The Effects of Rurality, Proximity of Other Traffic,
and Roadway Curvature on High Beam Headlamp
Use Rates
March 2016
Ian J. Reagan
Insurance Institute for Highway Safety
Matthew L. Brumbelow
Insurance Institute for Highway Safety
Michael J. Flannagan
University of Michigan Transportation Research Institute
ABSTRACT
Objective: The few observational studies of the prevalence of high beam use indicate the rate of
high beam use is about 25% when vehicles are isolated from other vehicles on unlit roads. Recent studies
were limited to two-lane rural roads and used measurement methods that likely overestimated use. The
current study examined factors associated with the rate of high beam use of isolated vehicles on a variety
of roadways in the Ann Arbor, Michigan, area. Methods: Twenty observation sites were categorized as
urban, rural, or on a rural/urban boundary and selected to estimate the effects of street lighting, road
curvature, and direction of travel relative to the city on high beam use. Sites were selected in pairs so that
a majority of traffic passing one site also passed through the other. Measurement of high beams relied on
video data recorded for two nights at each site, and the video also were used to derive a precise measure
of the proximity of other traffic. Nearly 3,200 isolated vehicles (10 s or longer from other vehicles) were
observed, representing 1,500-plus vehicle pairs. Results: Across the sample, 18% of the vehicles used
high beams. Seventy-three percent of the 1500-plus vehicle pairs used low beams at each paired site,
whereas 9% used high beams at both sites. Vehicles at rural sites and sites at the boundaries of Ann Arbor
were more likely to use high beams than vehicles at urban sites, but use in rural areas compared with
rural/urban boundary areas did not vary significantly. Rates at all sites were much lower than expected,
ranging from 0.9% to 52.9%. High beam use generally increased with greater time between subject
vehicles and leading vehicles and vehicles in the opposing lane. There were mixed findings associated
with street lighting, road curvature, and direction of travel relative to the city. Conclusion: Maximizing
visibility available to drivers from headlights includes addressing the substantial underuse of high beam
headlamps. Advanced technologies such as high beam assist, which switches automatically between high
and low beam headlamps depending on the presence of other traffic, can help to address this problem.
Key words: Nighttime driving, visibility, high beam headlamps, high beam assist, adaptive
driving beam.
1
INTRODUCTION
Twenty-nine percent of all fatalities during 2014 occurred in the dark on unlit roads (Insurance
Institute for Highway Safety, 2016). Although factors such as alcohol impairment and fatigue contributed
to many of these crashes, poor visibility likely also played a role. Since 1969, Federal Motor Vehicle
Safety Standard 108 has specified that all vehicles sold in the United States must have separate switches
for manually selecting low beam and high beam headlamps, with low beams required to limit glare for
oncoming or leading vehicles and high beams required to maximize forward illuminance in the absence of
other traffic. Drivers who consistently use high beam headlamps when appropriate will be able to see
farther ahead and will be better able to identify hazards and take appropriate action to avoid them (Reagan
and Brumbelow, 2016). This includes, in particular, pedestrians, bicyclists, and dimly lit roadside hazards
such as wildlife and disabled vehicles.
Substantial underuse of high beam headlamps was first documented in the 1960s (Hare and
Hemion, 1968), with observed use at 24% when averaged across 3 suburban and 14 rural sites throughout
the United States. The range of high beam use across the 17 sites was large, ranging from 3% to 85%, and
high beam use declined as traffic volume increased. The authors concluded that use varied significantly
by geographic region, but it is questionable whether this conclusion was warranted given that the 17 sites
were scattered across eight regions and three topographies. More recent, less comprehensive
observational studies (Iragavarapu and Fitzpatrick, 2012; Sullivan et al., 2004) have found that high beam
use remains low, with estimated use rates ranging from 25% to about 50%.
A limitation of Iragavarapu and Fitzpatrick (2012) and Sullivan et al. (2004) is that beam use was
assessed only on two-lane rural roads, where use is likely highest. Many roads located in urban or
suburban settings have segments where high beam use would be prudent, but drivers have to switch
between high and low beam headlamps due to the intermittent presence of other traffic or varying levels
of street lighting. While it is reasonable to expect high beam use to be lower in urban or suburban settings
than in rural areas, data about actual use in these locations are limited. Mefford et al. (2006) examined
high beam use among drivers involved in a field operational test of vehicles outfitted with data recorders
2
and prototype lane departure warning and curve speed warning systems. They reported that the
percentage of distance driven with high beams was highest when drivers were on rural roads and isolated
from other traffic, but the authors did not report the specific rates for urban and suburban conditions.
A second limitation of the more recent observation studies (Iragavarapu and Fitzpatrick, 2012;
Sullivan et al., 2004) is measurement error associated with the methodologies for classifying headlamp
patterns into high beam and low beam. Both studies used a single illuminance meter and an illuminance
threshold to classify beam patterns. A weakness of this approach is that low beams that are misaimed
upward may produce illuminance values that could be misclassified as high beams, thereby resulting in
overestimates of high beam use. As detailed below, the current study used a method that reduced the
imprecision associated with a single illuminance criterion to provide a more accurate estimate of high
beam use.
