type for HV driver at-fault incidents was Lane Change Without Sufficient Gap (26.6%), followed by Lateral Deviation of Through Vehicle (21.5%), and Left Turn Without Clearance (13.9%). Figure 8 shows a bar graph of the 79 HV driver at-fault incidents as a function of the Incident Type.

Table 7. Frequency, Percentage, and Rank Ordering of the Incident Types for HV Driver At-Fault Incidents (n HV = 79).

Figure 8 . Frequency of Incident Types for HV Driver At-Fault Incidents (n HV = 79).

Table 8 shows the frequency, percentage, and rank ordering for the Incident Types where the LV driver was at-fault. The most frequent Incident Type for LV driver at-fault incidents was Late Braking for Stopped/Stopping Traffic (41.3%) and Lane Change Without Sufficient Gap (21.7%). Figure 9 shows a bar graph of the 138 LV driver at-fault incidents as a function of the Incident Type.

Table 8. Frequency, Percentage, and Rank Ordering of the Incident Types for LV Driver At-Fault Incidents (n LV = 138).

Figure 9 . Frequency of Incident Types for LV Driver At-Fault Incidents (n LV = 138).

Table 9 shows the frequency, percentage, and rank ordering for the Incident Types when the at-fault driver was unknown. The most frequent Incident Type for Unknown at-fault incidents was Late Braking for Stopped/Stopping Traffic (27.6%), followed by Conflict With Oncoming Traffic (13.8%), Lane Change Without Sufficient Gap (10.3%), and Unable to Determine (10.3%). Figure 10 shows a bar graph of the 29 Unknown at-fault incidents as a function of the Incident Type.

Figure 10 illustrates the Incident Types, with respect to the driver assessed to be at-fault, by group (HV, LV, and Unknown). The figure shows that the Incident Types differed depending on whether the HV or LV driver was at-fault. Across all at-fault incidents, the most frequent Incident Type were Late Braking for Stopped/Stopping Traffic, Lane Change Without Sufficient Gap, and Lateral Deviation of Through Vehicle.

Table 9. Frequency, Percentage, and Rank Ordering of the Incident Types for Unknown At-Fault Incidents (n Un = 29).

Figure 10 . Frequency of Incident Types for Unknown At-Fault Incidents (n Un = 29).

Figure 11 . Frequency of Incident Types for HV, LV, and Unknown At-Fault Incidents (n HV = 79, n LV = 138, and n Un = 29).

Summary of Incident Type

Overall, the most common Incident Types were Late Braking for Stopped/Stopping Traffic (26.8%), Lane Change Without Sufficient Gap (22%), and Lateral Deviation of Through Vehicle (8.1%). These three Incident Types represented 56.9% of the LV-HV incidents.

A substantial number of LV-HV interactions were judged to have been the fault of the LV driver. Of the 246 LV-HV interaction incidents, 56.1% (63.6%, excluding the incident where it was unknown if the HV or LV driver was at-fault) of the LV drivers were at-fault, 32.1% (46.4% excluding the Unknown at-fault incidents) of the HV drivers were at-fault, while in the remaining 11.8%, it was unknown if the HV or LV driver was at-fault.

As can be seen in Figure 11, the Incident Types differed depending on whether the HV or LV driver was at-fault. The most prevalent Incident Types for HV driver at-fault incidents were Lane Change Without Sufficient Gap (26.6%), Lateral Deviation of Through Vehicle (21.5%), and Left Turn Without Clearance (13.9%). These three incident types accounted for 51.7% of the HV driver at-fault incidents. The most frequent Incident Types for LV drivers were Late Braking for Stopped/Stopping Traffic (41.3%) and Lane Change Without Sufficient Gap (21.7%). These two incident types accounted for 63.0% of the LV driver at-fault incidents. Note that the most prevalent Incident Type for at-fault LV drivers, Lane Change Without Sufficient Gap, was similar to the truck driver focus groups' reports in Hanowski et al. (1998) who indicated that "being cut-off by LV drivers" was a frequently occurring problem in local/short-haul trucking.

Primary Maneuvers, Secondary Maneuvers and Conflict Types

After each of the 246 incidents was classified by Incident Type, the next step in the analysis was to identify the "Primary Maneuvers" and "Secondary Maneuvers" involved in each incident. The Primary Maneuver refers to the maneuver of the driver who initiated the incident (not necessarily at-fault). Table 10 shows each Primary Maneuver and its corresponding definition. Across the 246 interaction incidents, 19 different Primary Maneuvers were identified.

Table 10. List and Definition of Each Primary Maneuver Types.

In addition to identifying the Primary Maneuver for each incident, the Secondary Maneuver, or the maneuver of the responding driver (i.e., driver of the second vehicle involved in the interaction), was also classified. Considering the maneuvers of both vehicles involved in the incident, a clear picture of the conflict, or Conflict Type, could be determined. Table 11 shows the Conflict Types that were identified in the 246 interaction incidents that were analyzed. As can be seen, Table 11 consists of 66 different Conflict Types (i.e., Primary Maneuver and Secondary Maneuver combinations).

Table 11 . The 66 Different Conflict Types Identified Across all LV-HV Incidents.

Table 12 shows the frequency, percentage, and rank ordering of each Primary Maneuver across the entire LV-HV 100-Car data set. The most frequent Primary Maneuver was Braking (22.8%), followed by Changing Lanes (21.1%) and Stopped (15%). Figure 12 shows a bar graph of the 246 incidents as a function of the Primary Maneuver.

Table 12. Frequency, Percentage, and Rank ordering of the Primary Maneuvers Across all LV-HV Incidents (N Total = 246).

Figure 12 . Frequency of Primary Maneuvers Across all LV-HV Incidents (N Total = 246).

Table 13 displays the frequency, percentage, and rank ordering for the Primary Maneuvers where the HV driver was at-fault. As can be seen in Table 13, the most frequent Primary Maneuver for HV driver at-fault incidents was Changing Lanes (32.9%), followed by Crossing Over Lane Line (20.3%) and Left Turn (15.2%). Figure 13 shows a bar graph of the 79 HV driver at-fault incidents as a function of the Primary Maneuver.

Table 13. Frequency, Percentage, and Rank Ordering of the Primary Maneuvers for the HV Driver At-Fault Incidents (n HV = 79).

Figure 13 . Frequency of Primary Maneuvers for HV Driver At-Fault Incidents (n HV = 79).

