Safety Evaluation of Billboard Advertisements on Driver Behavior in Work Zones
A thesis presented to
the faculty of
the Russ College of Engineering and Technology of Ohio University
In partial fulfillment
of the requirements for the degree
Master of Science
Patrick J. Fry
May 2013
© 2013 Patrick J. Fry. All Rights Reserved.
2
This thesis titled
Safety Evaluation of Billboard Advertisements on Driver Behavior in Work Zones
by
PATRICK J. FRY
has been approved for
the Department of Civil Engineering
and the Russ College of Engineering and Technology by
Deborah S. McAvoy
Associate Professor of Civil Engineering
Dennis Irwin
Dean, Russ College of Engineering and Technology
3
ABSTRACT
FRY, PATRICK J., M.S., May 2013, Civil Engineering
Safety Evaluation of Billboard Advertisements on Driver Behavior in Work Zones (207
pp.)
Director of Thesis: Deborah S. McAvoy
While the number of fatal crashes occurring in work zones has been declining
over the last few years, work zone safety is still considered to be a nationwide priority.
Recently, a work zone safety campaign consisting primarily of billboard advertisements
was found to be ineffective. Thus, a two-phase study was performed to determine how
the billboards could be improved to communicate more effectively the campaign’s safety
message to the public. The first phase consisted of a public opinion survey designed to
determine the preferred background and text color combination for the advertisement.
Survey responses were evaluated through statistical analyses and a yellow background
with black text is recommended for use in future campaign billboards. The second phase
consisted of a driving simulator study to examine the slogans and graphics used in the
advertisement along with the placement and orientation of the billboard with respect to
the driver. Driving performance and glance behaviors of the participants were monitored
throughout the simulated drive to determine what effect the billboards had on their
behavior. Based on statistical analyses of driving performance, no elements of the
billboard advertisements were found to improve safety on the roadway. Glance behaviors
were found to be significantly impacted by the presence of an advertisement, but were
not impacted by any particular element. Furthermore, as drivers made long glances to the
4
billboards their ability to maintain lane position was negatively impacted. Ultimately,
billboard advertisements should be used alongside other approaches to improve work
zone safety through an effective campaign.
5
ACKNOWLEDGMENTS
I would first like to express my sincere gratitude to my advisor, Dr. Deborah
McAvoy, for her continued support and encouragement throughout my Master of Science
program and into my professional career. I would not have succeeded in my graduate
studies without her guidance. Next, I would like to thank my committee members, Dr.
Eric Steinberg, Dr. Sang-Soo Kim, and Dr. Natalie Kruse Daniels, for their time and
assistance with the completion of my thesis research. I would also like to thank my fellow
researchers in the ORITE Safety and Human Factors Facility for their help in processing
the data throughout this research. Finally, I would like to thank my family, fiancée, and
friends for their constant love, encouragement, and support throughout my graduate and
professional careers.
6
TABLE OF CONTENTS
Page
Abstract ............................................................................................................................... 3
Acknowledgments............................................................................................................... 5
List of Tables ...................................................................................................................... 8
List of Figures ................................................................................................................... 10
Chapter 1: Introduction ..................................................................................................... 11
Chapter 2: Background ..................................................................................................... 14
2.1 Impact of Speeds in Work Zones ............................................................................ 14
2.2 Evaluating Safety Campaigns ................................................................................. 18
2.3 Use of Billboards in Safety Campaigns .................................................................. 22
2.4 Safety of Billboards ................................................................................................ 26
2.4.1 Impact on Glance Behavior ............................................................................. 26
2.4.2 Impact on Driving Performance ....................................................................... 37
Chapter 3: Public Survey Study ........................................................................................ 43
Chapter 4: Driving Simulator Study ................................................................................. 52
4.1 Selection of Methodology....................................................................................... 52
4.2 Research Equipment Used ...................................................................................... 53
4.3 Institutional Review Board Process ........................................................................ 55
4.4 Design of Scenarios ................................................................................................ 57
4.4.1 Scenario #1 – Slogans ...................................................................................... 58
4.4.1 Scenario #2 – Graphics .................................................................................... 62
4.4.2 Scenario #3 – Placement and Orientation ........................................................ 63
4.5 Study Procedure ...................................................................................................... 65
4.6 Data Collection ....................................................................................................... 73
4.7 Sample Size Determination .................................................................................... 80
4.7.1 Determination for Speed .................................................................................. 81
4.7.2 Determination for Lane Position ...................................................................... 83
4.7.3 Determination for Acceleration and Deceleration ........................................... 84
4.7.4 Determination for Mean Fixation and Total Fixation ...................................... 85
7
4.7.5 Determination for Number of Fixations........................................................... 88
4.7.6 Determination for Proportion of Fixation Duration ......................................... 89
4.7.7 Final Sample Size Determination .................................................................... 91
Chapter 5: Statistical Methodology .................................................................................. 93
5.1 Statistical Tests for Public Survey Study ................................................................ 93
5.2 Statistical Tests for Driving Simulator Study ......................................................... 97
Chapter 6: Results ........................................................................................................... 111
6.1 Public Survey Study.............................................................................................. 111
6.1.1 Increase Awareness of Work Zones............................................................... 114
6.1.2 Increase Caution in Work Zones .................................................................... 118
6.1.3 Billboard Seen Best in Daylight .................................................................... 123
6.1.3 Billboard Seen Best at Nighttime .................................................................. 126
6.1.4 Overview of the Statistical Analyses of Survey Responses ........................... 130
6.2 Driving Simulator Study ....................................................................................... 131
6.2.1 Scenario #1 – Slogans .................................................................................... 134
6.2.2 Scenario #2 – Graphics .................................................................................. 142
6.2.3 Scenario #3 – Placement and Orientation ...................................................... 146
6.2.4 Impact of Long Glances on Safety ................................................................. 148
Chapter 7: Conclusions ................................................................................................... 152
7.1 Public Survey Study.............................................................................................. 152
7.2 Driving Simulator Study ....................................................................................... 155
7.3 Recommendations for Future Research ................................................................ 162
References ....................................................................................................................... 164
Appendix A: Public Survey Study Materials .................................................................. 169
Appendix B: IRB Documents ......................................................................................... 173
Appendix C: Billboard Advertisements Used in Simulator Study ................................. 200
Appendix D: Results of the Games-Howell Tests on the Data Collected in the Driving
Simulator Study .............................................................................................................. 204
8
LIST OF TABLES
Page
Table 1: Background and text color combinations used in surveys.................................. 44
Table 2: Slogans used in Scenario #1 ............................................................................... 59
Table 3: Placement and orientation combinations used in Scenario #3 ............................ 63
Table 4: Initial sample size determination for speed ........................................................ 82
Table 5: Initial sample size determination for lane position ............................................. 83
Table 6: Initial sample size determination for acceleration .............................................. 85
Table 7: Initial sample size determination for mean fixation ........................................... 87
Table 8: Initial sample size determination for total fixation ............................................. 87
Table 9: Initial sample size determination for number of fixations .................................. 89
Table 10: Initial sample size determination for proportion of fixation duration .............. 90
Table 11: Summary of initial sample size determination ................................................. 91
Table 12: Results of z-test for total population on increased awareness of work zones 115
Table 13: Results of z-test by age group for increased awareness of work zones .......... 115
Table 14: Results of z-test by gender for increased awareness of work zones ............... 116
Table 15: Results of z-test by driving experience for increased awareness of work zones
......................................................................................................................................... 116
Table 16: Results of z-test by knowledge of rules for increased awareness of work zones
......................................................................................................................................... 118
Table 17: Results of z-test for total population on increased caution in work zones ..... 119
Table 18: Results of z-test by age group for increased caution in work zones ............... 120
Table 19: Results of z-test by gender for increased caution in work zones .................... 120
Table 20: Results of z-test by driving experience for increased caution in work zones . 121
Table 21: Results of z-test by knowledge of rules for increased awareness of work zones
......................................................................................................................................... 123
Table 22: Results of z-test for total population on billboard seen best in daylight ........ 124
Table 23: Results of z-test by age group for billboard seen best in daylight .................. 124
Table 24: Results of z-test by gender for billboard seen best in daylight ....................... 125
Table 25: Results of z-test by driving experience for billboard seen best in daylight .... 125
Table 26: Results of z-test for total population on billboard seen best at nighttime ...... 127
Table 27: Results of z-test by age group for billboard seen best at nighttime ................ 128
Table 28: Results of z-test by gender for billboard seen best at nighttime ..................... 128
Table 29: Results of z-test by driving experience for billboard seen best at nighttime .. 129
Table 30: Results of the one-way ANOVA tests on driver behavior in Pre-Work Zone of
Scenario #1...................................................................................................................... 135
Table 31: Results of the one-way ANOVA tests on driver behavior in Advanced Warning
Area of Scenario #1 ........................................................................................................ 139
Table 32: Results of the one-way ANOVA tests on driver behavior in Work Zone of
Scenario #1...................................................................................................................... 140
Table 33: Results of the one-way ANOVA tests on driver behavior in Pre-Work Zone of
Scenario #2...................................................................................................................... 143
9
Table 34: Results of the one-way ANOVA tests on driver behavior in Advanced Warning
Area of Scenario #2 ........................................................................................................ 145
Table 35: Results of the one-way ANOVA tests on driver behavior in Work Zone of
Scenario #2...................................................................................................................... 146
Table 36: Results of the one-way ANOVA tests on driver behavior in approaching the
billboards in Scenario #3 ................................................................................................ 147
Table 37: Results of the one-way ANOVA tests on total variation in lane position while
participants were glancing at the billboards by scenario ................................................ 150
10
LIST OF FIGURES
Page
Figure 1. Billboard advertisement used in Cleveland. ..................................................... 43
Figure 2. Driving simulator located in ORITE Safety and Human Factors Facility. ....... 54
Figure 3. Sample billboard advertisement used in Scenario #1. ...................................... 60
Figure 4. Work zone configuration used in Scenario #1. ................................................. 61
Figure 5. Example of a face model created using the faceLab 5 eye-tracking system..... 70
Figure 6. Public survey responses by age group. ........................................................... 112
Figure 7. Proportions for the total population on increased awareness of work zones. . 114
Figure 8. Proportions for the total population on increased caution in work zones. ...... 119
Figure 9. Proportions for the total population on billboard seen best in daylight. ......... 123
Figure 10. Proportions for the total population on billboard seen best at nighttime. ..... 127
Figure 11. Mean values for lane position in the Pre-Work Zone of Scenario #1. .......... 137
Figure 12. Mean values for number of fixations in the Pre-Work Zone of Scenario #1. 138
Figure 13. Results of post-test questionnaire on recalling the slogans used. ................. 141
11
CHAPTER 1: INTRODUCTION
Currently, the emphasis in roadway construction is to conduct maintenance and
rehabilitation on existing roadways rather than construct new roadways. Consequently,
there are an increased number of work zones present on the roadways. Work zones are
put in place in order to guide drivers through a set of roadway conditions which they
would typically not expect to encounter. As a result of these unexpected conditions, there
is an increased safety risk for both drivers and construction workers. During 2011, there
were 530 fatal crashes that occurred in work zones nationwide (1). Comparatively, there
were 586 and 680 fatal crashes that occurred in work zones across the nation during 2010
and 2009, respectively (1). While this data indicates that the total number of fatal crashes
in work zones has been decreasing over the last few years, work zone safety is still
considered to be a nationwide priority.