Advanced technologies can optimize visibility while limiting glare, largely eliminating the need
for drivers to switch between high and low beam headlamps. High beam assist systems use camera-based
sensors to switch automatically between high and low beam headlamps depending on the presence of
other traffic. Adaptive driving beam systems block out only the portion of the high beam pattern that
would otherwise glare the oncoming or leading drivers so as to provide the enhanced visibility of high
beams without glaring oncoming or leading vehicles. Another headlamp technology is curve adaptive
headlamps, which swivel with steering input to keep the headlamp pattern on the road when drivers
negotiate curves. Two recent studies reported that drivers detected dimly lit targets sooner with curve
adaptive low beam HID headlamps than with non-curve adaptive HID and halogen low beam headlamps
(Reagan et al. 2015), without introducing excessive levels of glare to oncoming drivers (Reagan et al.,
2014). All of these technologies have the potential to reduce nighttime crashes associated with driving in
the dark on unlit roads.
The goals of the current field observational study was to measure the rate of high beam headlamp
use when vehicles are on unlit or poorly-lit roads and isolated from traffic and to identify factors related
to high beam use. A primary research question was whether high beam use rates vary among rural,
3
transitional, or urban locations. Other research questions included whether high beam use varies by the
proximity of leading or opposing vehicles, the presence and amount of roadway lighting, on horizontal
curves compared with straight roadway segments, and by the direction of travel relative to an urban area.
Previous research has found that high beam use decreases with traffic volume (Hare and Hemion,
1968; Iragavarapu and Fitzpatrick, 2012; Sullivan et al. 2004). Given that traffic volume tends to be
higher on urban roads than on rural roads, high beam use rates were expected to be higher on rural roads
when other factors such as lighting were controlled for. We expected that high beam use would vary
inversely with the proximity of leading or opposing traffic but would not vary with the proximity of
following vehicles.
The effects of roadway lighting on high beam use are challenging to study because of the wide
variability in the quality of roadway lighting. In the current study, we examined the effect of an isolated
street light on a rural road and varied amounts of lighting in an urban environment. For the rural scenario,
we hypothesized that the high beam use rate would be greater prior to passing the light. For the urban
scenario, we expected high beam use to decline with increased roadway lighting.
Research on the effects of roadway curvature on high beam use is limited. Iragavarapu and
Fitzpatrick (2012) reported that high beam use was higher at observation sites located within 1 mile of
horizontal curves than at sites located farther away from curves. However, the researchers did not explain
these result. Similarly, there is little data about whether the direction of travel relative to an urban area
affects high beam use.
METHOD
Video cameras placed on the roadside and aimed at a measurement post were used to classify
vehicles as using high beam or low beam headlights. As a vehicle passed through an observation site, its
headlight pattern illuminated the post, and the video camera recorded the illumination pattern on the post.
Additionally, the camera kept a time signature of passing vehicles, and this was used to determine
whether a vehicle was isolated from opposing or leading traffic.
4
Photometry Equipment
Two cameras, recording opposite directions of traffic, and a measurement post were placed at
each observation site. Samsung Model HMX-F90 video cameras were mounted on tripods at a height of
approximately 1.4 m. Four identical measurement posts were constructed of light wooden trim boards, 8
feet [2.4 m] high and 1.125 inches [2.86 cm] wide, mounted vertically on plywood bases. Sand bags
secured the plywood bases to prevent the posts from tipping or swaying in wind. Strips of stiff white
paper mounted on two sides of the vertical trim boards provided diffusely reflecting surfaces upon which
headlamp illuminations fell. The white paper had a reflectance of 0.83 for tungsten-halogen light. Eight
identical cameras and four identical posts were assembled to provide for simultaneous photometry
(measurement of light) at multiple sites.
Observation Sites
Video field photometry was performed at 20 sites in Ann Arbor, Michigan, and the surrounding
county (Fig 1). The sites were organized into five groups of four sites each based on a variety of
characteristics including urban/rural location, road curvature, and presence and type of fixed road
lighting. Sites were classified as either in a rural area, on a boundary between a rural and urban area, or in
an urban area based on definitions used by the U.S. Census Bureau. For each group, photometry was
performed on two evenings. Within each group, the four sites were selected in pairs so that a majority of
traffic that passed through one site also passed through the other site. The paired sites allowed assessment
of changes in high beam use by individual drivers.
As indicated in Figure 1, the urban road site group was in Ann Arbor, and the sites had either no
roadway lighting, low levels of lighting, or high levels of lighting. The west and north site groups were
unlit, with sites selected to observe vehicles as they transitioned between rural and urban roadway
environments. The west boundary site group was west of Ann Arbor, and the north boundary group was
north of the city. The winding road site group was on unlit, winding roads located in rural areas or on the
boundary of Ann Arbor. The rural road site group was located in a rural environment, and each site pair
5
had an isolated street light on the section of road between them. With the exception of the urban road site
group in Ann Arbor, traffic flowing in both directions was observed. Eight of the 20 sites were on curved
roadway segments, with the remaining 12 on straight segments.
Video Collection Procedure
Data were collected on 12 evenings during May-August 2015. On most evenings, photometry
was performed at the four sites within a site group. One post was placed at each site. At all of the sites
except those in Ann Arbor, two cameras, one positioned on each side of the post and aimed at the post,
were used to observe traffic in both directions. At the sites in Ann Arbor, only one camera was used
because traffic on the other side of the road from the post was partially obscured; at these sites, high beam
use was measured for only the direction of traffic closest to the post. However, the camera did capture the
time gaps of traffic in the opposing direction. Video collection began at approximately the end of civil
twilight (when the sun is 6 degrees below the equator, or about an hour after sunset) and was conducted
only when the weather was clear and the roads were dry.
The measurement post was placed on the shoulder of the road, 2-3 m from the lane edge. Each
camera was approximately 10 m from the post. Figure 2 shows a video frame from a camera at a western
boundary site at dusk, prior to data collection. The second camera is behind the post in this view.
Video Data Processing
The video was processed with custom software developed in MATLAB 2015a and in
Mathematica 10.2.