Table 14 displays the frequency, percentage, and rank ordering for the Primary Maneuvers where the LV driver was at-fault. As can be seen in Table 14, the most frequent Primary Maneuver for LV driver at-fault incidents was Braking (32.6%), followed by Stopped (21.7%) and Changing Lanes (16.7%). Figure 14 shows a bar graph of the 138 LV driver at-fault incidents as a function of the Primary Maneuver.

Table 14. Frequency, Percentage, and Rank Ordering of the Primary Maneuvers for the LV Driver At-Fault Incidents (n LV = 138).

Figure 14 . Frequency of Primary Maneuvers for LV Driver At-Fault Incidents (n LV = 138).

Table 15 displays the frequency, percentage, and rank ordering for the Primary Maneuvers where it was unknown if the HV or LV driver was at-fault. As can be seen in Table 15, the most frequent Primary Maneuver for Unknown at-fault incidents was Braking (27.6%), followed by Through Traffic (20.7%), Changing Lanes (10.3%), Left Turn (10.3%), and Stopped (10.3%). Figure 15 shows a bar graph of the 29 Unknown at-fault incidents as a function of the Primary Maneuver.

Figure 16 illustrates the Primary Maneuvers, with respect to the driver assessed to have been at-fault, by group ( LV, HV, and Unknown). Figure 16 illustrates that the Primary Maneuvers differ depending on whether the HV or LV driver was at-fault. Across all incidents, the most frequent Primary Maneuvers were Braking, Changing Lanes, and Stopped.

Table 15. Frequency, Percentage, and Rank Ordering of the Primary Maneuvers for the Unknown At-Fault Incidents (n Un = 29).

Figure 15 . Frequency of Primary Maneuvers for Unknown At-Fault Incidents (n Un = 29).

Figure 16 . Frequency of Primary Maneuvers for HV, LV, and Unknown At-Fault Incidents (n HV = 79, n LV = 138, and n Un = 29).

Summary of Primary Maneuvers

Overall, the most frequent Primary Maneuvers were Braking (22.7%), Changing Lanes (21.1%), and Stopped (15%). These three Primary Maneuvers represented 58.9% of the recorded incidents. The most frequent Primary Maneuvers, excluding the Unknown at-fault incidents, were Braking (25.8%), Changing Lanes (21.1%), and Stopped (15%). These three maneuver types represented 60.4% of the Primary Maneuvers for both HV and LV driver at-fault incidents.

As can be seen in Figure 16, the Primary Maneuvers differed depending on whether the HV or LV driver was at-fault. The most prevalent Primary Maneuvers for HV driver at-fault incidents were Changing Lanes (32.9%), Crossing Over Lane Line (20.3%), and Left Turn (15.2%). These three Primary Maneuvers accounted for 68.4% of the HV driver at-fault incidents. The most prevalent Primary Maneuvers for LV driver at-fault incidents were Braking (32.6%), Stopped (21.7%), and Changing Lanes (16.7%). These three Primary Maneuvers accounted for 71% of the LV driver at-fault incidents. From this data it appears that most Primary Maneuvers for HV driver at-fault incidents occurred when the HV was in-motion and changing lanes or crossing the lane line. In contrast, a significant proportion of the Primary Maneuvers when the LV driver was at-fault occurred when the LV was decelerating or stopped.

Contributing Factors

Just as the Incident Types describe the action or what happened during an incident, Contributing Factors provide likely reasons why an incident occurred. For each incident that was analyzed, a number of Contributing Factors were identified. It should be noted that the Contributing Factor categories were taken from Wierwille, Kieliszewski, Hanowski, Keisler, and Olsen (2000) and from the GES Physical Impairments screen (USDOT/NHTSA, 2003, p. 434). Each Contributing Factor and its corresponding definition are shown in Table 16. It should be noted that an incident could receive multiple Contributing Factors (e.g., both Driver Impairment and Willful Behavior). In addition, the Contributing Factor was based solely on the behaviors of the LV driver. Without cameras inside the HV there was no way to determine, with any degree of certainty, the behavior(s) of the HV driver.

Even when the HV driver was judged to have been at-fault, the behaviors of the LV driver are identified. Put another way, for the events that were caused by the LV driver, the analyses considered the LV driver's behaviors that may have contributed to the event. For events where the HV-driver was at-fault, the analysis also considered the LV driver's behaviors. However, the consideration is for the LV driver behaviors that occurred as the driver reacted to the HV driver's actions. For example, an HV driver at-fault incident that was assigned a Contributing Factors of "Distraction" would refer to the LV driver being distracted (not the HV driver) as he/she reacted to the event. Table 16 shows each of the contributing factors identified in this study. A complete list of contributing factors can be found in Dingus et al. (2004).

Table 16. Contributing Factors Used to Identify the Primary Cause of the Incident.

Table 17 shows the frequency, percentage, and rank ordering for each Contributing Factor across all 246 incidents. As can be seen in Table 17, the most frequent Contributing Factor was Driving Techniques (49.5%), followed by Unknown (24%), Distracted (18.7%), and Aggressive Driving (15%). Figure 17 shows a bar graph of the 246 incidents as a function of the Contributing Factor.

Table 17. Frequency, Percentage, and Rank Ordering of the Contributing Factors Across all LV-HV Incidents (N Total = 246).

Figure 17 . Frequency of Contributing Factors Across all LV-HV Incidents (N Total = 246. )

Since more than one Contributing Factor could be coded with each incident, the percentages total more than 100%. This procedure was also used by Blower (1998). The denominator for determining the percentages was the total number of LV-HV interactions for that data set (Example: The denominators in Tables 15 and 16 were 246 and 79, respectively).

Table 18 shows the frequency, percentage, and rank ordering of each Contributing Factor for the HV driver at-fault incidents. As can be seen in Table 18, the most frequent Contributing Factor was Unknown (68.4%), followed by Driving Techniques (15.2%), and Distracted (11.4%). Figure 18 shows a bar graph of the 79 HV driver at-fault incidents as a function of the Contributing Factor.

Table 18. Frequency, Percentage, and Rank Ordering of the Contributing Factors for HV Drivers in the HV At-Fault Incidents (n HV = 79).

Figure 18 . Frequency of Contributing Factors for LV Drivers in the HV At-Fault Incidents (n HV = 79).