Work zone safety is an issue that is important to many entities that comprise our
roadway systems; from the drivers on the roadway, to the workers in the work zones, to
the transportation officials funding the construction projects. In 2005, through research
sponsored by the American Association of State Highway and Transportation Officials
(AASHTO) and the Federal Highway Administration (FHWA), the National Cooperative
Highway Research Program (NCHRP) Report 500 (Volume 17) was released. This report
outlined guidelines for how to reduce crashes in work zones and contained six primary
objectives for such (2). These objectives cover various approaches for improving safety
in work zones such as decreasing the impact of work zones on the roadway and
improving the traffic control devices used in work zones. Of particular interest to this
12
research is the objective on approaches for increasing public awareness of work zones.
One of the main approaches outlined for achieving this objective is a safety awareness
campaign such as the National Work Zone Awareness Week.
The first National Work Zone Awareness Week was held in April 2000 as a joint
effort between the FHWA, AASHTO, and the American Traffic Safety Services
Association (ATSSA) (3). The primary purpose of this campaign was to increase
nationwide public awareness regarding the issue of work zone safety through a large
coordinated event. The National Work Zone Awareness Week is still an event ongoing
today with the next event scheduled for April 15-19, 2013 (4).
Additional work zone safety campaigns are conducted throughout the year at the
local level by various groups including individual state Departments of Transportation
and local construction organizations. These campaigns typically focus on advocating the
campaign’s safety message through slogans. Recently in Ohio, a work zone safety
campaign using the slogan “Look Up, Hang Up, and Go Slow for the Cone Zone” was
conducted by the Ohio Department of Transportation in order to communicate to drivers
the increased risk of distracted driving while in work zones (3). Media used in this safety
campaign included displaying posters at rest areas and transmitting the campaign slogan
on electronic message boards located along the major highways. Additional media which
are typically used in local-level work zone safety campaigns include broadcasting
advertisements on the television, radio, newspaper, and internet, as well as placing
advertisements on billboards and transit vehicles.
13
In 2010, the Laborers’ Local 860 in Cleveland, Ohio conducted a work zone
safety campaign that primarily consisted of billboard advertisements. The billboards
used in the campaign included several lines of orange text on a black background with a
barrel caricature on the right and the Laborers’ Local 860 logo on the left. The billboards
were located on the rooftops of several buildings adjacent to Interstate-90 in Cleveland.
However, through an unofficial survey of individuals who drove past the signs, it was
found that the billboards were not easily noticed and the safety message was not received.
As a result, Laborers’ Local 860 desired to understand how the billboards could be
improved to communicate more effectively the campaign’s safety message to the public.
Researchers at Ohio University performed a two-phase study in order to
determine how a work zone safety campaign using billboards would be most effective at
communicating to drivers the need to slow down and drive safely through work zones.
During the first phase of this study, a public survey was created to test eight different
color combinations while using the current advertisement content. Public surveys were
distributed in the following three cities across the State of Ohio: Athens, Cleveland, and
Chillicothe. Results of the surveys were analyzed to determine which color combination
was preferred by the public. The second phase consisted of a driving simulator study to
examine the slogans and graphics used in the advertisement along with the placement and
orientation of the billboard with respect to the driver. Participants encountered the
billboards prior to entering work zones along an urban four-lane undivided highway.
Driving performance and glance behaviors of the participants were monitored throughout
the simulated drive to determine what effect, if any, the billboards had on their behavior.
14
CHAPTER 2: BACKGROUND
2.1 Impact of Speeds in Work Zones
Typically when a work zone is set up on a roadway a reduced speed limit is
implemented in an attempt to improve the safety of the work zone for both drivers and
workers. However, driver compliance with the reduced speed limits is widely recognized
as a work zone safety issue. This is evident from examining the recent crash data for
work zones. In 2010, speeding was identified as a factor in 34 percent of the fatal crashes
occurring in work zones nationwide (5). Likewise, 36 percent of the fatal crashes
occurring in work zones nationwide were reported as speeding-related during 2011 (1).
Examining the characteristics of the at-fault drivers in these crashes provides valuable
insight into potential groups that may be impacting work zone safety. From the 2011
crash data, the plurality of the at-fault drivers were in the 16-25 age range (28 percent)
and the majority of those drivers were male (86 percent) (1).
Li and Bai found similar results for crashes in highway work zones in Kansas that
occurred between 1992 and 2004, as 15 percent of the fatal crashes and 20 percent of the
injury crashes were speeding-related (6). Additionally, male drivers were found to be at
fault in 64 percent of the injury crashes and 75 percent of the fatal crashes investigated
(6). Based on the crash data, male drivers were determined to be almost two times more
likely to be involved in a fatal crash than female drivers. Upon examining the impact of
driving environment, Li and Bai found urban two-lane highways are the environment on
which fatal crashes in work zones were the most likely to occur (6). Thus, the nationwide
and Kansas crash data indicate that certain groups and/or areas could be potentially
15
targeted in efforts to reduce the number of crashes occurring in work zones and improve
safety.
Another means of examining crash data is to consider crash rates as they account
for the exposure level of drivers. Crash rates are typically provided based on the number
of vehicle-miles traveled (VMT). In 1999, the total number of fatalities nationwide was
41,717 and dropped to 33,808 in 2009, a decrease of 19 percent (7). However, when
examining the fatality rates per 100 million VMT the decrease is more substantial (26
percent) as the rate fell from 1.55 in 1999 to 1.14 in 2009 (7). Subsequently, the overall
safety of the nation’s roadways is viewed slightly different when comparing the number
of fatalities and the fatality rates. The fatality rates are a better indication of the actual
risk to drivers since they account for the amount of driving occurring each year.
Unfortunately, the amount of VMT for work zones nationwide is not specifically known.
However, the VMT for work zones would be considerably lower than the VMT for all
roadways since work zones are setup for limited durations and are only present on a
portion of the total roadway system. Consequently, the driving exposure to work zones
would be lower and, thus, work zones would most likely have higher fatality rates per
100 million VMT than for all roadways.
Researchers have previously used crash rates per VMT as a means of evaluating
safety. In 1999, Migletz et al. studied the impact of speeds on crashes for 66 work zone
sites nationwide which totaled 444.9 miles in length (8). The total crash rates per million
VMT was found to be 1.63 prior to construction beginning, but then increased to 1.74
during construction when work zones were in place (8). By examining only the fatal and
16
injury crashes, the crash rates were found to have increased by 4.1 percent before
construction began to during the construction period (8). This indicates that, for a given
section of roadway, the presence of work zones may create unsafe conditions for drivers.
Crash rates per 100 million VMT were used to explore the relationship between
speed and crashes as early as 1964. Solomon conducted spot speed studies on rural
highways at 35 study sections, each of which averaged 17 miles in length, across 11
different states (9). Rural highways were selected as they were determined to carry
approximately a third of the VMT occurring on highways across the nation. For each of
the study sections, volume counts were taken periodically and used to compute the total
VMT, which was then distributed among drivers based upon the speed distributions that
were determined through spot speed studies. Crash rates per 100 million VMT could then
be determined for drivers across an entire range of speeds. The crash rates by speed were
then compared to the measured average speed for each highway section and used to
determine the crash rates by the amount of variation from the average speed of traffic.
Solomon found that as the variation from mean speed increased, the crash risk was also
found to increase (9). This behavior was seen to be parabolic with the curve centered
between zero and ten miles per hour (mph) above the mean speed. Also, as speeds
increased, crash rates with high severity levels increase at faster rates than those with low
severity levels.
Further relationships between speed and crashes have been considered by other
researchers. Hauer developed models in which the number of fatal and injury crashes can
be estimated based on the change in mean speed and from the number of crashes
17
occurring (10). These models can help quantify the possible impact on crashes
corresponding to a specific change in mean speed. Based on previous research they
reviewed, Harsha and Hedlund were able to state a general rule of thumb in regards to the
relationship between speed and crashes: a one percent increase in speed resulted in a two
percent increase in injury crashes, a three percent increase in serious injury crashes and a
four percent increase in fatal crashes (11).
Not only are speeds in work zones recognized as a safety issue by researchers, the
public also recognizes the issue exists. In 1999, Kane et al. collected public opinions on
the issue of work zone safety through responses from focus groups and surveys (12).
From the focus group discussions, the prevailing opinion of the groups were that
speeding was the main safety problem in highway work zones. The majority of this
speeding is believed to stem from driver attitudes, such as impatience and indifference, or
from drivers simply ignoring traffic signs. With regards to communicating the need to
slow down and be cautious while driving in work zones, the focus groups indicated that
signs with slogans, such as “Let Them Work – Let Them Live,” would serve as effective
reminders (12).
Surveys were distributed across North Carolina and responses were collected
from 487 passenger car drivers and 58 truck drivers (12). Based on the survey responses
received, a majority of both passenger car and truck drivers indicated that they reduce
their speeds to either the posted speed limit or the speed of the surrounding traffic while
driving through work zones. The researchers later conducted spot speed studies in three
work zones that were located on four-lane highways in order to evaluate driver
18
compliance with the posted speed limits of 55 mph. Speeds were recorded in each of the
travel lanes at all three sites and it was determined that over 90 percent of drivers in the
outside lanes and 85 percent of drivers in the inside lanes exceeded the posted speed
limits (12). This study shows that while the public recognizes there is a safety risk posed
by speeding in work zones; however, the risk does not appear to be significant enough to
cause them to drive slower through work zones.
2.2 Evaluating Safety Campaigns
As discussed previously, safety campaigns are widely used to increase public
awareness of work zone hazards and to communicate specific behaviors to drivers. From
the previous section, speeds in work zones were identified to be a major component to
work zone safety. Accordingly, a multitude of work zone safety campaigns attempt to
communicate to drivers the need to reduce their speeds when driving through work zones.
In 2003, the California Department of Transportation (Caltrans) conducted a statewide
work zone safety campaign using the slogan “Slow for the Cone Zone” (13). The slogan
had been used in a previous campaign that was conducted in Sacramento and the San
Francisco Bay Area. The vast majority of drivers in both areas reported reducing their
speeds through work zones and being more alert to the presence of construction workers.
These self-reported behaviors also coincided with a 25 percent decrease in total crashes
occurring in work zones and no fatal crashes occurring over a two-year period (13).
For the statewide safety campaign, Caltrans utilized multiple media outlets in
order to reach drivers before they began their commute, by airing television commercials
during the morning news, or during their commute, by using billboards and radio
19
commercials. Through these media outlets, Caltrans expected the campaign message to
be seen an average of 38 times by a typical driver throughout the duration of the
campaign with the target audience being those drivers in the 25 to 49 age range (13).
Safety campaigns have been used in similar ways to target other issues and raise
public awareness. Phillips and Torquato reviewed 45 anti-speeding campaigns in 2009
which were mainly conducted in Europe and Australia in order to determine the target
audience, media, and type of content which were most commonly used (14). Not all of
the studies identified a target audience, but the majority of those studies that did focused
on young males. This follows the results of the speeding-related crash data discussed
previously in which young males were the most common group involved in the crashes.
With respect to media outlets, television advertisements were used in 80 percent of the
campaigns with the next highest media outlet being outdoor advertising, or billboards and
posters on buses, at 26 percent (14). In regards to content of the campaign message, the
top three types of approaches used were rational, hard-hitting emotional, and those
informing on the potential harm caused to others. Self-reported behaviors were available
for 20 percent of the studies reviewed, all of which indicated reductions in speed, yet the
actual change in driver behavior on the roadways was not quantified in any of the studies
(14).