Calibration
Field calibration was based on one evening of observations at one of the western boundary sites.
Video data on the illumination pattern of light from headlamps that shined on the post were calibrated
with the lux values that were recorded by Minolta lux meters placed at the top and bottom of the post. The
6
calibration data were used to translate the illumination patterns on the measurement post captured by the
video cameras to equivalent lux values.
Coding of High and Low Beam
The method for classifying headlamps as high or low beams combined an illuminance measure
taken when the vehicle was farther away from the measurement post with a measure of the vertical
gradient of the beam pattern when the vehicle was near the post. With regard to the measure of
illuminance, illuminance of the top 0.35 m of the post was averaged over 200 to 150 video frames before
the lamps of the approaching vehicle appeared in the field of view of the camera. At a speed of 45 mph,
this range of frames would correspond to a distance of 134 to 101 m from the measurement post. At this
distance, properly aimed low beam headlamps would not illuminate the top 0.35 m of the post. The
illumination that would be expected at the top 0.35 m of the post at these distances was simulated using
data on typical high and low beam headlamps (Schoettle and Flannagan, 2011). The simulation assessed
the effects of headlamp vertical aim on the illumination of the post, assuming a standard deviation of
vertical aim for vehicles of 0.75 degree (Flannagan, 2011).
Figure 3 presents the simulation results for median lux values for high and low beam headlamps
at five different vertical aims, ranging from ±2 standard deviations from the prescribed vertical aim (1.50, -0.75, 0.00, 0.75, 1.50 degrees). For ranges of ±1 standard deviation, there was no overlap between
high beam and low-beam lamps but substantial overlap for ±2 standard deviations. The criterion for an
initial classification of high beams based on illuminance was greater than 1.2 lux at the top 0.35 m of the
post averaged across the distance of approximately 134 to 101 m from the measurement post. This
criterion would properly classify high beams headlamps but would misclassify about 10% of low beams
that were misaimed upward as high beams.
For vehicles that met the lux criterion, finer discrimination between high and low beams was
provided by measuring the vertical gradient of the beam pattern. The measure was taken over the range of
75 to 30 video frames before the vehicle reached the observation site (corresponding to a range of
7
distance from 50 to 20 m at a speed of 45 mph). This is close enough to capture the low beam cutoff lines
for virtually all lamps with high misaim. For the cutoff line to be above the top of the post at 20 m, a low
beam lamp would have to be aimed more than 5 degrees up, or more than 6 standard deviations.
The vertical gradient measure was computed by first determining the amount of light on each of
seven equal vertical segments of the post (each about 0.35 m of the total height of the post of 2.44 m). For
the six boundaries between segments, a gradient measure was then computed by dividing the amount of
light on the segment below the boundary by the amount on the segment above the boundary. These values
will usually be close to 1.0 for high beam headlamps, which distributed light evenly on the post. The
gradient will be substantially higher than 1.0 for low beam headlamps, which project relatively more light
downward. The criterion level for the vertical gradient measure was determined by using the value that
returned the greatest phi value from a 2x2 table of lux and gradient measures, which was 2. Therefore,
headlamps that were classified as high beam on the basis of an illuminance greater than 1.2 lux were
reclassified as low beam if the gradient was greater than 2.0.
The resulting classifications are shown in Table 1. The 571 vehicles in the lower right cell of the
2x2 table (greater than 1.2 lux for illuminance and less than 2.0 for vertical gradient) were classified as
using high beams, and the vehicles in the three other cells of the table were classified as using low beams.
The phi coefficient for this table is -0.536, indicating a strong relationship between the two measures, as
would be expected based on the characteristic differences between low and high beams. As shown in
Table 1, 734 vehicles were classified as potentially using high beams by the illuminance criterion. The
upper right cell of the table indicates that the headlamps on 163 vehicles may have been low beams that
were aimed high, thus producing high illuminance levels at the top of the post from a distance of about
100 m.
Definition of Isolated Vehicles
All vehicle types were eligible for the study, including large trucks, buses, and motorcycles,
because all vehicle types have high beam and low beam headlamps. Only vehicles that were isolated from
8
other traffic were included. A vehicle was considered to be isolated if there was a minimum of 10 seconds
between it and a lead vehicle in the same lane (henceforth termed the leading vehicle) and no vehicle
passed the measurement post in the opposing lane of traffic in the 10 seconds before (the preceding
vehicle) or after (the oncoming vehicle) the subject vehicle passed the post.
Matching Vehicles
All observation sites were paired so that it was likely that a large portion of the vehicles passing
any site in either direction would soon pass or would have just passed the other site in the pair. For some
pairs, there was little opportunity for vehicles to leave or enter the traffic stream between the two sites, so
virtually all vehicles probably passed both sites. However, vehicles that were isolated at a single site may
have been too close to opposing or leading vehicles at the other site to be considered isolated.
For all site pairs, vehicles passing both sites were matched probabilistically on the basis of
relative timing between the sites. For each pair of sites and each direction of traffic, the two timelines for
vehicles passing the two sites were compared. Beginning with a rough estimate of the expected time
interval between when a single vehicle would be expected to pass the first and second sites (based on
distance and approximate speed), we determined the temporal offset between the two timelines that most
closely aligned the passing times of the highest number of potential matches between vehicle passes at the
two locations. The set of passes that were closely aligned in time by the chosen temporal offset were
considered to be matches, that is, both passes involved the same vehicle.
The data set excluded vehicles that could not be matched. The initial unmatched data set had
5,294 vehicle observations. The matched data set included 3,162 vehicle observations representing 1,581
vehicles that were observed across the 10 site pairs.