Table 19 shows the frequency, percentage, and rank ordering of each Contributing Factor for the LV driver at-fault incidents. As can be seen in Table 19, the most frequent Contributing Factor for LV driver at-fault incidents was Driving Techniques (70.3%), followed by Distracted (22.5%) and Aggressive Driving (22.5%). Figure 19 shows a bar graph of the 138 LV driver at-fault incidents as a function of the Contributing Factor.

Table 19. Frequency, Percentage, and Rank Ordering of the Contributing Factors for the LV Driver At-Fault Incidents (n LV = 138).

Figure 19 . Frequency of Contributing Factors for LV Driver At-fault Incidents (n LV = 138).

Table 20 shows the frequency, percentage, and rank ordering of each Contributing Factor for the Unknown at-fault incidents. As can be seen in Table 20, the most frequent Contributing Factor for Unknown at-fault incidents was Driving Techniques (48.3%), followed by Distracted (20.7%), Unknown (17.2%), Roadway Alignment (10.3%), and Aggressive Driving (10.3%). Figure 20 shows a bar graph of the 29 Unknown at-fault incidents as a function of the Contributing Factor.

Table 20. Frequency, Percentage, and Rank Ordering of the Contributing Factors for the Unknown At-Fault Incidents (n Un = 29).

Figure 20 . Frequency of Contributing Factors for Unknown At-Fault Incidents (n Un = 29).

Contributing Factors X Incident Type Summary

Table 21 illustrates the frequency of LV driver at-fault incidents by Incident Type as well as Contributing Factor. Table 21 provides a more descriptive and comprehensive illustration of the Contributing Factors for each Incident Type. The far left column lists the Incident Types while the Contributing Factors are listed in the first row. As indicated above, more than one Contributing Factor could be coded to a particular incident, thus, 176 occurrences of Contributing Factors were coded to 138 LV driver at-fault incidents.

As can be seen in Table 21, the most frequent Contributing Factors for the Late Braking for Stopped/Stopping Traffic Incident Type was Driving Techniques (n = 47) and Distracted (n = 20). The Driving Techniques and Distracted Contributing Factors were coded in 82.5% and 35.1% of the LV at-fault Late Braking for Stopped/Stopping Traffic Incident Types, respectively. The most frequent Contributing Factors for the Lane Change Without Sufficient Gap Incident Types were Aggressive Driving (n = 16) and Driving Techniques (n = 14). The Aggressive Driving and Driving Techniques Contributing Factors were coded in 53.3% and 46.7% of the LV at-fault Lane Change Without Sufficient Gap Incident Types, respectively.

Table 21. Frequency of Incident Types X Contributing Factors Across LV Driver At-Fault Incidents (n LV = 138).

Tree Taxonomy

Figures 21-23 illustrate another way of viewing the data in Figures 18-20. Note that the numbers in the contributing factors boxes do not necessarily sum to the total in the higher at-fault category as multiple contributing factors could have been selected for each incident.

Figure 21 is a tree taxonomy of the General and Specific Contributing Factors for the 79 incidents where the HV driver was at-fault. An important point for the HV driver at-fault incidents is that the Contributing Factors are for the LV driver and not the HV driver. Again, this is due to the methodology used where cameras and other data collection equipment were in the LVs only. As such, the Contributing Factors in Figure 29 should be considered driver behaviors identified when the LV driver was responding to the at-fault HV driver.

Figure 22 provides a tree taxonomy of the 138 incidents where the LV was at-fault and provides a tree taxonomy of the 29 incidents where it was unknown if the HV or LV drivers was at-fault.

Figure 21. Taxonomy of the Contributing Factors for LV Drivers in the HV At-Fault Incidents (n HV = 79).

Figure 22 . Taxonomy of the Contributing Factors for LV Driver At-Fault Incidents (n HV = 138).

Figure 23 . Taxonomy of the Contributing Factors for Unknown At-Fault Incidents (n Un = 29).

Contributing Factor Taxonomy

A taxonomy to highlight the characteristics of the incidents was also developed. The Incident Types, Contributing Factor Categories, and brief descriptions of the incidents were arranged hierarchically to create three taxonomies: (1) Figure 24 illustrates the HV driver at-fault incidents, (2) Figure 25 illustrates the LV driver at-fault incidents, and (3) Figure 26 illustrates the Unknown at-fault incidents. At the highest level of the taxonomy, the events are grouped by the driver who was at-fault for the incident (i.e., HV, LV, or Unknown). The second highest level in the taxonomy is the Incident Type, followed by the General Contributing Factor, Specific Contributing Factor, and brief description of the incident, respectively. The incident descriptions explain the basic cause and result of the incidents. The taxonomy structures are shown below in Figures 24-26.

Figure 24 . Taxonomy Structure Used to Characterize the HV Driver At-Fault Incidents (n HV = 79).

Figure 24. (Continued.)

Figure 24. (Continued.)

Figure 24. (Continued.)

Figure 24. (Continued.)

Figure 24. (Continued.)

Figure 25. Taxonomy Structure Used to Characterize the LV Driver At-Fault Incidents (n LV = 138).

Figure 25. (Continued.)

Figure 25. (Continued.)

Figure 25. (Continued.)

Figure 25. (Continued.)

Figure 25. (Continued.)

Figure 25. (Continued.)

Figure 25. (Continued.)

Figure 26. Taxonomy Structure Used to Characterize the Unknown At-Fault Incidents (n Un = 29).

Figure 26. (Continued.)

Figure 26. (Continued.)

Contributing Factors Summary

Overall, the most frequent Contributing Factors were Driving Techniques (49.5%), Unknown (24%), Distracted (18.7%), and Aggressive Driving (15%). The most prevalent Contributing Factors for HV driver at-fault incidents were Unknown (68.4%), Driving Techniques (15.2%), and Distracted (11.4%). The most prevalent Contributing Factors for LV driver at-fault incidents were Driving Techniques (70.3%), Distracted (22.5%), and Aggressive Driving (22.5%). Please note that all Contributing Factors were coded with respect to LV driver at-fault incidents.