In 2011, Adamos and Nathanail developed a tool for evaluating safety campaigns
based on the design of the campaign and then estimating the potential impact on driver
behavior (15). The proposed tool consists of a multidimensional matrix of correlations
that uses variables corresponding to how a particular safety campaign is structured. Of
20
primary interest to this research are the variables associated with the level on which the
campaign is conducted, the target audience, the primary purpose of the campaign
(increasing awareness, changing driver behavior, etc.), and how media is used to
communicate the message. Each of these variables have been identified as means of
evaluating safety campaigns in previous research and are known to affect the success of a
campaign.
The evaluation tool developed by Adamos and Nathanail was tested using a locallevel campaign related to drinking and driving and a national campaign about driving
while fatigued (15). The local-level campaign was focused on increasing awareness of
the problem of drinking and driving among young drivers (18-30 years old) by
distributing leaflets, displaying posters, and conducting a workshop. The aim of the
national campaign was to increase risk awareness of driving while fatigued. The
campaign had a target audience of professional drivers, but was also expected to reach
non-professional drivers. The campaign message was communicated through a variety of
mediums including broadcasts of television and radio commercials and the distribution of
leaflets. For the local-level campaign, drivers were not found to significantly change their
intentions to not drink and drive from the before to after campaign periods. However, for
the national campaign, drivers were found to be significantly less likely to drive while
fatigued. Each of the campaigns was evaluated solely based on self-reported behaviors
and evaluations were not made regarding actual driver behavior. This is likely due to the
campaign design and the evaluation tool used to assess the impact on public awareness.
21
Wundersitz, Hutchinson, and Woolley examined fourteen safety campaigns which
used mass media and were conducted between 2001-2009 in order to assess what
campaign mediums were found to be effective (16). However, conclusions could not be
drawn on specific elements as the campaigns reviewed had varying characteristics, such
as their objectives or outcomes. The researchers did find that a large portion of the
campaigns were only evaluated based upon self-reported behaviors or the ability to
remember the message instead of changes in actual driver behavior. This is consistent
with previous research reviewed as most focus on whether or not the campaign messages
were seen.
Wundersitz, Hutchinson, and Woolley also argue that the frequency of crashes
may not be an appropriate evaluation measure for safety campaigns due to the variable
nature of crash data (16). While it is true that crash data can vary on a year-to-year basis,
observations of crash data over an extended period of time can indicate trends that are
occurring on the roadways. Furthermore, crashes are an essential safety measure and need
to be considered in some capacity. Thus, a combination of measures should be examined
when evaluating safety campaigns in order to assess their overall impact.
In 2011 Hoekstra and Wegman discussed the need to evaluate the safety
campaigns that are conducted in order to assess the effectiveness of a campaign and
determine what improvements could be made to future campaigns (17). Safety campaigns
will generally have a specific behavior and audience which are selected to be the primary
focus of the campaign. By examining how the safety campaign attempts to alter driver
behavior, a campaign may be adapted to produce better results. Hoekstra and Wegman
22
present driving as a type of automatic behavior, which people are typically more likely to
alter when they are in a position to reconsider their behavior and implement a change
(17). Thus, presenting a work zone safety message to drivers prior to their entry into a
work zone provides drivers an opportunity to implement the behavior presented in the
campaign message. Hoekstra and Wegman also suggest that safety campaigns need to be
pretested to see if the campaign will potentially produce the desired change in behavior
(17). Testing safety campaigns in a simulated driving environment could provide
opportunities to refine how the campaign is presented and examine the effects in a
controlled setting.
2.3 Use of Billboards in Safety Campaigns
Billboards have been used in many safety campaigns throughout the nation in
order to target a variety of issues related to driving. Lee et al. studied a safety campaign
conducted in Florida during 2009 that primarily focused on educating the public on
behaviors associated with aggressive driving (18). One of the driver-education
approaches used in the campaign was displaying campaign slogans on six roadside
billboards along a target corridor of Interstate-75 in Florida. In order to assess the
effectiveness of the safety campaign, surveys were distributed to truck drivers and
passenger car drivers.
The surveys were distributed during the following three time periods: prior to the
campaign, during the campaign, and after the campaign ended. The percentage of truck
drivers who reported encountering aggressive driving every day decreased throughout the
campaign from 95 percent before the campaign started, to 70 percent during the
23
campaign, and then down to 63 percent after the campaign ended (18). Accordingly, the
percentage of passenger car drivers who reported encountering aggressive driving every
day went from 53 percent to 43 percent to 56 percent over the same time periods (18).
These results show that the safety campaigns were able to produce the desired behavioral
change of reducing the amount of aggressive driving exhibited along the target corridor.
However, the lasting effects of the campaign were non-existent among passenger car
drivers since a larger percentage of drivers reported seeing aggressive driving after the
campaign had ended than prior to the start of the campaign.
Gantz, Fitzmaurice, and Yoo examined a seat belt safety campaign in Indiana that
was conducted in 1987, shortly after the mandatory seat belt law was passed in that state
(19). The safety campaign was composed of multiple media approaches, including
television, radio, and billboard advertisements. Survey questionnaires were administered
to 811 drivers across the state through telephone interviews (19). The questions in the
surveyed covered a range of topics such as perceived seat belt use, exposure to the seat
belt campaign, and awareness of the mandatory seat belt law. Based on the survey results,
53 percent of drivers recalled seeing campaign advertisements on television while only
13 percent recalled seeing campaign advertisements on billboards (19).
For those drivers who recalled seeing billboard advertisements, 54 percent
believed the advertisements had little to no impact, 17 percent said the advertisements
reinforced the need to wear a seat belt, and 17 percent said the advertisements caused
them to think more about seat belts (19). These percentages were comparable to those for
the other media approaches. Television commercials had 47 percent of drivers who
24
believed that the commercials had little to no impact, 13 percent said the commercials
reinforced the need to wear a seat belt, and 19 percent said the commercials caused them
to think more about seat belts (19). The results of this study indicate that while billboards
may not be the most effective approach for communicating safety campaigns to the
majority of the public, they are able to provide a similar impact on drivers when
compared to the other media approaches used in such campaigns.
Eby et al. examined a seat belt safety campaign conducted in Michigan during
2002 which combined a mass media campaign with an enforcement program (20). The
campaign was designed to have media and enforcement activities occur in an
experimental region while no specific activities were to occur in a control region. The
mass media campaign included television, radio, newspaper, and billboard
advertisements. The researchers conducted three separate direct observation studies to
determine the percentage of drivers using their safety belt. The studies were done for
three different periods: prior to the start of the campaign (Pre), immediately after the
campaign (Post 1), and about a month after the end of the campaign in order to study the
long-term effects (Post 2).
From the direct observation study, the overall safety belt use in the experimental
region was found to be 74.7 percent, 72.2 percent, and 77.9 percent across the Pre, Post 1,
and Post 2 periods, respectively, while the usage in the control region was 77.5 percent,
81.4 percent, and 82.3 percent across the same periods (20). Consequently, the
researchers concluded that the campaign did not increase safety belt use in the
experimental region. However, the safety belt use increased significantly in the
25
experimental region from Post 1 to Post 2 for drivers in the 16-29 age group and for
males. This could indicate that populations which are known to be a safety concern can
successfully have their behavior modified through a safety campaign, despite not being a
specific target audience for the campaign.
Along with the direct observation study, telephone surveys were conducted in the
experimental and control regions for approximately the same periods (Pre, Post 1, and
Post 2). Self-reported seat belt usage was approximately 90 percent for both regions
across all three periods (20). The percentage of drivers surveyed who reported being
exposed to the campaign on increased safety belt enforcement was significantly higher
for the Post 1 and 2 periods (38 percent and 26 percent) than for the Pre period (11
percent) in the experimental region (20). These levels were comparable to those of the
control region indicating that campaign activities were likely seen by drivers in the
control region despite no specific activities having occurred there. Thus, the use of a
control region was not successful and hinders the ability to make comparisons between
the regions. Within the experimental region, the drivers who reported being exposed to
the campaign from having seen television commercials was 56 percent while those
having seen billboards was 21 percent for the Post 1 period (20). These percentages are
similar to those reported by Gantz, Fitzmaurice, and Yoo for drivers in Indiana, hence
corroborating the potential exposure to the safety campaign from using billboards.
Phillips, Ulleberg, and Vaa examined the effect of road safety campaigns on
crashes based upon a meta-analysis conducted on the results of 67 studies which had
occurred between 1975 and 2007 (21). Meta-analysis is described by the researchers as a
26
method for summarizing the results of a group of similar studies which have common
evaluation criteria. The results of the meta-analysis show that road safety campaigns are
able to reduce crashes by nine percent (21). This suggests that, overall, safety campaigns
are able to produce at least a slight change in driver behavior, resulting in additional
safety benefits along the roadway. Furthermore, the results of the meta-analysis showed
that the use of roadside billboards in a safety campaign is linked to a greater reduction in
the number of crashes on the roadway. Phillips, Ulleberg, and Vaa conclude that roadside
billboards present an immediate delivery of the safety message to drivers which, in turn,
may result in drivers selecting to alter their driving behavior upon viewing the safety
message.
2.4 Safety of Billboards
2.4.1 Impact on Glance Behavior
The ability of billboard advertisements to reach drivers while they are on the
roadway can be beneficial for communicating a safety campaign message. However, the
use of billboards in this context requires drivers to look away from the roadway ahead in
order to see and comprehend the safety message. This can lead to a potential driver
distraction issue which may introduce other potential safety issues along the roadway.
Research into the safety effects of billboard advertisements has included studying the
glance behavior of drivers to determine if billboards are a significant distraction.
Research has shown that drivers have a natural limit for the length of time that
they are willing to look away from the roadway. In 1988, Rockwell mentioned a two
second rule that has been found by researchers in regards to driver eye movements away
27
from the roadway ahead (22). This limit on glance duration is included in the discussion
of in-vehicle glances recorded during three on-road studies conducted by Rockwell. The
two second rule was found to be consistent with the results of the three studies. The mean
glance durations towards the car radio were found to be statistically similar for all three
studies with glances averaging 1.4 seconds overall, whereas glances towards the rear
view mirror averaged 1.0 seconds and speedometer glances averaged 0.8 seconds (22).
Also in 1988, Zwahlen, Adams Jr., and Schwartz proposed a set of limits for
glance behavior when a driver is using the keypad on a cellular telephone to dial a phone
number (23). The number of glances towards the phone that were required to complete
the dialing task was related to the average glance duration for both a maximum and
acceptable limit. The maximum limits on average glance duration were proposed as two
seconds for a single glance with the limit decreasing by 0.2 seconds for each additional
glance made up until 1.4 seconds for four glances (23). Glances that exceeded these
maximum limits were considered to be unacceptable as they were determined to impact
driver safety. The acceptable limits on average glance duration were proposed as 1.2
seconds for one glance, 1.0 seconds for two glances, and 0.9 seconds for three glances
(23). The glances occurring between the acceptable and maximum limits were considered
to potentially impact driver safety.
In 1993, Wierwille proposed a limit of 1.6 seconds for which drivers would
normally be willing to look away from the roadway (24). Wierwille reviewed the results
of several studies on driver glance behavior and found the 1.6 second limit for a single
glance to apply to all of the mean glances reported in the studies. In 2007, Horrey and
28
Wickens used the 1.6 second limit in the evaluation of glances made towards a LCD
screen mounted on the dashboard of a driving simulator (25). The LCD screen was
intended to represent the in-vehicle technologies present in modern vehicles. Participants
were given the task of reading a series of numbers to determine whether there were more
even or odd numbers while driving in the simulated environment. The task was classified
as either simple or complex depending on the amount of numbers presented to the driver
(five or 11). Additionally, drivers were presented with a hazard condition where either a
low or high wind gust hit the side of the car.