Analysis Approach
Data were analyzed with SAS version 9.3 by conducting descriptive analyses and a series of
repeated measures Poisson regressions that estimated the likelihood that a vehicle would be using high
beams. Vehicles matched across site pairs served as the repeated measure. One model assessed
9
differences in high beam use across all sites. Five models examined the high beam use rate within each
site group. The criterion for statistical significance was p < 0.05.
All models included the repeated measure for matched vehicles, a categorical independent
variable denoting road curvature (curved or straight), and four continuous variables denoting the time
gaps (in seconds) between the subject vehicle and the vehicles ahead of and behind it in the same lane and
the vehicles traveling in both directions in the opposing lane.
The model that assessed high beam use across all sites also included a categorical independent
variable for rurality (seven urban sites, five boundary sites, or eight rural sites).
With regard to the models for the five site groups, the model for the urban site group included a
categorical independent variable denoting the amount of street lighting (high, low, or absent). The models
that assessed high beam use within each of the two site groups on boundaries between Ann Arbor and
rural areas included categorical variables for rurality and for travel direction relative to Ann Arbor
(heading towards or away from Ann Arbor). The model for the rural site group included a categorical
independent variable to indicate a site’s position relative to an isolated street lamp located halfway
between each site pair.
RESULTS
Descriptive Statistics
Across the 20 sites, there were 3,162 observations of isolated vehicles. The quality of the
nighttime video was sufficient to distinguish at three categories of vehicles: (1) passenger cars, pickups,
SUVs, and vans; (2) larger vehicles such as trucks and buses; and (3) motorcycles. In a random sample of
200 cases, the counts in those categories were: 198, 1, and 1. High beams were used in 571 (18%) of these
observations. Table 2 provides the rates of high beam use at each site pair, along with the direction of
travel, rurality, and road curvature. The high beam use rate at individual sites was as low as 0.8% for the
eastbound traffic at the urban end of the west boundary site group. The highest rate of high beam use, just
10
more than 50%, occurred among northeast-bound vehicles at the second site pair (both rural sites) in the
north boundary group. Generally, high beam use was well below 50% at the other rural sites.
Table 3 indicates that the means and standard deviations of the time gaps between the subject
vehicles and other traffic passing through the sites were highly variable. As expected, mean time gaps
increased with rurality; for example, the mean (standard deviation) number of seconds between the
subject vehicles and leading vehicles in the urban, boundary, and rural sites was 76.41 (96.9), 127.76
(171.17), and 192.24 (356.57), respectively.
Overall Model Results
Seventeen (0.5%) of the 3,162 observations were excluded from analysis due to missing values
for the time gap between the subject vehicle and the closest vehicle traveling behind it, which reduced the
number of observations used in the model to 3,145.
Table 4 summarizes the estimated effect that each independent variable had on the percentage
increase in the likelihood that a vehicle would be using high beams, based on the overall Poisson
regression model. The curvature of the road and the proximity of following vehicles did not significantly
affect high beam use, but sites classified as boundary or rural areas had significantly higher rates of high
beam use than urban sites. Vehicles observed at boundary sites and at rural sites were estimated to be
94.1% and 79.8%, respectively, and more likely to have used high beams than vehicles observed at urban
sites.
For every 30-second increase in time between a subject vehicle and a preceding vehicle in the
opposing lane, high beam use was estimated to increase by 1.1%. Similarly, for every 30-second increase
in time between a subject vehicles and an oncoming vehicle, high beam use was estimated to increase by
1.4%. Finally, for every 30-second increase in time between a subject vehicle and a leading vehicle, high
beam use was estimated to increase by 2.3%.
Based on these results, we estimated rates of high beam use for curved and straight road sections
in urban, boundary, and rural areas, using the median values across observations for the four time-gap
11
variables that are seen in Table 3. With time gaps kept at these median values, estimates of high beam use
for straight road sections at urban sites, boundary sites, and rural sites were 9.2%, 17.8%, and 16.5%,
respectively. Predicted high beam use for curved sections at urban sites, boundary sites, and rural sites
were 10.2%, 19.9%, and 18.4%, respectively.
Model Results for Urban Site Group
Of the 163 pairs of vehicle observations made at either of the two pairs of urban sites, 81% used
low beams at each site, 11.7% used high beams at the first observation site and low beams at the second
observation site, 5.5% used low beams at the first site and high beams at the second observation site, and
1.8% used high beams at both sites.
The model indicated that subject vehicles on curved sections were 540% more likely to be using
high beams than subject vehicles on straight road sections (Table 4). Counter to expectations, high beam
use was 617% more likely at the site with high levels of street lighting compared with the two sites with
no street lighting. Compared with no lighting, the presence of low levels of street lighting did not affect
the likelihood of high beam use. As indicated in Table 4, the amount of the gap between the subject
vehicle and the closest vehicles in the same or opposing lanes did not affect high beam use.
Model Results for Western Boundary Site Group
There were 561 pairs of vehicles in the western boundary site group. About three-quarters
(78.1%) of the vehicles were using low beams at both sites, 7.3% used high beams at the first site and low
beams at the second site, 9.3% used low beams at the first site and high beams at the second, and 5.4%
used high beams at both sites.