These findings are similar to what Hankey et al. (1999) found when they conducted a database analysis using the state of Pennsylvania's crash database. The Hankey et al. (1999) study found that 77% of crashes occurring from 1995 to 1996 had "human error" cited as a primary factor in the crashes. By summing the frequency of incidents that had at least one Contributing Factor associated with human error in Figures 18 and 19 (e.g., Driving Techniques, Aggressive Driving, and Purposeful Violation of Traffic Laws) for both HV and LV driver at-fault incidents yields a total of 145 incidents out of 217. Thus, 66.8% of all the LV-HV interactions were determined to have at least one human error as a Contributing Factor. Yet, when only the LV driver at-fault incidents are considered, 94.2% (130 incidents out of 138) of the incidents had at least one human error as a Contributing Factor. It should be noted that in the Hankey et al. (1999) study, human error was the primary factor (i.e., excluding all other factors) in the crash, while in the present study it was a contributing factor. Another difference is that the Hankey et al. (1999) study was a crash database analysis that included PARs, which may be biased as they rely, in part, on verbal reports from drivers. The current study had cameras inside the vehicle recording the behaviors of the driver as the incident occurred; this methodology eliminates driver verbal report bias.

Hankey et al. (1999) also conducted a database analysis using the 1996 FARS database. In this analysis, "Aggressive Driving" was found to be a primary factor in 31.1% of the fatal crashes. The current research found that Aggressive Driving contributed to 15% of all the LV-HV incidents. However, the Hankey et al. (1999) study assessed all crash types. While there are many similarities between the current data and the Hankey et al. (1999) study, the current research assessed LV-HV interactions rather than all crash types (e.g., LV-LV interactions were not considered).

Stuster (1999) assessed driver-related factors in LV-HV fatal crashes. He found that 67.3% of passenger vehicles were cited with a driver-related factor in fatal LV-HV crashes (these were similar to the Willful Behavior and Driver Proficiency categories in Contributing Factors). Stuster's (1999) results are similar to the results in the current research, where 66.8% of the LV-HV interactions were determined to have human error as a Contributing Factor.

Driver Distraction

A substantial number of the LV-HV incidents had Distraction listed as a Contributing Factor. Again, as indicated previously, the incidents where Driver Distraction was indicated refer to the behavior of the LV driver. The Distraction Contributing Factor was sub-divided into several discrete categories. Table 22 shows the frequency, percentage, and rank ordering for each sub-category in the Distraction Contributing Factor. As can be seen in Table 22, the most frequent sub-category for the Distraction Contributing Factor was Talking/Listening on Cell Phone (21.7%), followed by Passenger in Adjacent Seat (13%) and Dialing Hand-Held Phone (8.7%). Figure 27 displays a bar graph of the 46 Distraction Contributing Factors as a function of the discrete sub-categories.

Table 22. Frequency, Percentage, and Rank Ordering of Each Sub-Category in the Distraction Contributing Factor (n = 46).

Figure 27 . Frequency of the Distraction Contributing Factor in Each Sub-Category.

Table 23 shows the frequency, percentage, and rank ordering for each sub-category in the Distraction Contributing Factor for HV driver at-fault incidents. As can be seen in Table 23, the most frequent Distractions for HV at-fault incidents were Talking/Listening on Cell Phone (22.2%), Combing or Fixing Hair (22.2%), Dialing Hand-Held Cell Phone (22.2%), and Passenger in Adjacent Seat (22.2%).

Table 23. Frequency, Percentage, and Rank Ordering of Each Sub-Category in the Distraction Contributing Factor for the HV At-Fault Incidents (n = 9).

Table 24 shows the frequency, percentage, and rank ordering for each sub-category in the Distraction Contributing Factor for LV driver at-fault incidents. As can be seen in the table, the most frequent Distractions for LV at-fault incidents were Talking/Listening on Cell Phone (19.4%) and Passenger in Adjacent Seat (12.9%).

Table 24. Frequency, Percentage, and Rank Ordering of Each Sub-Category in the Distraction Contributing Factor for the LV At-Fault Incidents (n = 31).

Table 25 shows the frequency, percentage, and rank ordering for each sub-category in the Distraction Contributing Factor for the Unknown driver at-fault incidents. As can be seen in Table 25, the most frequent Distraction for Unknown at-fault incidents was Talking/Listening on Cell Phone (33.3%).

Table 25. Frequency, Percentage, and Rank Ordering of Each Sub-Category in the Distraction Contributing Factor for the Unknown At-Fault Incidents (n = 6).

Summary of Driver Distraction

Overall, Distraction was cited in 18.7% of the LV-HV interaction incidents (see Figure 17). This is higher than what Stutts et al. (2003) found when they reviewed the U.S. Crashworthiness Data System (CDS) from 1995-1999. Stutts et al. (2003) found that 8.3% of drivers involved in a fatal LV-HV interaction crash were identified as "Distracted" in the CDS database. Similarly, Stuster (1999) found that 8.7% of the LV drivers were cited with the driver-related factor "Driving Inattentively." It should be noted that Stuster (1999) only listed the twelve most frequent driver-related factors. Thus, the percentage noted may be incomplete as it only refers to driving inattentively. Hanowski, Olson, Perez, and Dingus (2001) assessed the occurrence of "driver distraction" as a contributing factor in a naturalistic study using long-haul drivers. A total of 2,737 critical incidents were recorded in the Sleeper Berth study. Of these, 178 (6.5%) had "driver distraction" as a contributing factor. As such, there was a considerably higher percentage of "driver distraction" related incidents in the current study as compared to previous studies.

As seen in Figure 17, the Distraction Contributing Factor was divided into discrete sub-categories. When Stutts et al. (2003) assessed the CDS, they also divided the 8.3% of the "Distraction" crashes into more discrete sub-categories. They found that the most frequent distracter in the CDS involved an External Distraction (29.4%), followed by Adjusting Radio/Cassette/CD (11.4%) and Other Occupant (10.9%), respectively. The current research found very different results. For example, as shown in Table 24, External Distractions (summing Left Window, Right Window, Center Mirror, and Other External Distraction) accounted for only 19.5% of the distraction incidents for at-fault LV drivers. Furthermore, Adjusting the Radio and Passenger in Adjacent Seat only accounted for 6.5% and 12.9% of the distraction incidents for LV at-fault drivers, respectively.

The most frequent sub-category for the Distraction Contributing Factor for LV at-fault incidents in current research was Talking/Listening on Cell Phone (19.4%), followed by Passenger in Adjacent Seat (12.9%) and Looking Out Center Mirror (9.7%), and Looking Out Left Window (9.7%). Stutts et al. (2003) found that Using/Dialing Phone was cited in only 1.5% of the distraction crashes, while Other Occupant (similar to Passenger in Adjacent Seat in the current research) was cited in 10.9% of the distraction crashes. Note that Stutts et al. (2003) include both talking/listening and dialing phones in their definition of phone use.