The mean glance duration towards the screen was below the 1.6 second limit for
each of the four conditions studied. However, the complex task was found to have
significantly more long glances as 21 percent of the total glances recorded exceeded the
1.6 second limit compared to 6 percent for the simple task (25). Overall, the driver’s
ability to respond to the hazard condition was found to be negatively impacted by
increased glance duration towards the LCD screen. As a result, the researchers developed
a model to estimate the crash risk based on the percentage of glances longer than 1.6
seconds and certain aspects of the driving situation, including speed of the vehicle.
These previous studies have focused primarily on in-vehicle glances while other
studies have focused on those glances made outside of the vehicle. In 2006, Klauer et al.
used the data collected during a naturalistic driving study to determine the crash risk
when drivers looked away from the forward roadway (26). The researchers analyzed
glance behavior during the period prior to the precipitating factor of a crash or near-crash
where the subject was the at-fault driver. The analysis began five seconds before the
29
precipitating factor and ended one second afterwards. Glances away from the roadway
during this time period was found to be a contributing factor for 60 percent of the
crashes, near-crashes, and incidents that occurred in the study (26).
Odds ratios were calculated for a variety of conditions when the driver was
looking away from the forward roadway. Odds ratios describe the likelihood of an event,
such as a crash or near-crash, occurring and would be 1.0 for normal conditions with
greater values indicating an increased crash risk. The odds ratio for all glances away from
the forward roadway that were greater than two seconds was calculated to be 2.19 (26).
However, those glances made towards the side and rearview mirrors or to the sides of the
roadway environment as part of the driver scanning for potential hazards merit separate
consideration from other glances away from the roadway as these behaviors are required
as a part of normal driving. For these types of long glances that exceed 2.0 seconds the
odds ratio was calculated to be 0.45 indicating that these behaviors increase safety (26).
Similarly, when these long glances related to the driving task are removed from the data
set for long glances away from the forward roadway the odds ratios increases to 2.27
(26). Moreover, the length of time and frequency of glances away from the roadway were
found to be significantly higher for crashes than near-crashes. Thus, as the driver’s
attention is averted from the roadway for instances not related to the driving task at-hand,
the safety of the driver is negatively impacted.
Based on past research, there is an apparent acceptable limit of two seconds for
driver glances. This limit applies to glances made both inside and outside of the vehicle.
Additionally, these long glances are shown to have an impact on a driver’s ability to
30
safely operate their vehicle as seen by the increased crash risk. Thus, when assessing the
impact of glance behavior on safety, the two second limit can be a useful guideline.
In 2002, Sodhi, Reimer, and Llamazares studied the duration of glances away
from the forward roadway during an on-road study that took place over a 22 mile route
(27). Drivers were instructed to complete a series of tasks while a head-mounted eyetracking system recorded their glance behavior. The eye-tracking data was only analyzed
for five participants for the task requiring drivers to change the radio to a certain station,
and for the task in which participants were asked to look in the rearview mirror and
describe the vehicle present. Based upon the analysis, two of the 113 glances made
towards the radio and two of the 95 glances made towards the rearview mirror exceeded
the 1.6 second limit proposed by Wierwille (27). The average of the longest glances made
in the radio and rearview mirror tasks were 1.56 and 1.63 seconds, respectively (27).
These results support the proposed limits that drivers will naturally set for themselves,
but is based on a very small sample of drivers.
Beijer completed an on-road study examining the glance behavior of drivers
towards roadside advertising as they drove along an expressway in Toronto during 2002
(28). As participants drove along the expressway, they passed 37 different advertising
signs which were categorized into the following four sign types: billboard, scrolling text,
video image, and roller bar (28). The billboard signs were static advertisements whereas
the other three sign types had some form of active component included in the sign.
Throughout the experimental drive an eye-tracking system was used to collect the
participant’s eye movements and a scene camera was used to collect the visual scene as it
31
was presented to the driver. The videos from the eye-tracking system were then analyzed
to determine various glance behavior parameters such as average glance duration and the
number of glances per advertising sign.
The study showed no significant difference in the average glance duration
between the four sign types, but the number of glances per sign was significantly
different between the billboard signs and the three active component signs. Beijer found
that there were 0.64 glances per subject per sign towards billboard signs, yet there were
1.32, 1.31, and 1.45 glances per subject per sign towards the roller bar, scrolling text, and
video signs, respectively (28). Additionally, Beijer analyzed long glances, or those
glances longer than 0.75 seconds, to determine if there was a particular sign type which
participants tended to fixate on for longer time periods. It was determined that 88 percent
of participants made at least one long glance towards an advertising sign, with billboards
and scrolling text signs receiving significantly less long glances per sign than the roller
bar and video signs (28). These results imply that advertising signs with active
components may cause drivers to take longer and more frequent glances towards these
signs. Thus, active component signs could be considered a potentially greater distraction
for drivers as they divert their attention from the roadway ahead for longer time periods,
which could result in unsafe driving conditions.
Smiley, Smahel, and Eizenmann further examined glance behavior towards
roadside video advertising signs during an on-road study in Toronto during 2002 (29).
Drivers passed signs which were located near three intersections in the downtown area
and along an urban expressway while an eye-tracking system recorded their glance
32
behavior. Once the eye-tracking data was analyzed, there were a total of 69 intersection
approaches (33 were by video signs) and 14 expressway approaches. Drivers made at
least one glance towards a video sign during 48 percent of the intersection approaches
and 36 percent of the expressway approaches (29). The average glance duration towards
video signs was determined to be 0.48 seconds while the longest glance was 1.47 seconds
(29). This indicates that the video signs do not pose a significant safety risk with regard
to the amount of time that drivers look away from the roadway.
In 2011, Perez and Bertola conducted an on-road study in which participants
drove two 30 minute routes past locations with and without advertising billboards and
recorded the driver’s glance behavior with an eye-tracking system (30). Data was
collected in four data collection zones where billboards were present. The location of the
billboard (left or right side of the roadway) and the level of clutter (high or low) in the
visual scene were varied to create the four zones. Additionally, there were four data
collection zones which did not have billboards but featured high and low levels of visual
clutter for two zones apiece. The researchers selected the data collection zones to begin
960 feet prior to the billboards based on the Manual on Uniform Traffic Control Devices
(MUTCD) legibility guidelines. The glance behavior during these zones were analyzed
by first examining where the driver was looking (roadway ahead, billboard, etc.) and
noting the glance durations for each object. The proportion of glances were determined
by comparing the total glance durations per object to the total amount of time driving
through the data collection zone.
33
For the zones with billboards present, drivers were seen to have looked less
frequently at the roadway ahead as the mean proportion of glances was 0.82 in
comparison to 0.85 for the zones without billboards (30). However, the mean proportion
of glances towards billboards was 0.02 in areas with high clutter and 0.04 in areas with
low clutter (30). Furthermore, the mean glance duration towards billboards was 0.04
seconds for the entire study (30). This indicates that the billboards evaluated in the study
were not very distracting as they were glanced at for very short durations and for
approximately 3 percent of the entire time participants were driving in the study.
Lee, McElheny, and Gibbons conducted a naturalistic driving study in Cleveland,
Ohio during 2006 where participants drove past five digital billboards, 15 conventional
billboards, 12 comparison sites, and 12 baseline sites (31). The comparison sites had
visual elements that drivers would see on a daily basis, such as commercial signs, while
the baseline sites had no signs present. The participant’s glance behavior was recorded
using an eye-tracking system and the data was analyzed for an eight second period which
ended when the participant passed the object of interest. Data was collected for 36
participants on a 50-mile route during daytime conditions and then for 12 of those same
participants on a 48-mile route during nighttime conditions (31). Four sites were not
included in the abbreviated route used for the nighttime conditions.
Based on the analysis of the glance behavior for daytime conditions, there were
no significant differences between the four types of sites for both the proportion of
glances to the roadway ahead and the number of glances made towards the side of the
roadway with the sign. However, the mean glance duration was found to be significantly
34
longer for digital billboards (0.87 seconds) than for conventional billboards (0.73
seconds) (31). There were no differences in the amount of long glances made towards
each of the four sites when using the 1.6 second limit proposed by Wierwille. For the
glance behavior during nighttime conditions, there were trends seen across the glance
behavior measures which indicated that digital billboards may be more of a distraction
than conventional billboards. However, no statistical analyses were done due to the small
sample size of data collected during nighttime conditions.
The location of the billboard may influence the driver’s glance behavior. In 2005,
Crundall, Loon, and Underwood compared the distraction capabilities of street-level
advertisements (SLAs) and raised-level advertisements (RLAs) through the glance
behavior of participants as they viewed multiple video clips (32). As participants viewed
the video clips, an eye-tracking system was used to record their eye movements. From the
eye-tracking data, it was determined that the SLAs had 2.1 fixations per advertisement
and RLAs had 1.5 fixations per advertisement (32). Similarly, SLAs had mean fixation
durations of 0.433 seconds and RLAs had mean fixation durations of 0.354 seconds (32).
These results suggest that drivers may look for longer periods and more frequently at
advertisements that are located lower and closer to their normal field of view. This seems
logical as drivers would not have to deviate their gaze very far from the roadway in order
to see these advertisements.
Content of the billboard advertisements may be another factor which influences
glance behavior. In 2011, Megias et al. conducted a motorcycle driving simulator study
where drivers passed static billboards located above the roadway which had either
35
positive, negative, or neutral emotional content (33). The emotional content of the
pictures used on the billboards were rated by a standard produced by the International
Affective Picture System. Examples of the types of images used in the study are:
romantic scenes for positive content, vehicles from crashes for negative content, and
inanimate objects for neutral content. Eight images for each type of content were used in
the study in combination with six hazardous and six non-hazardous situations. The
hazardous situation was created by having an entity unexpectedly enter the roadway in
front of the subject. The scenarios for the study were 7.00 seconds in length with the
billboards visible for the first 4.80 seconds of the scenario while the entity entered the
roadway 0.80 seconds after the billboard was not visible (33).
The total fixation time on the billboards was significantly different between the
different types of content as positive images were looked at for 1.76 seconds, negative
images for 1.91 seconds, and neutral images for 1.60 seconds (33). The average glance
durations were the same for the positive and negative images (0.33 seconds) with the
neutral images being significantly higher (0.35 seconds) (33). Similar to the total fixation
times, the total number of fixations was significantly different as negative images had the
most fixations and neutral images had the least. These results indicate that the content of
a billboard can impact the glance behavior of drivers and should be considered in the
development of billboard advertisements.
Consideration of glance behavior differences between drivers needs to be
included in a potential safety evaluation. In 2011, Owens et al. tested novice and
experienced drivers on their ability to assess the risk presented in a driving environment
36
by viewing several photographs (34). Participants were shown two photographs of actual
driving environments and instructed to decide which of the scenes posed the greater risk
to the driver. As soon as a participant had decided which scene was riskier, they were
instructed to enter their response on a keyboard. The system used to run the test
controlled the order in which the 26 pairs of test scenes were displayed, recorded the time
it took for the participant to enter their response after the image was shown, and collected
data on the participant’s eye movements. The pairs of driving scenes were setup in an
either an “Easy” group where there were large differences in the risk presented by the
scenes or a “Difficult” group where the difference was smaller (34). As expected, the
response times for the “Difficult” group were significantly longer than those for the
“Easy” group. The analysis of the glance behaviors showed that experienced drivers
fixated on the riskier roadway scene to a significantly greater degree than novice drivers
(34). This supports the idea that as drivers gain more experience, they become more
adept at assessing the potential risk on the roadway. Consequently, younger drivers may
be more at risk when glancing towards distracting objects as they may misinterpret the
risk present on the roadway and decide to look away during these times.