As shown in Table 4, a significant effect for travel direction relative to Ann Arbor indicated that
vehicles heading into Ann Arbor were 24.5% less likely to use high beams than vehicles heading out of
Ann Arbor. Subject vehicles at the two sites in Ann Arbor were 60.1% less likely to use high beams than
vehicles observed at the two sites on the boundary between Ann Arbor and the rural outlying area. For
every 30-second increase in the gap between a subject vehicle and an oncoming vehicle, high beam use
12
was estimated to increase by 2.0%. High beam use did not vary reliably between the two sites on curves
compared with the two sites on straight road segments or for the gaps between a subject vehicle and a
leading or trailing vehicle or preceding vehicle in the opposing lane.
Model Results for Northern Boundary Group
There were 207 paired observations of vehicles in the northern boundary group. Fifty-five percent
of the vehicles used low beams activated at both sites, 10.1% used high beams at the first site and low
beams at the second site, 12.1% used low beams at the first site and high beams at the second, and 22.7%
used high beams at both sites.
As shown in Table 4, vehicles on curves were 19.5% more likely to use high beams compared
with vehicles on straight roadway segments. For every 30-second increase in time between the subject
vehicle and a leading vehicle, high beam use was estimated to increase by 1.5%. The amount of time
between the subject vehicle and an oncoming vehicle also significantly affected high beam use such that
for every 30-second increase in time, high beam use was estimated to increase by 0.9%. All of these
effects were significant, but the remaining effects in the model were not.
Model Results for Winding Road Group
There were 309 paired observations of vehicles passing the site pairs in the winding road group.
Seventy-three percent of the vehicles used low beams at both sites, 8.7% used high beams at the first site
and low beams at the second site, 10% used low beams at the first site and high beams at the second, and
8.1% used high beams at both sites.
As shown in Table 4, vehicles at the rural site were 52.7% less likely to use high beams than
vehicles at the boundary sites. For every 30-second increase in the gap between a subject vehicle and a
lead vehicle, high beam use increased by 3.2%. For every 30-second increase in the gap between a subject
vehicle and a preceding vehicle in the opposing lane, high beam use increased by 2.3%. All of these
effects were significant, but the remaining model effects were not (see Table 4).
13
Model Results for Rural Group
There were 341 paired observations of vehicles in the rural site group. Seventy-three percent of
the vehicles used low beams at both sites, 6.5% used high beams at the first site and low beams at the
second site, 7% used low beams at the first site and high beams at the second, and 13.2% used high beams
at both sites.
As indicated in Table 4, for every 30-second increase in the gap between a subject vehicle and a
leading vehicle, high beam use increased significantly by 4.1%. For every 30-second increase in the gap
between a subject vehicle and a preceding vehicle in the opposing lane, high beam use increased
significantly by 3.1%. The remaining effects were not significant.
DISCUSSION
The underuse of high beam headlamps first reported almost 50 years ago by Hare and Hemion
(1968) remains a significant problem. In the current study, vehicles were observed on unlit or poorly lit
roads and when isolated from other traffic in Ann Arbor, Michigan, and the surrounding county. Yet,
even when observed in these conditions, only 18% of vehicles used high beams. Most states, including
Michigan, direct drivers who are within 500 feet of an oncoming vehicle not to use lights that will aim
glaring beams into the eyes of the oncoming driver. Based on the definition of an isolated vehicle used in
the current study, the subject vehicles were well past this range. States generally encourage drivers to use
high beams when appropriate, although they do not require use.
Visibility with high beams is superior compared with low beams. For example, a recent field
experiment found that drivers could detect small (8 × 12 in.) gray roadside targets as much as 28.4 meters
earlier with high beams than with low beams (Reagan and Brumbelow, 2016). At the speed of 30 mph
driven during the study, 28.4 meters translates to an increase of 2.1 seconds for drivers to detect and react
to potential road hazards. While high beam use has been consistently low in prior research, the current
18% use rate is lower than the 24%-50% use rates reported in previous studies (Hare and Hemion, 1968;
Iragavarapu and Fitzpatrick, 2012; Sullivan et al., 2004). Part of this discrepancy may be because the
14
previous studies observed headlamp use primarily on two-lane rural roads, whereas the current study
included sites located in urban areas as well as in boundary areas between urban and rural regions. The
difference in use rates between the current and previous research may also have resulted from the use of
the vertical gradient of the headlamp beam to classify beams as low or high. This method decreased the
likelihood that low beams misaimed upward would have been misclassified as high beams, had
illuminance been the sole classification criterion as in previous studies (Iragavarapu and Fitzpatrick,
2012; Sullivan et al., 2004).
In the current study, a vehicle’s high beam use was tracked over two sites. Across the sample,
73% of the 1500-plus paired observations used low beams at each site in a pair, whereas 9% were
observed using high beams at both sites, so that less than 20% of drivers changed their high or low beam
settings between sites. Thus, not only was high beam use low at all sites, but drivers rarely modulated
headlight beams between locations located about a half-mile apart, on average.
As expected, vehicles at rural sites and sites at the boundaries of Ann Arbor were significantly
more likely (80% and 94% more likely, respectively) to use high beams than vehicles observed at urban
sites. However, counter to the expectation that the use rate would be highest at rural sites, there was no
significant difference between the likelihood of high beam use in rural areas compared with rural/urban
boundary areas. Although high beam use rates were higher at boundary and rural sites, rates at these sites
were still much lower than expected. A telephone survey about drivers’ motivations for high beam use
interviewed 600 drivers residing in Ann Arbor and the surrounding county (Reagan and Cicchino, 2016).
Respondents were asked how often they used high beams in the same types of driving environments as
those studied in the current project. Respondents reported greater high beam use rates than were observed
in the current study, with the largest discrepancy between self-reported and observed use in rural settings.
Specifically, 80% of drivers said they always or almost always use high beams on unlit rural roads, but
the current study estimated high beam use on unlit rural roads at 18% on straight sections and 20% on
curves.