Again, these differences might highlight the discrepancy between crashes and near crashes and/or the methodologies used to obtain the data. Another likely reason for the discrepancy is that the Stutts et al. study used crash data from 1995-1999. There is little doubt that the number of cell phones in use has increased substantially from the time period of 1995-1999, as compared to 2003-2004, when the current data were collected. In addition, in many states, PARs have not been developed to account for cell phone use (Wierwille et al., 2002) and this may have also contributed to the lower percentage found in the Stutts et al. (2003) research.

Classification Strategy Used in the LTCCS

Accident Types

Each of the 246 LV-HV interactions were grouped by Accident Type based on the methodology used in the LTCCS (Thieriez, Radja, and Toth, 2002). Note that there was only one LV-HV crash recorded in the 100-Car Study. Therefore, using the Accident Types from the LTCCS does not reflect an absolute match, but rather a relative match. However, to facilitate future data comparisons with the near-crash data collected in the current study with other studies using the LTCCS, each of the 246 LV-HV interactions were coded using the LTCCS classification scheme. Because only one crash occurred, the closest match with respect to Accident Types was recorded for each incident. Table 26 shows the LTCCS Accident Type descriptions.

Table 26. Description of the LTCCS Accident Types (Thieriez, Radja, and Toth, 2002).

This table lists categories, configurations, and accident types (including intent) that were identified in the LTCCS.

Category I. Single Driver; there are three configurations.

Configuration A: Right Roadside Departure; there are five accident types:

1. Drive Off Road
2. Control/Traction Loss
3. Avoid Collision with Vehicle, Pedestrian, Animal
4. Specifics Other
5. Specifics Unknown

Configuration B: Left Roadside Departure; there are five accident types:

1. Drive Off Road
2. Control/Traction Loss
3. Avoid Collision with Vehicle, Pedestrian, Animal
4. Specifics Other
5. Specifics Unknown

Configuration C: Forward Impact; there are six accident types:

1. Parked Vehicle
2. Stationary Object
3. Pedestrian/Animal
4. End Departure
5. Specifics Other
6. Specifics Unknown

Category II. Same Trafficway, Same Direction; there are three configurations.

Configuration D: Rear-End; there are 14 accident types:

20 - 23. Stopped

24 - 27. Slower

28 - 31. Decelerating

32. Specifics Other

33. Specifics Unknown

Configuration E: Forward Impact; there are 10 accident types:

34 - 35. Control/Traction Loss

36 - 37. Control/Traction Loss

38 - 39. Avoid Collision With Vehicle

40 - 41. Avoid Collision With Object

42. Specifics Other

43. Specifics Unknown

Configuration F: Sideswipe Angle; there are six accident types, including:

48. Specifics Other

49. Specifics Unknown

Category III. Same Trafficway, Opposite Direction; there are three configurations.

Configuration G: Head-On; there are four accident types:

50 - 51. Lateral Move

52. Specifics Other

53. Specifics Unknown

Configuration H: Forward Impact; there are 10 accident types:

54 - 55. Control/Traction Loss

56 - 57. Control/Traction Loss

58 - 59. Avoid Collision With Vehicle

60 - 61. Avoid Collision With Object

62. Specifics Other

63. Specifics Unknown

Configuration I: Sideswipe Angle; there are six accident types:

64 - 65. Lateral Move

66. Specifics Other

67. Specifics Unknown

Category IV. Change Trafficway, Vehicle Turning; there are two configurations.

Configuration J: Turn Across Path; there are eight accident types:

68 - 69. Initial Opposite Directions

70 - 73. Initial Same Direction

74. Specifics Other

75. Specifics Unknown

Configuration K: Turn Into Path; there are 10 accident types:

76 - 79. Turn Into Same Direction

80 - 83. Turn Into Opposite Directions

84. Specifics Other

85. Specifics Unknown

Category V. Intersecting Paths (Vehicle Damage); there is one configuration.

Configuration L: Straight Paths; there are six accident types, including:

90. Specifics Other

91. Specifics Unknown

Category VI. Miscellaneous; there is one configuration.

Configuration M: Backing, Etc.; there are five accident types:

92. Backing Vehicle

93. Other Vehicle or Object

94. Other Accident Type

95. Unknown Accident Type

100. No Impact

Table 27 shows the frequency, percentage, and rank ordering for the Accident Types across the entire data set. As can be seen, the most commonly occurring Accident Type that involved an interaction between a LV and HV was Scenario 38/39: Same Trafficway Same Direction, Forward Impact, Avoid Collision With Vehicle. Again, it is important to note that the Accident Types listed above in Table 26 were intended to be used for crashes and not for near crashes. Since this was the case, a "what if" approach was taken where data analysts coded the Accident Type to reflect what would likely have occurred had there been a crash. Thus, because of this subjective interpretation, not all events fit neatly into a category. As such, the results from the Accident Type categorization are not as "clean" as those from the Incident Type categories used in Hanowski, Keisler, and Wierwille (2004). Figure 28 shows a bar graph of the 246 incidents as a function of Accident Type.

Table 27. Frequency, Percentage, and Rank Ordering of the LTCCS Accident Types Across the Entire Data Set (N Total = 246).

Figure 28 . Frequency of Accident Types Across the Entire Data Set (N Total = 246).

Table 28 illustrates the frequency, percentage, and rank ordering of the Accident Types for HV driver at-fault incidents. The most frequent Accident Type for the HV driver at-fault incidents was Scenario 44/45: Same Trafficway/Same Direction: Sideswipe Angle: In Blind Spot (27.8%), followed by Scenarios 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (15.2%) and 25/25: Same Trafficway/Same Direction: Rear-End: Approaches Slower Constant Speed Vehicle (8.9%). Figure 29 shows a bar graph of the 79 HV driver at-fault incidents as a function of the Accident Type.

Table 28. Frequency, Percentage, and Rank Ordering of the Accident Types for HV Driver At-Fault Incidents (n HV = 79).

Figure 29 . Frequency of Accident Types for HV Driver At-Fault Incidents (n HV = 79).