As these studies show, there is a need to evaluate a driver‘s glance behavior in
considering the safety of billboard advertisements. If the billboard is too distracting from
a visual perspective than the use of that sign in a safety campaign would be selfdefeating. Similarly, the sign must be read in order to obtain the safety message, so a
balance on the glance duration must be found. Ultimately, the impact on a driver’s
performance in safely operating the vehicle while looking at billboards is unclear for
37
most of these studies reviewed in this section and needs to be considered alongside the
impact on glance behavior.
2.4.2 Impact on Driving Performance
Research into the safety of billboards has also included analyzing crash data to
see if billboards were associated with an increase in crashes. Tantala and Tantala
analyzed the crash data at 20 sites with digital billboards and 57 sites with regular
billboards located in the greater Reading area in Pennsylvania (35). Crash data for the
areas near the billboard sites was taken over a nine year period (2001 to 2009) which
included data from before the digital billboards replaced the regular billboards. Also
traffic volume data for the same time periods was used to determine the crash rates based
on the annual average daily traffic (AADT). This was done to factor in how often crashes
occurred based on the average amount of traffic on the roadways. For the 20 digital
billboard sites, the number of crashes that occurred within a half-mile of the site was
examined for equivalent periods prior to and after the billboard conversion. The total
number of crashes was found to decrease by 13.0 percent from the before to after periods
while the crash rate per million AADT also dropped from 1.52 to 1.35 (35). However,
these reductions in crashes were not statistically significant at a 95 percent level of
confidence.
Tantala and Tantala also conducted an Empirical Bayes method, as specified by
the 2010 Highway Safety Manual, to predict the number of crashes that would have
occurred at the digital billboard sites if the signs had not been converted (35). The
Empirical Bayes method used the crash data from the 57 regular billboard sites when
38
developing the predicted number of crashes. The results of the Empirical Bayes method
were 917 predicted crashes if the conversion did not occur in comparison to 925 actual
crashes, a difference that is not statistically significant (35). Consequently, the analysis of
the crash data shows no effect on crashes from digital billboards in comparison to regular
billboards.
In addition to the glance behavior discussed previously, Smiley et al. studied
crashes and driving performance near video signs in Toronto during 2002 (36). Crash
data was obtained for a period of approximately seven years which spanned the
construction of the video signs at the three intersections and along the expressway.
Empirical Bayes methodology was used to account for traffic volumes at the
intersections, but was not used for the expressway segment. The total crashes on the
video approaches at intersections were not significantly different for the before and after
periods as the crashes increased by 0.6 percent (36). For the expressway video approach,
the number of crashes across the before and after periods was not significantly different
with a comparable non-video sign expressway segment. Combined with the results found
by Tantala and Tantala, billboards appear to not have an impact on the number of crashes
occurring at the locations where the signs are visible.
Driving performance on the expressway was studied by Smiley et al. through use
of traffic detector stations on the travel lanes approaching the video sign and on the travel
lanes directly opposite for which the sign was not visible (36). The speed and headways
for the vehicles traveling through this section were evaluated for periods before and after
the installation of the video sign. The data showed that the average speed decreased over
39
the video approach, but there were increases in the standard deviation of speed and a
decrease in headways. As a result, the researchers argue that safety on the video approach
was improved through the decrease in speed, yet degraded as headways decreased and
speed variance increased (36).
From the previous section on glance behavior, a few of the studies discussed did
examine driving performance in addition to glance behavior. Lee, McElheny, and
Gibbons also collected data on the driving performance of the participants as they drove
past the four types of sites during daytime conditions (31). Across all road types,
conventional billboards had a significantly greater standard deviation of speed than
digital billboards; though this difference was not present between the two sign types
when the participant was driving on the interstate. The standard deviation of lane position
was higher for digital and conventional billboards than the comparison and baseline sites,
but this difference was not statistically significant. These results indicate that the type of
billboard does not have an effect on driving performance. Conversely, for the motorcycle
driving simulator study, Megias et al. found the braking responses to the hazardous event
to be significantly faster for the negative images in comparison to the positive and neutral
images (33). The content of billboard advertisements may impact driving performance;
however, this evaluation was made during scenarios which were very short in duration.
In 2011, Edquist et al. conducted a driving simulator study examining the impact
of billboards on driving performance (37). The study consisted of participants completing
a lane change task as instructed by a traffic sign located on the side of the road. This lane
change task was conducted at the following three locations: those with static billboards,
40
those with changeable billboards, and those with no billboards present. The driving
simulator was also equipped with an eye-tracking system which was used to collect data
on the participant’s eye movements. The study showed that participants took significantly
longer to change lanes at locations with static or changeable billboards than at locations
with no billboards. Furthermore, the results of the eye-tracking data showed that for
almost all of the participants, drivers spent less time looking at the roadway ahead when
driving past a location with a billboard present than when for locations without a
billboard. The results of this study suggest that drivers may take longer to complete
simple driving actions, such as completing a lane change, due to the presence of billboard
advertisements. This may occur because drivers are not devoting their full attention to the
driving task at hand and the roadway ahead, but instead are becoming distracted by an
outside influence.
In 2009, Young et al. conducted a driving simulator study to determine the effect
of billboards on driving performance and glance behavior for three different roadway
environments: urban, rural, and highway (38). Participants in the study drove each of the
three environments with and without billboards located along the roadway. The primary
driving performance measures evaluated were time to contact and the lateral control of
the vehicle. The time to contact was measured relative to the vehicle ahead of the
participant. Lateral control of the vehicle was evaluated as the number of times the
vehicle left the travel lane and how long the vehicle was out of the lane. Time to contact
was determined to not be significantly different for the presence of billboards, but the
lateral control measures were both significantly different. While driving past billboards,
41
participants drove outside of the travel lane more frequently and for longer time periods.
Furthermore, an eye-tracking system was used to collect data on the duration and
frequency of glances towards billboards. Participants were found to look towards the side
of the roadway significantly more when passing billboards, but the duration of these
glances were not affected by the presence of billboards.
In 2007, Jamson and Merat completed a driving simulator study which examined
the effectiveness of displaying safety campaign messages on variable message signs
(VMS) (39). Participants drove past a total of 24 VMS which displayed either two
slogans, “WATCH YOUR SPEED” or “KEEP YOUR DISTANCE.” The number of
signs which were displaying a safety message was varied across four groups with no
signs on for a baseline condition and either 33 percent, 66 percent, or 100 percent of the
VMS displaying one of the two slogans. The two slogans were used equally throughout
each of the three groups with active signs. After passing the 24 VMS used for the safety
messages, a final sign was presented with a Tactical Incident Management (TIM)
message informing drivers of an incident ahead and instructing them to drive in a specific
lane. This sign was setup in order to require the driver to complete a lane change
maneuver. The length of time needed to complete the lane change was recorded along
with the speed and headway of the vehicle.
The participant’s eye movements were recorded using an eye-tracking system.
The researchers were primarily interested in the eye movements as the participants
approached the VMS. This resulted in using a section of 820 feet prior to the sign as this
was the area over which the sign was determined to be legible. The upstream section was
42
denoted as the 1640 feet prior to the legibility section while the downstream section was
1640 feet after the VMS. The change in speed and headway were each calculated as the
difference between the mean values for the upstream and downstream sections. There
was no significant difference found between the active and inactive signs for the change
in either mean speed or mean headway. The length of time needed to complete the
required lane change was found to be significantly shorter for the 33 percent of VMS
active condition in comparison to the other conditions. However, for the same condition
the gazes towards the final VMS were significantly longer in duration than for the other
conditions. Over the course of the scenarios, it was seen that drivers continued to glance
at the signs with safety messages which helped them be able to respond sooner to the
TIM message (39). Thus, despite the safety messages not having a direct effect on the
intended driving performance measures, there was an improvement in the driving
performance required by the warning message.
Thus, it is apparent that billboards can be effective at communicating a safety
campaign to drivers; however, they can also serve as a significant distraction from the
driving task at hand. The amount of time required for a driver to look towards a billboard
and comprehend the safety message should be balanced with the amount of time in which
they can safely look away from the roadway ahead. Based on the research reviewed, it
can be seen that the impact on driver behavior from using billboards to deliver a work
zone safety message is unknown at this time. This study aims to examine the effects of
campaign billboards on driver behavior in work zones and to determine the most
effective content and placement of billboard advertisements from a safety perspective.
43
CHAPTER 3: PUBLIC SURVEY STUDY
The first phase of the study consisted of a public opinion survey designed to
determine the background and text color combination that was preferred by the public for
use in the billboard advertisement. Previously, Laborers’ Local 860 used the billboard
advertisement, shown in Figure 1 below. This sign was found to be ineffective at
communicating the safety message after conducting an unofficial survey of drivers who
had passed the sign. Researchers initially focused on assessing which elements of the
advertisement were potentially contributing to the sign not being seen by drivers. The
combination of a black background with orange text was believed to be difficult for
drivers to see when traveling at high speeds on a busy highway. Also the advertisement
was believed to be too cluttered for drivers to easily comprehend the safety message as
multiple lines of text with varying fonts were used alongside two graphics.
Figure 1. Billboard advertisement used in Cleveland.
44
In order to assess whether the black and orange color combination used in the
advertisement was considered ineffective by drivers, the advertisement was reproduced in
several color combinations which are listed in Table 1 below. The black background with
orange text used in the original sign was included along with seven new combinations.
The background and text colors used in the reproduced advertisements were selected
from those colors specified for use on traffic signs by the National Manual on Uniform
Traffic Control Devices (MUTCD) (40).
Table 1: Background and text color combinations used in surveys
Set of Billboards
Background Color
Text Color
Black
Orange
Green
White
Set 1
Orange
Black
Yellow
Black
Fluorescent Yellow-Green Black
Red
White
Set 2
Light Blue
Black
Red
Black
There are 13 background colors identified in the National MUTCD for use on
traffic signs (40). Each of the colors is associated with conveyance of specific
information to drivers. Since the purpose of the billboards is to communicate a safety
campaign message to drivers, it was considered to be important to remain consistent with
the manner in which traffic signs are used. This approach builds upon the knowledge
already obtained by drivers through their years of experience and provides an opportunity
to build upon that driver expectancy.
45
The black background used in the previous advertisement is associated with
traffic signs that communicate regulations to drivers. This color would be good for the
billboard as it would reinforce the regulation of reduced speed limits that are typically
used in work zones. Red is used in traffic signs with prohibitive regulations, primarily for
stop signs. Using red would communicate to drivers that exceeding the speed limit is a
behavior that should be stopped since it is a traffic offense. Both yellow and fluorescent
yellow-green are the background colors used for warning signs, with fluorescent yellowgreen specifically for school, pedestrian, and bicycle warning signs. Each of these colors
is consistent with the slogans used in the advertisement which are warning drivers to be
more cautious in work zones by reducing their speed. Orange is the background color for
temporary traffic control signs which are used to guide drivers safely through work
zones. Thus, using an orange background for the billboard advertisement would build
upon the connotation with work zones and be consistent with helping to guide drivers
through the work zone.