15
Sullivan et al. (2004) and Hare and Hemion (1968) found that high beam use decreased with
increased traffic density. Rather than using a general measure of traffic density, the current method
obtained precise measurements of the proximity of other vehicles to the subject vehicles. By design, all of
the subject vehicles were far enough away from other vehicles that high beams could have been used
without bothering other drivers. However, high beam use still generally increased with greater time
between subject vehicles and leading vehicles and between subject vehicles and preceding and oncoming
vehicles in the opposing lane, although findings were not consistent across site groups. Despite the effects
for the proximity of other traffic, further inspection of the results suggests low high beam use rates were
very low even when vehicles were isolated from leading or opposing vehicles for several minutes.
The overall model indicated no significant change in high beam use for vehicles observed on
curves versus straight roadway segments. However, vehicles observed on curves within the northern
boundary site group and the urban site group were significantly more likely to use high beams than
vehicles observed on straight roads. Although not significant, the results for curves versus straight roads
for the rural site group and western boundary site group also suggested high beam use was higher on
curved roads. However, high beam use was not significantly different on curved versus straight roads for
the winding road site group. Iragavarapu and Fitzpatrick (2012) reported that sites within 1 mile of a
horizontal curve were associated with increased high beam use. Future research is needed to establish
whether high beam use differs on curved versus straight roads.
The current project also studied whether high beam use is affected by roadway lighting in two
respects. For each pair of sites in the rural site group, there was a street light located in the middle of each
pair about one-quarter mile from each site. It was expected that high beam use would be greater among
vehicles that had not reached the light than among vehicles that had passed it because drivers might
switch to low beams after having seen the street light and forget to reactivate high beams. However, this
hypothesis was not confirmed. In other analyses examining the effects of street lighting, sites in the urban
group allowed comparison of high beam use on streets with high and low levels of street lighting with an
urban parkway with no street lighting. Counter to expectations, the site with the most street lighting had
16
significantly higher high beam use than the sites with low levels of street lighting and no street lighting.
The site with the most lighting has more pedestrian traffic than the other urban sites, which may explain
in part why more vehicles used high beams based at this site. Despite differences in high beam use across
the urban sites, it is important to reiterate that the urban site group had the lowest mean high beam use
rate (10.5% based on a weighted average of the four sites) of the five site groups
Limitations
The findings of the current study may not generalize to other roads in and around Ann Arbor or to
roads in other areas of the United States. Hare and Hemion (1968) reported wide variability in high beam
use at sites located in the southeast, northeast, midwest, and northwest United States, and the current
study also observed a wide range in high beam use across sites. Relatedly, there were few candidate
winding roads available for site selection in the Ann Arbor area, and many of the curved road sites
observed had large radii (ranging from X to Y feet/meters). Observations occurred during late spring and
early summer. Drivers report using their headlights more often when they believe the risk of hitting deer
or other wildlife is increased (Reagan and Cicchino, 2016). Therefore, high beam use may be higher
during the fall when deer are most active than was observed in the current study.
The time-based definition of what constituted isolated vehicles in the current study was fairly
conservative. Drivers of vehicles observed at straight road sites may have dimmed from high beams to
low beams if they saw any vehicle, even if it were well beyond the range of a vehicle’s high beams. Hare
and Hemion (1968) noted that the distances at which drivers dimmed high beams due to oncoming
vehicles ranged from 1,082 feet to 2,573 feet. Thus, some observations in the current data may represent
drivers that dimmed their headlights out of concern of glaring drivers of vehicles that could be seen but
were further away than 10 seconds.
Advanced headlight technologies could address the substantial underuse of high beams. High
beam assist systems (also referred to as semi-automatic high beams) switch between low and high beams
based on the presence or absence of other traffic and ambient light. These systems would largely
17
eliminate underuse. However, in the survey of drivers in the Ann Arbor area (Reagan and Cicchino,
2016), only 43% said they would want a vehicle with a system that automatically switched between high
and low beams. Drivers who did not want such a headlamp system most often cited concerns that the
systems might not switch to low beams in time or a general mistrust of automation (see also Fekety et al.,
2013).
Adaptive driving beams also could reduce underuse of high beams. Like high beam assist, these
systems automatically provide high beam illumination in the absence of other road traffic. However,
adaptive driving beams continuously modulate the high beam pattern to create a shadow around other
vehicles. In this way, adaptive driving beams offer high beam visibility except for the segment of the
beam that is blocked out to limit glare for other oncoming or lead drivers. Such a continuously adjusting
system may be more acceptable to drivers than high beam assist systems that abruptly switch between low
and high beams.
However, adaptive driving beam systems are currently prohibited in the United States because
federal motor vehicle safety standards require headlights to have separate high beams and low beams.
One implementation of adaptive driving beams available in Europe was found to be associated with
acceptable levels of perceived discomfort glare and less discomfort glare than the low beams of some
vehicles presently available to consumers in the United States (Reagan and Brumbelow, 2015). Further,
the Society of Automotive Engineers plans to publish standards for adaptive driving beams to facilitate
the revision of the federal safety standards to allow this technology.
In the absence of advanced lighting technologies, manufacturers could optimize headlight
systems to create better visibility for drivers while still being mindful of glare. Two recent developments
should encourage such efforts. In spring 2016, the Insurance Institute for Highway Safety (2016) debuted
a performance-based rating scheme for headlight systems. Ratings are based on low beam and high beam
performance using five roadway scenarios (approaches from four curved trajectories and an approach
from a straightaway) and on acceptable glare from low beams. The system also credits vehicles that are
equipped with high beam assist. Similarly, the National Highway Traffic Safety Administration plans to
18
include low-beam headlight illumination performance and the provision of semi-automatic high beam
switching in the New Car Assessment Program (Office of the Federal Register, 2015).