Table 29 illustrates the frequency, percentage, and rank ordering of the Accident Types for LV driver at-fault incidents. The most frequent Accident Types for HV driver at-fault incidents was Scenario 20/21: Same Trafficway/Same Direction: Rear-End: Approaches Stopped Vehicle (26.8%), followed by Scenarios 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (22.5%) and 28/29: Same Trafficway/Same Direction: Rear-End: Approaches Decelerating Vehicle (17.4%). Figure 30 shows a bar graph of the 138 LV driver at-fault incidents as a function of the Accident Type.

Table 29. Frequency, Percentage, and Rank Ordering of the Accident Types for LV Driver At-Fault Incidents (n LV = 138).

Figure 30 . Frequency of Accident Types for LV driver At-Fault Incidents (n LV = 138).

Table 30 shows the frequency, percentage, and rank ordering of the Accident Types for Unknown driver at-fault incidents. The most frequent Accident Type for the Unknown Accident Types was Scenario 38/39: Same Trafficway Same Direction; Forward Impact, Avoid Collision With Vehicle (20.7%); followed by Scenarios 28/29: Same Trafficway/Same Direction: Rear-End: Approaches Decelerating Vehicle (17.2%) and 52/52: Same Trafficway/Opposite Direction: Head-On: Other (10.3%). Figure 31 shows a bar graph of the 29 Unknown at-fault incidents as a function of the Accident Type.

Table 30. Frequency, Percentage, and Rank Ordering of the Accident Types Where It was Unknown if the LV or the HV driver was At-Fault (n Un = 29).

Figure 31 . Frequency of Accident Types for Unknown At-Fault Incidents (n Un = 29).

Figure 32 shows a bar graph comparing the Accident Types, with respect to the driver that was assessed to have been at-fault, for the three groups ( LV, HV, and Unknown). The figure shows that the Accident Types varied depending on whether the HV or LV driver was at fault. The most commonly occurring Accident Types were Scenarios 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle; 20/21: Same Trafficway/Same Direction: Rear-End: Approaches Stopped Vehicle; 28/29: Same Trafficway/Same Direction: Rear-End: Approaches Decelerating Vehicle.

Figure 32 . Frequency of Accident Types for HV, LV, and Unknown At-Fault Incidents (n HV = 79, n LV = 130, & n Un = 29).

Accident Types Summary

Overall, the most frequent Accident Types were Scenarios 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (19.9%); 20/21: Same Trafficway/Same Direction: Rear-End: Approaches Stopped Vehicle (16.3%); and 28/29: Same Trafficway/Same Direction: Rear-End: Approaches Decelerating Vehicle (13%). These three Accident Types represented 49.2% of the Accident Types for all LV-HV incidents.

As can be seen in Figure 32, the Accident Types for HV and LV driver at-fault incidents differed markedly. The most prevalent Accident Types for HV driver at-fault incidents were Scenarios 44/45: Same Trafficway/Same Direction: Sideswipe Angle: In Blind Spot (27.8%); 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (15.2%); and 25/25: Same Trafficway/Same Direction: Rear-End: Approaches Slower Constant Speed Vehicle (8.9%). These three Accident Types accounted for 51.9% of the HV driver at-fault incidents. The most prevalent Accident Types for LV driver at-fault incidents were Scenarios 20/21: Same Trafficway/Same Direction: Rear-End: Approaches Stopped Vehicle (26.8%); 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (22.5%); and 28/29: Same Trafficway/Same Direction: Rear-End: Approaches Decelerating Vehicle (17.4%). These three Accident Types accounted for 66.7% at the LV driver at-fault incidents.

Figure 32 highlights some of the differences between HV and LV driver at-fault incidents with respect to Accident Type. The most prevalent Accident Type for HV driver at-fault incidents involved a Sideswipe Angle. By summing all the HV driver at-fault Accident Types that involved a Sideswipe Angle, it was found that 41.8% of the HV driver at-fault incidents were coded with this Accident Type. Conversely, the most prevalent Accident Type for LV driver at-fault incidents involved a Rear-End Approach. By summing all LV driver at-fault Accident Types that involved a Rear-End approach, it was found that 55.1% of the LV driver at-fault incidents were coded with this Accident Type. Thus, most of the HV driver at-fault incidents involved a Sideswipe Angle, while most of the LV driver at-fault incident involved a Rear-End Approach.

These results are quite different than Blower's (1998) review of fatal LV-HV crashes. He found that 13.9% of fatal LV-HV interactions, where only the LV was coded with a driver-related factor, involved a rear-end approach. Further, Blower found that 9.4% of the fatal LV-HV interactions, where only the HV driver was coded with a driver-related factor, involved a sideswipe angle. Similarly, when Council et al. (2003) reviewed all types of LV-HV crashes in North Carolina, they found that 23.2% of the HV driver at-fault crashes involved a sideswipe and 28.5% of the LV driver at-fault crashes involved a rear-end approach. One possible reason for these discrepancies is due to the difference in the event types being classified; that is, non-crashes in the current study and fatal LV-HV crashes in Blower (1998) and all types of crashes in Council et al. (2003). Also, the driver/subject population that was included in the study was skewed towards high mileage and younger drivers, and this may also impacted the types of events that were recorded in this study.

Critical Reason for the Critical Event

To be consistent with the LTCCS (Thieriez, Radja, and Toth, 2002), the LV driver at-fault incidents were coded with a Critical Reason for the incident. The Critical Reason for the incident was considered the primary reason for why the incident occurred (as compared to the Contributing Factors where all the factors that contributed to the incident's occurrence were coded). More than one Critical Reason could be coded for each incident, but this was a rare occurrence (ten of the recorded incidents were coded with two Critical Reasons). Only the LV driver at-fault incidents were coded with a Critical Reason because, as noted previously, only those vehicles were equipped with video recording equipment. For the HV driver at-fault incidents, it was not possible to determine with any certainty what the driver was doing that contributed to the event; therefore, all HV driver at-fault incidents were coded as "Unknown reason for the critical event."

Table 31 shows the frequency, percentage, and rank ordering of the Critical Reasons for LV driver at-fault incidents. The most frequent Critical Reason for the LV driver at-fault incidents was Aggressive Driving Behavior (24.6%), followed by Too Fast for Conditions (15.2%) and Internal Distraction (13.8%). As in the previous analysis to determine the Contributing Factors (Chapter 1), more than one Critical Reason could be coded to a particular incident, thus, the percentage total is greater than 100%. Figure 33 shows a bar graph of the 138 LV driver at-fault crashes as a function of the Critical Reason.

Table 31. Frequency, Percentage, and Rank Ordering of the Critical Reasons for LV Driver At-Fault Incidents (n LV = 138).