Green and blue are the background colors which are used for traffic signs that
provide some type of guidance or information to drivers. Green is used for directional
guidance and to provide recreational information. Blue is used to inform drivers of
services available, provide tourist information, or denote an evacuation route. Thus, the
associations for both green and blue are fairly consistent with the intent of the
advertisement to offer guidance to drivers. However, the meanings for the two colors do
not line up as closely as the other background colors discussed.
46
For the background colors selected from the National MUTCD for the reproduced
billboard advertisements, it was determined that most of the colors were darker shades.
For the surveys, a balanced mix of dark and light colors was considered to likely be more
beneficial in determining the color combination preferred by drivers. Based on which
colors were shown to the people taking the surveys, a general trend towards a lighter
shade may be due to the other shades being darker and somewhat similar to one another,
or vice versa. In this way a balanced approach could help to eliminate any bias towards a
lighter or darker color based on the colors presented. Thus, light blue was selected to
replace blue as the background color for the reproduced advertisements. Light blue is
currently an unassigned color for traffic signs, but would likely be recognized by drivers
as having a similar meaning to that of blue signs.
Five background colors specified for traffic signs were not selected as they were
thought to be inappropriate for use in the reproduced advertisements for a number of
reasons. Brown is used for providing drivers guidance to recreational areas and cultural
interests. While guidance is being offered to drivers through the billboard, the type of
guidance intended by the color is considered too dissimilar to be used for the
advertisement. Conversely, a white background color is used for regulatory signs which
would have a similar meaning to that intended by the billboards. However, white was not
selected for use in the reproduced advertisements since it was considered not to be a good
option for attracting a driver’s attention. Both brown and white were thought to be too
likely to blend in easily with the surrounding environment and, subsequently, not be seen
by drivers traveling at highway speeds.
47
The background colors of purple, fluorescent pink, and coral were not selected as
their meanings do not line up with the intended purpose of the billboard. Purple is the
background color used on signs to indicate which lanes are for electronic toll collection
accounts only at a collection point on a toll highway. Fluorescent pink is for incident
management signs which are used to guide drivers along a certain route in instances
where a traffic incident has occurred. Traffic incidents are specified in the MUTCD to
include emergency road user occurrences such as a hazardous material spills, natural
disasters, or any unplanned event which impacts the normal flow of traffic (40). These
emergency situations are not similar to the message of the safety campaign, so the
association should be avoided. Finally, coral is a background color that is unassigned at
this point and should not be used in the advertisement in order to avoid potential driver
confusion.
Based on the background color of the traffic sign, the MUTCD also specifies the
colors which may be used in the legend (the text and symbols) of the sign (40).
Accordingly, the text colors for the billboard advertisements were selected to be either
black or white depending on the background color used. The lone exception was the
black background with orange text used in the original advertisement. This color
combination is listed in the MUTCD as an option for changeable message signs. The
original color combination was primarily maintained in the reproduced advertisements so
that it could be compared to the new color combinations developed.
For the orange, yellow, and fluorescent yellow-green backgrounds, black text was
selected as it was the only legend color specified for each. Likewise, white text was
48
selected as it was the only option for use with a green background. The light blue
background did not have a legend color specified since it is currently unassigned, so
black text was selected for the advertisement since it better contrasted with the lighter
shade. Regulatory signs with a red background can have either black or white for the
legend color. As a result, both colors were selected to be used for the text in the
reproduced advertisements. This would also provide an evaluation of which text color
might be preferred given a constant background.
There were only two variations to the reproduced advertisements other than the
different background and text color combinations used. The first was to change the
slogan from “Caution in Work Zones” to “Slow Down in Work Zones.” The new slogan
was selected since it was a more direct version of the message being communicated to
drivers. The phrase “Slow Down” plainly communicates to drivers an action they should
take to improve safety whereas “Caution” leaves room for the interpretation as to how
drivers should be more cautious through work zones. The second variation was to display
the “Slow Down in Work Zones” slogan in the same font as the “Road Work Season”
slogan. The slogans were displayed in the same font in order to remain consistent within
the advertisement in regards to how the message was communicated to drivers.
The reproduced advertisements were separated into two groups so that the
participants would only be required to evaluate four billboards when taking the survey.
Presenting the participants with eight billboards was thought to provide them with
potentially too many options to consider at one time and may not produce a preferred
option. Also, dividing the advertisements into two groups provided the ability to evaluate
49
certain color combinations directly. For example, grouping the two red backgrounds with
differing text colors allowed for an assessment on how text color affected public
preference. Similarly, grouping the orange text on a black background and the black text
on an orange background allowed for a direct assessment on how public preference is
affected by reversing the colors in a particular combination.
Survey questionnaires were developed to collect opinions from the public about
the reproduced billboard advertisements. The surveys consisted of thirteen questions
which ranged from information about the driver to their opinions on the various
advertisements. Two demographics questions were first used to assess whether or not the
surveyed population matched the overall population. The next six questions addressed the
participant’s driving experience and their knowledge of the rules and regulations for work
zones. For the next four questions, participants were asked to select which of the four
billboards in the set they were shown best met the following conditions: made them more
aware of work zones, made them more cautious in work zones, would be seen the best
during daytime conditions when driving on the freeway at 65 mph, and would be seen the
best during nighttime conditions when driving on the freeway at 65 mph. The final
question in the survey asked if the sign was too cluttered with information.
Survey responses were solicited from volunteers across three cities in Ohio:
Athens, Chillicothe, and Cleveland. Distribution of the surveys was done outside of high
-traffic locations, such as malls and stores, in each of the cities. Participants were handed
a copy of the survey to fill out and were shown either one of the two sets of billboard
advertisements. Copies of the questionnaire and the sets of billboard advertisements that
50
were shown to the participants are included in Appendix A. The responses obtained from
the survey were compiled into a database and evaluated to determine the preferred
billboard color combination for each of the four different conditions presented in the
survey. The responses to the demographics and driving experience questions were used to
evaluate which color combination was preferred among the different groups (i.e. age
groups, gender, etc.) for the different conditions.
The resulting color combination from each of the conditions and groups evaluated
were then compared to make a determination on the color combination which would be
most effective overall. Statistical tests were done on the responses to the survey questions
to determine if there were statistically significant differences between the proportions of
responses for each color combination. Since there were two sets of billboard
advertisements used in the survey, the proportion of responses for each color combination
were used in order to make comparisons between the two sets. The null hypothesis for the
statistical tests was that there is no difference between the billboards that were selected.
The sample size for the public opinion survey was selected based upon the effect
size that would be determined. Effect size is commonly used in statistical analysis as a
means of defining the differences that are found to exist between the treatments that were
used in a study. According to Field, effect size is an objective and standardized measure
of the magnitude of the observed effect (41). This allows the effect of a treatment to be
equated across different variables which are used in a study. Thus, for the public opinion
surveys the effect of the color combinations on the four conditions in the survey can be
evaluated in a similar manner for both sets of billboard advertisements.
51
Several measures of effect size have been proposed by statisticians including the
specification of the difference that would be detected in the study. This detectable
difference is often presented as a proportion of the sample standard deviation (σ) with
0.25σ for a small effect, 0.50σ for a medium effect, and 1.00σ for a large effect (42). For
the public opinion survey, it was determined that a medium or large effect size would be
preferred for the evaluation of the color combinations. Furthermore, it was selected that
for the entire study, the statistical tests would be conducted at a 95 percent level of
confidence (α = 0.05) and the statistical power of a test to detect an effect would be 80
percent (β = 0.20), based upon a 4:1 ratio of α to β.
From these parameters the sample size per treatment level for the medium and
large effect sizes could be determined for the entire study evaluating the eight billboard
advertisements and for both sets of four billboards. From Hinkle, Wiersma, and Jurs, the
entire study would require 936 participants to detect a medium effect size and 240
participants to detect a large effect size (42). For each set of four billboard
advertisements, 352 and 92 participants would be needed to detect a medium and large
effect size, respectively (42). Due to time constraints, it was determined that the sample
size for a large effect size would be used and the minimum number of survey responses
that would be collected was set at each of the values listed above. The total number of
survey responses collected was 376, with 156 participants completing surveys for Set 1
and 220 participants completing surveys for Set 2.
52
CHAPTER 4: DRIVING SIMULATOR STUDY
4.1 Selection of Methodology
The second phase of the study used the Ohio Research Institute for Transportation
and the Environment’s (ORITE) Safety and Human Factors Facility’s state-of-the-art
driving simulator. As seen in the Background chapter, the primary methods used to
evaluate the impact of billboards on driver behavior were a driving simulator or an onroad instrumented vehicle in a naturalistic driving environment. These are the same
approaches that were recommended by Molino et al. for studies on electronic billboards
that are used for advertising (43). Of the previous research reviewed in the Background
chapter, there were five studies which evaluated both driving performance and glance
behavior. Lee, McElheny and Gibbons was the only naturalistic driving study examining
both aspects of driver behavior while the other studies reviewed (Megias et al., Edquist et
al., Young et al., and Jamson and Merat) were all conducted in a driving simulator.
The driving simulator provides the ability to evaluate driver behavior in a
controlled environment. One of the main benefits of this type of study is that researchers
have full control over the conditions presented to the driver. This helps limit the number
of factors which may potentially affect the different variables being studied. For example,
the presence of traffic on the roadway may affect the speed at which drivers select to
travel. Vehicles that are driving ahead of the subject could cause the subject to drive at a
slower speed than they would if there was no traffic present on the roadway. The driving
behavior exhibited during naturalistic driving studies can be affected by these types of
factors which may produce potentially confounding effects. However, driving simulator
53
studies are only a representation of a real-world driving environment and the driver
behavior exhibited may not directly correlate to that for an actual driving environment.
For example, the speed at which a driver travels in the simulator may not be the exact
same speed they may travel on actual roadways; thus, illustrating a primary advantage of
conducting a naturalistic driving study.
Yet, trends seen in the driving behavior exhibited during a simulator study can be
indications of how drivers would behave on actual roadways despite the behavior not
being exactly the same. Additionally, driving simulator studies protect the safety of
drivers during the study since they would not be exposed to the risk of physical harm
associated with a naturalistic driving study. Furthermore, the logistical issues to conduct
each of the two methodologies influenced the decision on which evaluation tool to utilize.
Based upon these considerations, a driving simulator study was selected as the best
methodology for this study since drivers could be exposed to multiple billboards in a
controlled environment within a short period of time.
4.2 Research Equipment Used
The state-of-the art driving simulator located in the ORITE Safety and Human
Factors Facility is a Drive Safety DS-600c Research Simulator and is shown in Figure 2.
The simulator cab is composed of the front half of a Ford Focus which features all of the
standard controls found in a typical passenger car, such as an accelerator, a brake pedal, a
steering wheel, two side mirrors, and a rear-view mirror. Three panels wrap around the
front of the cab to provide a 180 degree field of view, so that the driver is fully-immersed
within the simulated environment. Projectors display the simulated world on the panels in
54
front of the cab while the side and rearview mirrors are small screens which display the
simulated world behind the vehicle. The cab is mounted on a fixed motion base which
provides a total of five degrees of pitch and five inches of longitudinal motion to mimic
the starting and stopping of a vehicle. The simulator collects a wide array of driving
performance data which can be specified by the researchers during the design of the
scenarios. Data can be collected at rates up to 60 Hz, hence providing the ability to
thoroughly analyze a participant’s driving performance.
Figure 2. Driving simulator located in ORITE Safety and Human Factors Facility.