In sum, maximizing visibility available to drivers from motor vehicle headlights includes
addressing the substantial underuse of high beam headlamps. Advanced technologies can help to address
this problem.
ACKNOWLEDGMENT
This work was supported by the Insurance Institute for Highway Safety.
REFERENCES
Flannagan MJ. Visual performance of headlighting systems and maintenance of aim in use. In TQ Khanh
(ed.) 9th International Symposium on Automotive Lighting. Munich, Germany: Herbert Utz Verlag;
2011:1-12.
Hare CT, Hemion RH. Headlamp Beam Usage on U.S. Highways. San Antonio, Texas: Southwest
Research Institute; 1968. Report no. AR-666.
Insurance Institute for Highway Safety. Headlight test and rating protocol (ver. I). Arlington, VA: Author;
2016.
Insurance Institute for Highway Safety. [Unpublished analysis of data from the Fatality Analysis
Reporting System]. Arlington, VA: Author; 2016.
Iragavarapu V, Fitzpatrick K. High beam usage on low volume rural roads in Texas. Transp Res Rec.
2012;2298:88-95.
Mefford ML, Flannagan MJ, Bogard SE. Real-World Use of High-Beam Headlamps. Ann Arbor, MI:
University of Michigan Transportation Research Institute; 2006. Report no. UMTRI-2006-11.
National Highway Traffic Safety Administration. Traffic Safety Facts, 2013: A Compilation of Motor
Vehicle Crash Data from the Fatality Analysis Reporting System and the General Estimates System.
Washington, DC: U.S. Department of Transportation; 2015. Report no. 812139.
Office of the Federal Register. Federal Register, vol. 80, no. 241, pp. 78522-78591. National Highway
Traffic Safety Administration, Request for comments, Docket no. NHTSA-2015-0119; New Car
Assessment Program. Washington, DC: National Archives and Records Administration; 2015.
Reagan IJ, Brumbelow ML. Drivers’ Detection of Roadside Targets when Driving Vehicles with Three
Different Headlight Systems during High Beam Activation. Arlington, VA: Insurance Institute for
Highway Safety; 2016.
Reagan IJ, Brumbelow ML. Perceived discomfort glare from an adaptive driving beam headlight system
compared with three low beam lighting configurations. Procedia Manufacturing 2015;3:3214-3221.
19
Reagan IJ, Brumbelow ML, Frischmann T. On-road experiment to assess drivers’ detection of roadside
targets as a function of headlight system, target placement, and target reflectance. Accid Anal Prev.
2015;76:74-82.
Reagan IJ, Cicchino, J. Phone survey of motivations for and attitudes about high beam use. Arlington,
VA: Insurance Institute for Highway Safety; 2016.
Reagan IJ, Frischmann T, Brumbelow ML. Test Track Evaluation of Headlight Glare Associated with
Adaptive HID, Fixed HID, and Fixed Halogen Low Beam Headlights. Arlington, VA: Insurance Institute
for Highway Safety; 2014.
Schoettle B, Flannagan MJ. A Market-Weighted Description of Low-Beam and High-Beam Headlighting
Patterns in the U.S., 2011. Ann Arbor, MI: University of Michigan Transportation Research Institute;
2011. Report no. UMTRI-2011-33.
Sullivan JM, Adachi G, Mefford ML, Flannagan MJ. High-beam headlamp usage on unlighted rural
roadways. Lighting Res Tech. 2004;36:59-67.
20
Table 1. Counts of vehicles classified by the illuminance and
vertical gradient criteria.
Illuminance
< 1.2 lux ≥ 1.2 lux
Total
Vertical
≥2
1979
163
2142
gradient
<2
449
571
1020
Total
2428
734
3162
Table 2. Vehicle direction, rurality, road curvature and observed high beam use for each site pair.
Site pairs
Vehicle
Rurality
Road curvature
High beam use N
by group
direction
Site 1
Site 2
Site 1
Site 2
Site 1 Site 2 (pairs)
Urban roads1
Pair 1
South
Urban
Urban
Straight Curve
2.7
17
112
Pair 2
West
Urban
Urban
Straight Curve
17.7
5.8
52
North boundary
Pair 1
North
Urban
Rural
Straight Straight 20.9
13.4
67
Pair 2
Northeast Rural
Rural
Straight Curve
51
52.9
68
Pair 2
Southwest Rural
Rural
Curve
Straight
37
33.3
27
Pair 1
South
Rural
Urban
Straight Straight 42.2
17.8
45
West boundary
Pair 1
West
Urban
Urban
Straight Curve
9.1
13.1
176
Pair 2
West
Boundary Boundary Curve
Straight 22.3
24.1
166
Pair 2
East
Boundary Boundary Straight Curve
16.7
18.9
90
Pair 1
East
Urban
Urban
Curve
Straight
3.1
0.8
129
Winding roads
Pair 1
East
Boundary Rural
Curve
Straight 16.4
12.9
140
Pair 1
West
Rural
Boundary Straight Curve
4
5
101
Pair 2
East
Boundary Boundary Straight Curve
38.7
41.9
31
Pair 2
West
Boundary Boundary Curve
Straight
46
43.2
37
Rural roads2
Pair 1
East
Rural
Rural
Straight Curve
22.4
22.4
58
Pair 1
West
Rural
Rural
Curve
Straight
24
14
50
Pair 2
East
Rural
Rural
Straight —
22.1
19.8
86
Pair 2
West
Rural
Rural
Straight —
18.4
19.1
147
1
Street light levels for urban roads group: site pair 1 had no street lighting; pair 2 site 1 had high levels
of street lighting; pair 2 site 2 had low levels of lighting.