Figure 33 . Frequency of Critical Reasons for LV Driver At-Fault Incidents (n LV = 138)

Critical Reasons X Incident Type Summary

Table 32 illustrates the frequency of LV driver at-fault incidents by Incident Type as well as Critical Reason. Table 32 provides a more descriptive and comprehensive illustration of the Critical Reasons for each Incident Type. The Incident Types were chosen, rather than the Accident Types, because, based on the results shown previously, they seen to be more appropriate for classifying near-crashes and incidents. The far left column of Table 32 lists the Incident Types while the Critical Reasons are listed in the first row. As indicated above, more than one Critical Reason could be coded to each incident, thus, there are 147 Critical Reasons coded to 138 LV driver at-fault incidents.

As can be seen, the most frequent Critical Reasons for the Late Braking for Stopped/Stopping Traffic Incident Type was Internal Distraction (n =11) and Too Fast for Conditions (n = 8). The Internal Distraction and Too Fast For Conditions Critical Reasons were coded in 19.3% and 14% LV at-fault Late Braking for Stopped/Stopping Traffic Incident Types, respectively. The most frequent Critical Reasons for the Lane Change Without Sufficient Gap Incident Types were Aggressive Driving (n = 16) and Too Fast for Conditions (n = 6). The Aggressive Driving and Lane Change Without Sufficient Gap Critical Reasons were coded in 53.3% and 20% of the LV at-fault Lane Change Without Sufficient Gap Incident Types, respectively.

Table 32. Frequency of Incident Types X Critical Reasons for LV Driver At-Fault Incidents (n LV = 138).

Critical Reason Summary

The most frequent Critical Reasons for LV driver at-fault incidents were Aggressive Driving Behavior (24.6%), Too Fast for Conditions (15.2%), and Internal Distraction (13.8%). These were markedly different than what Blower (1998) found when he assessed fatal LV-HV interactions. For incidents where only the LV driver was cited with a driver-related factor, Blower (1998) found that only 7.1% of the incidents were coded with the Aggressive Driving Critical Reason, while 19.2% and 0.11% of the incidents were coded with Too Fast for Conditions and Internal Distraction, respectively.

There were other interesting trends in the current data set. Sixty-four of the 138 LV at-fault incidents (46.4%) were coded with at least one Critical Reason that was a risky driving behavior (i.e., Aggressive Driving Behavior, Too Fast for Conditions, Following too Closely, and Illegal Maneuver), while 22.5% of the LV driver at-fault incidents involved some type of awareness variable (i.e., Internal Distraction, Inattention, External Distraction). In comparison, Stuster's (1999) analysis of fatal LV-HV crashes found that 57.2% of the incidents involved the at-risk driving behaviors by LV drivers. Stuster (1999) also found that 10.1% of fatal LV-HV crashes involved the LV driver driving inattentively.

Conclusions

The analyses that were conducted with the LV-HV interactions captured in the 100-Car Study (Dingus et al., 2004) provide convincing evidence to support the contention that LV-HV interactions are a serious problem. Consider that of the 9,125 critical incidents captured in the 100-Car Study, 246 LV-HV interactions were identified. Put another way, of the large critical incident data set that was obtained in the 100-Car Study, 2.7% of these incidents involved a LV-HV interaction. While 2.7% may appear to represent a small proportion of the overall critical incident picture, it should be noted that LV-HV interactions easily have the potential to become serious, and even fatal because of the tremendous difference in weight between an HV and LV.

There are seven important findings that stem from the analyses conducted on the interactions between HVs and LVs. First, of the 246 interactions that were analyzed, 138 (56.1%) were assessed to have been the fault of the LV driver. HV drivers were at-fault in 79 (32.1%) of the incidents, while in the remaining 29 (11.8%) incidents it was unknown if the HV or LV driver was at-fault. Excluding the incidents where it was unknown if the HV or LV driver was at-fault, 63.6% and 36.4% of the incidents were the fault of the LV and HV drivers, respectively. Thus, LV drivers were responsible for a significant proportion of the LV-HV interactions.

These findings support what the drivers in the Hanowski et al. (1998) focus groups reported about LVs being their most important safety concern. Further, the results are similar to prior published studies that used a crash database approach to assess LV-HV interactions (cf. Blower, 1998; Stuster, 1999; Wang, Knipling, and Blincoe, 1999). Based on these findings, it is suggested that focusing on the LV driver, and their errors, may provide the largest area of opportunity for reducing such events.

The second important finding from these analyses was in regard to the different Incident Types that were frequent among HV and LV drivers. For LV driver at-fault incidents, the most frequent Incident Types were: Late Braking for Stopped/Stopping Traffic (41.3%), Lane Change Without Sufficient Gap (21.7%), and Aborted Lane Change (8%). These particular Incident Types are indicative of at-risk driving behaviors. Once again, the objective data support the sentiment of the L/SH drivers in the Hanowski et al. (1998) focus group who indicated that during their daily travel they were often "cut-off" by LV drivers. And, the data supports the results from the L/SH on-road study (Hanowski, Keisler, and Wierwille, 2004) where the most prevalent Incident Type for LV driver at-fault incident was Lane Change Without Sufficient Gap (accounting for 24.8% of the LV driver at-fault incidents in the L/SH study). In contrast, the most frequent Incident Types for HV drivers were: Lane Change Without Sufficient Gap (26.6%), Lateral Deviation of Through Vehicle (21.5%), and Left Turn Without Clearance (13.9%). There were substantial differences in the most prevalent Incident Types as a function of driver group.

The third finding is the difference in the Primary Maneuvers for HV and LV drivers. The most prevalent Primary Maneuvers for LV driver at-fault incidents were: Braking (32.6%), Stopped (21.7%), and Changing Lanes (16.7%). The two most frequent Primary Maneuvers for LV driver at-fault incidents involved assumed difficulties on the part of the LV driver decelerating or stopping. In contrast, the most frequent Primary Maneuvers for HV driver at-fault incidents were: Changing Lanes (32.9%), Crosses Over Lane Line (20.3%), and Left Turn (15.2%). The two most frequent Primary Maneuvers for HV driver at-fault incidents involved difficulties changing or crossing over the lane line while the vehicle was in motion. These results make intuitive sense because HV drivers have limited visibility and deal with blind spots thereby making lane changes difficult in traffic.