55
The driving simulator is also equipped with a faceLab 5 eye-tracking system from
Seeing Machines, as seen in Figure 2, which can be used to record the driver’s eye
movements throughout the study. The eye-tracking system consists of two cameras and
an infrared light pod installed on a bracket mounted to the dashboard in front of the
driver’s seat. Software included in the system is used to create a three-dimensional model
of the participant’s face in order to track their eye movements. As a result, there is no
video recorded of the participants and the eye-tracking system does not contact the
participant in any way which, unlike some head-mounted systems, allows them freedom
to behave as they normally would. Finally, the eye-tracking system collects data at a rate
of 60 Hz, thus providing the ability to thoroughly analyze a participant’s glance behavior
in correlation with the driving performance data collected in the simulator.
4.3 Institutional Review Board Process
Since human subjects were needed for their participation in the driving simulator,
the study had to gain approval from the university’s Institutional Review Board (IRB)
before the research could commence. The IRB process is designed to protect the rights of
human subjects in regards to personal privacy and minimize the risk that they may incur.
The process states that the research should maximize the amount of benefits that would
result from conducting the study and the researchers must be able to show that those
benefits justify the use of human subjects. Participants must also be selected in a fair
manner, so as to not unnecessarily target any particular groups. Prior to submitting the
56
study to the IRB for approval, the researchers completed a research ethics course which
tested these issues.
In order to submit the study to the IRB for approval, the researchers completed an
IRB Project Outline Form. This form described the structure of the study and detailed the
steps which were taken to protect the participants. The study was described in regards to
the objectives, methodology, and the risks and benefits associated with the study. Along
with describing the risks to participants, researchers described the procedures that were
taken in order to minimize the risks. Also the compensation for those individuals
participating in the study was covered in the submittal. For this study, it was determined
to use the Psychology Department’s pool of research participants for recruitment and
those individuals received research credit compensation. However, for other participants
in the study there was no compensation offered.
The submittal also required the researchers to provide copies of the consent forms
and questionnaires which were used in the study. The informed consent process included
communicating the voluntary nature of the study and disclaiming the possible risk of
incurring simulator sickness, often characterized by headaches and/or nausea.
Participants were informed of these issues prior to their involvement in the study. Copies
of the informed consent forms were included in the submittal to the IRB. The study was
first approved by the IRB in January 2011 for a period of one year. In December 2011, an
IRB Periodic Review Form was submitted and approved for another year. All of the
forms that were submitted to the IRB for this study and the approval documents are
included in Appendix B.
57
4.4 Design of Scenarios
The simulator study was designed to evaluate three specific aspects of the
billboard advertisements to determine which of those elements were effective at
communicating the safety message. The aspects of the billboards that were evaluated
were the slogans and graphics used in the advertisement and the placement and
orientation of the billboard with respect to the driver. Each of these three elements was
evaluated in individual scenarios that were specifically designed to test that particular
element. The driver behavior displayed as participants drove past the billboards was used
to evaluate the effectiveness of the billboards. If participants displayed safer driving
behaviors, such as driving slower through work zones or maintaining better lane
placement, after having seen a billboard then the particular sign was considered to be
effective at communicating the safety message.
For all three scenarios, participants drove on an urban four-lane undivided
highway which was designed similar to Interstate-90 in Cleveland where the original
billboard advertisements were located. No traffic was present on the highway during the
three scenarios in order to avoid having the participant’s behavior impacted by other
vehicles. While lack of traffic present on the roadway could potentially produce
differences in the observed driver behavior, this condition was considered the best
assessment of the billboards’ true impact on driver behavior. Since the only portions of
the scenarios which varied were the billboard elements being studied, any changes in
driving behavior exhibited by the participants can be associated with the billboards
presented. However, by observing driver behavior in this manner, the emphasis would be
58
on examining directional trends rather than the exact magnitude of the impact on driver
behavior.
Based on the results of the public survey study, which are presented in Section
6.1, the billboard advertisements for the simulator study were selected to have a yellow
background with black text. The slogan and graphics which were used in the
advertisement were dependent on the element being studied in the particular scenario.
The billboard advertisements which were used throughout the three scenarios can be seen
in Appendix C. The next three subsections provide further descriptions on how each of
the three scenarios was designed to evaluate a particular element of the billboards.
4.4.1 Scenario #1 – Slogans
The first scenario was designed to test 12 different slogans in the billboard
advertisement to see which were effective from a safety perspective. The slogans selected
for the study were either used in previous work zone safety campaigns or variations of
those slogans that were previously used. The slogans from previously conducted safety
campaigns included those that were conducted in Ohio, Michigan, California, and
Pennsylvania (3). All twelve slogans have the theme of driving slower through work
zones in order to create a safer environment for both work zone personnel and the driving
public. In order to limit the amount of time that a participant drove in the simulator, the
first scenario was divided into four different versions, as shown in Table 2, each of which
consisted of three slogans.
59
Table 2: Slogans used in Scenario #1
Version
1st Slogan
A
Slow Down in
Work Zones
B
Slow Down Respect
the Barrel
C
Slow Down
Save Lives
D
Go Slow Thru
the Zone
2nd Slogan
Let ‘em Work
Let ‘em Live
Give ‘em a
Brake
Brake for
Barrels
My Daddy
Works Here
3rd Slogan
See Orange
Drive Slow
Drive Slow
Drive Safe
Go Slow for
Safety
Slow for the
Cone Zone
The slogans were divided across the four versions of Scenario #1 in an attempt to
vary the keywords used to communicate the safety message across the three slogans for a
particular setup. For example, “Slow Down Respect the Barrel” and “Brake for Barrels”
were assigned to different groups since they each included the keyword barrel within the
slogan. Also there were four slogans which did not contain the keyword slow, so those
four slogans were separated. The slogans were divided up in this manner so participants
were receiving different messages as they passed the various billboards within the
scenario. If participants found the messages to be too repetitive then they may tend to
ignore the billboards and not receive the safety message. Furthermore, in order to
eliminate a possible bias from the order in which the billboards were presented to the
drivers, each of the four versions was designed into three subsets (i.e., A1, A2, A3) so
that each of the slogans was presented as the first, second, and third slogan encountered.
Throughout Scenario #1, the driving environment presented was kept constant
with only the billboard advertisements varying across the four versions and their
respective subsets. All of the test billboards in Scenario #1 featured a slogan with the
graphics used in the original billboard advertisement. As seen in Figure 3, the slogans
60
were framed by the Laborer’s Local 860 logo on the left side of the advertisement and the
construction barrel caricature on the right. The same layout for the advertisement was
used throughout with only the slogan changing from one billboard to the next.
Figure 3. Sample billboard advertisement used in Scenario #1.
Control billboards were also used in the study and consisted of a yellow
background with no content. The yellow background was maintained for the control
billboards rather than using the blank billboards in the simulated environment which have
a white background. Changing the background color between the control and test
billboards would introduce another factor which may impact glance behavior as
participants may glance towards the billboards solely due to the different background
color.
In the layout for Scenario #1, participants drove through four work zones with a
billboard located prior to the driver entering the work zone. All of the billboards in
Scenario #1 were located 100 feet from the right edge of pavement and were placed 250
feet before the “Road Work 1 Mile” sign in the advanced warning area of the work zone.
The billboards were also rotated towards the driver by 15 degrees and were placed on the
61
ground so that they were approximately 20 feet in height. The four work zones consisted
of the same right-lane closure based on the configuration outlined in the Ohio Manual of
Uniform Traffic Control Devices. The work zones were approximately 1.5 miles in
length and the posted speed limit was reduced to 55 mph within the work zone. Speed
limit signs were posted at the end of the work zone to restore the speed limit back to 65
mph. The work zone configuration showing how the warning signs and construction
barrels were setup along with the location of the billboard can be seen in Figure 4.
Figure 4. Work zone configuration used in Scenario #1.
The four work zones were each separated by a mile-long stretch of regular
highway. These sections were provided for buffer areas between the work zones in order
to provide participants enough time to adjust back to their normal driving habits after
having driven through a work zone. Throughout Scenario #1, the first billboard
encountered was always a control billboard with the remaining billboards being the test
advertisements. The control billboards were used to establish a baseline condition for
62
driver behavior so comparisons could be made with the driver behavior exhibited in the
presence of the campaign billboards.
4.4.1 Scenario #2 – Graphics
The second scenario was used to test three combinations of graphics that were
featured in the billboard advertisement to see which combinations were effective from a
safety perspective. The graphics that were used for this part of the study were the
Laborer’s Local 860 logo and the construction barrel caricature that were used in the
original billboard advertisement. The first combination consisted of the logo and barrel
framing the slogan as used in the first scenario. The second combination of graphics
framed the slogan with the construction barrels whereas the third combination featured
the logo to the left of the slogan. All three of the advertisements featured the same yellow
background with black text that was used in the first scenario. The “Slow Down in Work
Zones” slogan was selected for the test billboards since it was the slogan that was used
during the public opinion survey. Thus, consistency in the evaluation of the other
elements of the billboard advertisements was maintained.
The setup of the roadway environment for Scenario #2 was the same as that used
throughout Scenario #1 with billboards placed prior to a driver’s entry into each of the
four work zones. The work zone configurations were again right lane closures for a
distance of 1.5 miles and the four work zones spaced a mile apart from one another.
Locations and placement of the billboards were the same as for Scenario #1. The first
billboard encountered was the control billboard consisting of a blank yellow background,
with the three remaining billboards displaying the different graphic combinations. Three
63
versions of Scenario #2 were created so that each graphic combination was presented on
the second, third, or fourth billboard encountered.
4.4.2 Scenario #3 – Placement and Orientation
The third scenario was used to test 12 different combinations of the physical
placement and orientation of the billboard advertisement with respect to the driver to
determine which combinations were effective from a safety perspective. In order to create
the 12 combinations shown in Table 3 below, the billboards were placed at different
heights, on either side of the highway, and orientated at three different angles with
respect to the driver. For the vertical placements, as listed in Table 3, billboards were
placed either on the ground (Low) or on top of an industrial building that was 20 feet in
height (High). Billboards were always centered at 100 feet from the edge of pavement on
either the left or right side of the highway. Finally, the orientation angles were measured
perpendicular to the centerline of the highway.
Table 3: Placement and orientation combinations used in Scenario #3
Combination # Vertical Placement Side of Road Orientation Angle
1
High
Left
0°
2
High
Left
15°
3
High
Left
60°
4
High
Right
0°
5
High
Right
15°
6
High
Right
60°
7
Low
Left
0°
8
Low
Left
15°
9
Low
Left
60°
10
Low
Right
0°
11
Low
Right
15°
12
Low
Right
60°
64
The angles were selected to be either zero, 15, or 60 degrees so that the billboard
was either directly facing the driver, slightly angled towards the driver, or almost parallel
to the driver, respectively. These three angles represent how billboards are typically
found along actual roadways. The 60 degree angle was primarily selected to represent the
billboard used in the original campaign that was parallel to the highway. Orienting the
billboard parallel to the highway was believed to be part of the reason why the original
sign was believed to be ineffective. For a parallel orientation, drivers would need to look
farther away from the roadway in order to see the sign than if the billboard was angled
more towards the driver’s direct line of sight. By selecting the 60 degree angle, the
billboard is almost parallel to the roadway, yet is angled towards the driver slightly in
order to attract their attention.
For Scenario #3, the participants again drove along the four-lane highway in an
urban area as in the previous two scenarios. However, unlike the previous scenarios,
work zones were not setup on the highway and the posted speed limit was 65 mph
throughout the entire scenario. As participants drove along the highway they passed a
total of 12 billboards which were spaced approximately 0.6 miles apart. For Scenario #3,
it was not considered feasible to have the participants drive through work zones after
seeing each of the billboards as that would result in a scenario which was too lengthy.