2
For rural roads group, high beam use site 1 column reflects pre-street lamp, site 2 column reflects poststreet lamp.
Table 3. Mean (standard deviation), median, minimum, and maximum time gaps (seconds) between
subject vehicles and nearest traffic in same or opposing lane of travel.
Other vehicle’s position
relative to subject vehicle
Mean (std dev)
Median Minimum Maximum
Trailing
101.38 (203.27)
41.91
0.27
3470.37
Leading
117.40 (168.87)
62.28
10.01
2683.78
Oncoming vehicle, opposing lane
136.84 (253.70)
63.03
10.04
2859.09
Preceding vehicle, opposing lane
131.57 (234.35)
61.78
10.01
2789.79
21
Table 4. Summary of results from Poisson regression models.
Percent change in
Independent variable
likelihood (95% CI)
Overall
Curved vs. straight road
11.4 (-2.4, 27.2)
Boundary vs. urban site
94.1 (51.8, 148.1)
Rural vs. urban site
79.8 (40.2, 130.5)
Proximity of leading vehicle (30 s)
2.3 (1.2, 3.4)
Proximity of trailing vehicle (30 s)
0.1 (-0.9, 1.2)
Proximity of approaching vehicle (30 s)
1.4 (0.9, 2.0)
Proximity of preceding vehicle (30 s)
1.1 (0.3, 1.8)
Urban site group
Curved vs. straight road
540 (114, 1815)
High street lighting vs. no lighting
617 (116, 2278)
Low street lighting vs. no lighting
-64.6 (-89.4, 17.6)
Proximity of leading vehicle (30 s)
2.9 (-9.6, 17.1)
Proximity of trailing vehicle (30 s)
9.5 (-2.3, 22.7)
Proximity of approaching vehicle (30 s)
-2.7 (-15.4, 12.0)
Proximity of preceding vehicle (30 s)
-3.2 (-13.0, 7.7)
Western boundary site group
Curved vs. straight road
14.8 (-10.6, 47.3)
Heading toward vs. away from Ann Arbor
-24.5 (-37.7, -8.6)
Urban vs. boundary site
-60.1 (-73.1, -41.0)
Proximity of leading vehicle (30 s)
2.5 (-2.8, 8.1)
Proximity of trailing vehicle (30 s)
-0.9 (-6.9, 5.5)
Proximity of approaching vehicle (30 s)
2.0 (0.0, 4.0)
Proximity of preceding vehicle (30 s)
1.4 (-1.7, 4.5)
Northern boundary site group
Curved vs. straight road
19.5 (0.2, 42.5)
Heading toward vs. away from Ann Arbor
3.5 (-14.1, 24.7)
Rural vs. urban site
33.8 (-7.1, 92.6)
Proximity of leading vehicle (30 s)
1.5 (0.4, 2.6)
Proximity of trailing vehicle (30 s)
-0.4 (-1.7, 0.9)
Proximity of approaching vehicle (30 s)
0.9 (0.3, 1.5)
Proximity of preceding vehicle (30 s)
0.3 (-0.6, 1.2)
Winding road site group
Curved vs. straight road
-23.7 (-45.6, 7.1)
Rural vs. boundary site
-52.7 (-72.9, -17.3)
Proximity of leading vehicle (30 s)
3.2 (0.8, 5.7)
Proximity of trailing vehicle (30 s)
-0.1 (-2.5, 2.4)
Proximity of approaching vehicle (30 s)
1.7 (-1.5, 5.1)
Proximity of preceding vehicle (30 s)
2.3 (0.0, 4.7)
Rural site group
Curved vs. straight road
14.7 (-11.2, 48.2)
Heading away from vs. toward street light
-5.0 (-14.1, 5.0)
Proximity of leading vehicle (30 s)
4.1 (0.2, 8.0)
Proximity of trailing vehicle (30 s)
0.9 (-2.4, 4.3)
Proximity of approaching vehicle (30 s)
0.7 (-2.3, 3.7)
Proximity of preceding vehicle (30 s)
3.1 (0.6, 5.7)
22
z
p
1.60
5.29
4.62
4.07
0.20
5.31
2.88
0.11
<0.0001
<0.0001
<0.0001
0.84
<0.0001
0.004
3.32
3.22
-1.69
0.43
1.57
-0.38
-0.60
0.0009
0.001
0.09
0.67
0.12
0.71
0.55
1.08
-2.87
-4.60
0.91
-0.29
2.00
0.87
0.28
0.004
<0.0001
0.37
0.77
0.05
0.38
1.98
0.36
1.57
2.71
-0.63
2.81
0.67
0.05
0.72
0.12
0.007
0.53
0.005
0.51
-1.56
-2.63
2.66
-0.08
1.03
1.98
0.12
0.009
0.008
0.93
0.30
0.05
1.05
-1.00
2.08
0.54
0.44
2.40
0.29
0.32
0.04
0.59
0.66
0.02
Fig 1. Twenty roadside observation sites; shading shows
urban areas as defined by the U.S. Census Bureau.
Fig 2. Video frame from a western boundary site showing
position of post just rightward of pavement edge. Video
was taken at dusk, prior to data collection.
Fig 3. Results of simulated illumination of top of measurement post
from median low beam and high beam headlamps at a distance of
100 m, with five levels of vertical aim.
23