It appears that the most frequent Primary Maneuvers for the HV drivers were fundamentally different than those of LV drivers. Whereas the HV drivers seemed to have difficulty changing lanes (which, it is assumed, reflects the inherent difficulty of driving with significant blind spots), LV drivers exhibited difficulties decelerating or stopping (which supports the findings that a significant proportion of the incidents involved a lack of attention and/or at-risk driving behaviors). Also, it is important to point out that there were few similarities in the Primary Maneuvers for LV and HV drivers. This highlights the fundamental differences in the characteristics of at-fault incidents by HV and LV drivers.

The fourth important finding is related to the Contributing Factors that were most frequent with HV and LV drivers. For LV drivers, the most frequent Contributing Factors for at-fault incidents were: Driving Techniques (70.3%), Distracted (22.5%), and Aggressive Driving (22.5%). The most frequent Contributing Factors for HV driver at-fault incidents were: Unknown (68.4%), Driving Techniques (15.2%), and Distracted (11.4%). The large number of Unknown Contributing Factors for HV driver at-fault incidents is indicative of the methodology used to code these events. Because the HV did not have any video cameras, the Contributing Factor was coded with respect to the behaviors of the LV driver. As the LV driver was not responsible for the incident, it was unlikely they would be coded with a Contributing Factor, thus the high frequency of Unknown Contributing Factors. Further, the methodology used to code the Contributing Factors also explains the similarities between LVs and HVs (i.e., they were all coded with respect to the LV driver, and therefore, might be expected to be similar).

The fifth noteworthy finding from the current research involves the Accident Types that were most prevalent for LV and HV drivers. The most prevalent Accident Types for LV driver at-fault incidents were: Scenarios 20/21: Same Trafficway/Same Direction: Rear End: Approaches Stopped Vehicle (26.8%); 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (22.5%); and 28/29: Same Trafficway/Same Direction: Rear End: Approaches Decelerating Vehicle (17.4%). Approximately 55% of the LV driver at-fault incidents involved a Rear-End approach. These Accident Types also support the findings from the analysis of the most prevalent Primary Maneuvers for LV driver at-fault incidents: decelerating or stopped.

Conversely, the most prevalent Accident Types for HV driver at-fault incidents were: Scenarios 44/45: Same Trafficway/Same Direction: Forward Impact: Sideswipe Angle: In Blind Spot (27.7%); 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (15.2%); and 25/25: Same Trafficway/Same Direction: Rear End: Approaches Constant Speed Vehicle (8.9%). Approximately 42% of the HV driver at-fault incidents involved a Sideswipe Angle. These Accident Types also support the findings from the most prevalent Primary Maneuvers for HV driver at-fault incidents: changing lanes and crossing the lane line.

The seventh noteworthy finding from the current research reflects some of the similarities and discrepancies found between the current study and prior studies using a crash database approach in analyzing LV-HV interactions. While both approaches found that LV drivers were responsible for the majority of LV-HV interactions, the reasons why these interactions occurred differed with respect to the methodologies used to assess these interactions. For example, the current research found that 22.5% of the LV driver at-fault incidents were cited with the Contributing Factors of Aggressive Driving. In Stuster's (1999) analysis, only 4.3% of the LVs were cited with the driver-related factor "Erratic/Reckless Driving." Moreover, Hankey et al. (1999) found that 31.1% of the fatal crashes in the FARS database were cited with Aggressive Driving. As such, the results from the current study (22.5%) are within the range reported by Stuster (4.3%) and Hankey (31.1%).

Table 33 compares the Critical Reasons for the LV at-fault incidents in the current study to the LV driver-related factors in the Blower (1998), Stuster (1999), and Council et al. (2003) studies. Although Table 33 provides a convenient way to compare the prior LV-HV interaction studies to the current study, the reader should be aware that there were significant differences in the data and classification strategy used in the different studies. For example, both Blower (1998) and Stuster (1999) assessed only fatal LV-HV crashes, and Council et al. (2003) assessed all types of LV-HV crashes (including fatal crashes). More than one driver-related factor could have been selected in the Blower (1998), Stuster (1999), and Council et al (2003) studies. Lastly, the Stuster (1999) study only reported the twelve most important (or frequent) driver-related factors.

Table 33. Comparison of a Selection of Results in the Current Study with Prior LV-HV Interaction Studies.

The current research also found that 41.8% of the HV driver at-fault incidents involved a Sideswipe Angle, while 55.1% of the LV driver at-fault incidents involved a Rear End approach. These results differed from Blower's (1998) review of fatal LV-HV crashes. He found that 9.4% of fatal LV-HV interactions involved a sideswipe angle. Further, Blower's (1998) analysis found that 13.9% of the fatal LV-HV interactions involved a rear-end strike. When Council et al. (2003) reviewed all types of LV-HV crashes in North Carolina, they found that 23.2% of the HV driver at-fault crashes involved a sideswipe and 28.5% of the LV driver at-fault crashes involved a rear-end approach. These discrepancies might highlight the differences between analyzing crashes and near crashes and/or the methodologies used analyze the data (i.e., a crash database approach versus a naturalistic or in situ data collection approach, and the younger age bias in the current study).

Foot Notes

This was a unique event that was not identified from any of the triggering methods. This event involved the driver reporting that he had received a ticket for illegally passing a stopped school bus. This self-report was then confirmed by reviewing the video.

Since more than one Contributing Factor could be coded with each incident, the percentages total more than 100%. This procedure was also used by Blower (1998). The denominator for determining the percentages was the total number of LV-HV interactions for that data set (Example: The denominators in Tables 15 and 16 were 246 and 79, respectively).

Passenger vehicles were only considered because the Contributing Factors in the current research were based solely on the behavior(s) of the LV driver

Only LV at-fault incidents were considered because only the LV drivers had instrumented vehicles.

It should be noted that Stuster (1999) only listed the twelve most frequent driver-related factors.

Only includes HV driver initiated events recorded in the Sleeper Berth study (Hanowski, Olson, Perez, and Dingus, 2001).

Only includes driving inattentively.

Called "Unsafe Speed" rather than Too Fast for Conditions in Stuster (1999) and Council et al. (2003).

The first number indicates the percent of all LV-HV crashes, while the second number indicates the percent of fatal LV-HV crashes

Called "Improper Following" rather than Following Too Closely in Blower (1998).

Called "Erratic/Reckless Driving" in both Blower (1998) and Stuster (1999).