The participants would likely begin to become fatigued by driving through such a long
scenario and would likely drive in a manner which would cause them to finish the
scenario sooner. Consequently, having participants alter their behavior in order to
65
complete the study earlier would likely not represent their true driving behavior in an
actual environment and would not be beneficial to this study.
Three versions of Scenario #3 were created with the placement and orientation
combinations randomized in each to prevent the driver from being able to predict where
the next billboard will be located. If drivers were able to anticipate where the billboard
would be located, they may start looking at that area on the screen before actually seeing
the billboard. This would affect the accuracy of the eye-tracking data and the evaluation
of the billboards from a safety perspective. The placement and orientation combinations
were considered to likely have a greater impact on glance behavior than the other
billboard elements studied. This was due to the participants having to look at different
areas of the roadway environment in order to see the different billboards. However,
driving performance was still evaluated throughout Scenario #3, so that the effects from
the billboards on both aspects of driver behavior would be evaluated as done for the two
previous scenarios.
4.5 Study Procedure
The entire simulator study was composed of the testing scenarios as described
above along with the informed consent process, two questionnaires, and two adaptation
scenarios. The length of time in which a participant needed to complete the study was
selected to be a maximum of one hour, so as to limit the amount of time that participants
would be driving in the simulator. Having participants drive for periods of longer than
one hour increases the potential of fatigue setting in which may cause the driver to alter
their behavior so that they could finish the study faster. In order to provide enough time
66
for participants to complete the entire study, general guidelines were set to help limit the
time spent on each aspect. The informed consent process and completion of the
questionnaires was allotted 15 minutes. The adaptation scenarios and set up of the eyetracking system were allotted 15 minutes while the testing scenarios were allotted 30
minutes.
The testing scenarios were conducted in different phases of the driving simulator
study with the first phase consisting of Scenario #1 while the second phase was Scenarios
#2 and #3. Each of the three testing scenarios were calculated to take approximately 10
minutes to complete based upon an individual driving at the posted speed limit for the
entire length of the scenario. Since Scenario #1 was split into four versions that each had
different slogans, a single participant could drive through three of the versions within the
30 minute window allotted for the testing scenarios. While Scenarios #2 and #3 each had
different versions, the same billboards are used throughout so participants only had to
complete those scenarios once. As a result, participants in the first phase of the study
drove through three different versions of Scenario #1 while the participants in the second
phase drove one version of Scenario #2 and one version of Scenario #3.
Participants were largely recruited from the Psychology Department’s human
subject research system. Those students who completed the study through this method
earned a research credit for their participation. Additional students were recruited from
the general student body by direct contact and flyers, but those individuals were offered
no compensation for completing the study. All of the participants recruited for the study
67
were required to be at least 18 years of age, have a valid driver’s license, and have at least
two years of driving experience.
When participants first entered the Safety and Human Factors Facility, they were
introduced to the study through a narrative paragraph which informed them of the
procedure and provided a general overview of the study. The overview of the study
indicated that they would be driving past several billboard advertisements, but no
description was given as to their content. After being introduced to the study, participants
were asked to complete an informed consent form and were allowed to ask any questions
in regards to the study. Participants were then asked to fill out a pre-test questionnaire
which included questions on driver demographics, driving experience, and opinions on
safety campaigns. After these two elements had been completed, the driving simulator
portion of the study could commence.
Participants were then shown the driving simulator and given an opportunity to
adjust the seat as necessary and locate the controls. Next, participants were required to
complete two adaptation scenarios prior to proceeding on with the study. The adaptation
scenarios were designed to provide participants with ample opportunity to become
familiar with the driving simulator controls and reach a point where they were as
comfortable driving in the simulated environment as they would be in an actual driving
environment.
The first adaptation scenario featured a two-lane rural roadway. During the
scenario, subjects had to maintain their speed and lane position for approximately thirty
seconds. Feedback tools were provided to the participants in order to help them adjust to
68
how the driving simulator operates. The first feedback tool guided participants as to
where the vehicle was positioned within the lane so they could adjust to how that appears
on the screens. The lane position feedback tool consisted of five circles that were present
on the screen directly in front of the driver. The circles would light up based upon where
the vehicle was positioned within the lane with only one circle on at a time.
When the vehicle was in the center of the lane then the center circle was green,
but as the vehicle moved to the left side of the lane, yet was still within the lane, the
circle directly left of center turned yellow. If the vehicle crossed the centerline into the
oncoming lane then the far left circle turned red. The same procedure happened as the
vehicle moved to the right with the circle right of center turning yellow when the vehicle
was on the right side of the lane and the far right circle turning red when the vehicle
crossed the edge line onto the shoulder. If the vehicle went all the way outside of the lane
then all five circles turned red.
The second feedback tool provided to the participants was used to assist them
with getting adjusted to the vehicle’s acceleration. If participants drove over the posted
speed limit of 45 mph by more than 5 mph, the simulator issued a verbal command which
instructed the driver to “Slow Down.” After participants were able to maintain their speed
below 50 mph and keep the vehicle within the lane for approximately 30 seconds, then
they successfully completed the first adaptation scenario. At that point the simulator
again issued another verbal command which said “Practice Complete” and the screens of
the simulator faded to black and the vehicle came to a stop automatically.
69
The next adaptation scenario featured a four-lane undivided highway in an urban
environment. No traffic was present on the highway and the virtual world was vacant of
work zones and billboards. Participants were instructed to drive along the highway until
they felt comfortable with the controls of the simulator and being able to drive in that
environment. Observations on how the participant was driving were made by the
researcher to ensure that they were able to successfully keep the vehicle in a travel lane
and drive at an appropriate speed. On average, participants drove through the second
adaptation scenario for approximately five minutes. Once participants indicated they felt
comfortable driving in the simulator and the researcher observed this to be true, the
second adaptation scenario was stopped and the study proceeded.
Once a participant had completed the two adaptation scenarios, the eye-tracking
system was configured for that individual. Participants were asked to sit in their normal
driving position and look straight ahead at the screen. The cameras were then adjusted so
that the participant’s face was centered within the frame of each camera. Then a threedimensional face model was created for that participant using the eye-tracking software.
The face model was created by locating six reference points on the image of the
participant’s face. Reference points were placed on the corners of both eyes and on the
corners of the mouth, as shown in Figure 5. After the reference points were established,
the faceLab 5 software was able to lock on to the glint of light on the eye which resulted
from the infrared light being reflected onto the participant’s face. The system was able to
track either the iris or the pupil, depending on which provided greater accuracy. Gaze
vectors were available for each eye as can be seen by the green lines centered on each of
70
the eyes in Figure 5. The eye-tracking system used a world model, as seen in the frame
on the right in Figure 5, to relate the location of the participant to the simulator screens.
Figure 5. Example of a face model created using the faceLab 5 eye-tracking system.
After the three-dimensional model was created, a gaze calibration procedure was
completed in order to increase the accuracy of the eye-tracking system. Participants were
instructed to follow a flashing white dot with their eyes while keeping their head in the
same position. The dot would appear on the visible part of the center screen which was
indicated by a box that was directly in front of the driver. The dot started in the upper
left-hand corner of the box and moved across the top of the box to the middle and then to
the upper right-hand corner. The dot then proceeded to move down across the middle of
the box from the right side to the middle to the left side. Then finally down across the
bottom of the box from the left corner to the middle to the right corner. At each of these
nine positions the dot would stop and the center of the dot would flash which indicated
that the eye-tracking system was capturing the participant’s gaze.
71
After the dot had moved through the nine positions, a message appeared on the
screen displaying the accuracy of the calibration. If the accuracy of the calibration was
within one degree of error for both eyes, the calibration was considered acceptable. If the
error was greater than one degree for either eye, the gaze calibration procedure was
repeated. In some instances, an error less than one degree could not be obtained through
three trials due to issues with how the eye-tracking system was operating for that
particular participant. For these instances, the minimum error that could be obtained was
accepted and the problems with the eye-tracking system were noted for possible reference
during the analysis of the eye-tracking data.
Next, the testing scenarios were run with instructions given for each of the
particular scenarios prior to the participant being allowed to begin the scenario. For
Scenarios #1 and 2, participants were instructed to drive along the highway as they
normally would and were told that they would be driving through four work zones and
past several billboards, but no indication was given on the content of the billboards. For
Scenario #3 participants again were instructed to drive along the highway as they
normally would and were told they would be driving past several billboards, but were not
told the content of the signs. In between the testing scenarios, participants were given an
opportunity to exit the driving simulator if they were in need of a break.
During the evaluation of the slogans in the first phase of the simulator study,
participants were asked to fill out a short post-test questionnaire upon completing
Scenario #1. No post-test questionnaires were administered after Scenarios #2 and #3 as
the same slogan was used in the billboards for those scenarios. The questionnaire had two
72
questions in regards to the billboards they had just seen during the scenarios. The first
question sought the opinion of the participant in regards to billboard clutter. This question
was included so a comparison could be made to the billboard advertisements that were
presented in the public survey study. The billboard advertisements for this portion of the
study were simplified down to a single slogan and two graphics whereas the
advertisements in the survey had multiple lines of text and two graphics. Thus, this
question was used to see if participants still found these simplified billboards to be
cluttered with information.
The second question in the post-test questionnaire had participants recall the
slogans which they remembered best from the scenarios. This was done to determine the
most memorable slogans out of the nine slogans that the participants were shown.
Participants were required to write down the slogans when answering the questionnaire to
assess how accurately the slogan was recalled. This option was selected rather than
providing a list of slogans which the participants could select from as it was considered to
be a better evaluation of the ability of a slogan to reach drivers.
After completing the post-test questionnaire during the first phase or after
completing the two scenarios during the second phase, participants were debriefed on the
purpose of the study being an evaluation of billboards as a means of communicating a
work zone safety campaign message. Participants were given the opportunity to ask any
questions they had in regards to the study and to offer any feedback or comments. At that
point in time, participants were informed that they had completed the study and were
thanked for their participation.
73
4.6 Data Collection
Throughout all of the testing scenarios, data was collected on the participant’s
driving performance and glance behavior. The driving simulator produced log files of the
driving performance for each of the scenarios. The eye-tracking system was run
throughout each of the scenarios and produced log files which detailed the participant’s
glance behavior. For both the driving simulator and the eye-tracking system, the data
collection rates were set at 60 Hz so that the data could be correlated during the analysis
process. This provided the ability to analyze how the two aspects of driver behavior were
each affected by the billboards and to examine the relationship between the two.
For the driving simulator, the data collection was divided into the three zones for
Scenarios #1 and 2. The first zone was denoted as Pre-Work Zone and ranged from 1250
feet before the billboard to the “Road Work 1 Mile” sign. This area was selected to begin
1250 feet prior to the billboard as this was determined to be the point at which
participants could first detect the billboard. The second data collection zone, also known
as the Advanced Warning Area, consisted of the traffic signs that provided drivers with
information about the work zone ahead. This area stretched from the “Road Work 1
Mile” sign to the first construction barrel in the work zone. Finally, the third data
collection zone (Work Zone) spanned the length over which the construction barrels were
set up. These three data collection zones were selected since these areas would indicate
the extent to which the billboards impacted the driver’s behavior.
For Scenario #3, the driving simulator was set up to collect data for the entire
length of the scenario since no work zones were in place. However, the data was only
74
analyzed for sections prior to each of the 12 billboards in the same manner that was done
for the Pre-Work Zone in Scenarios #1 and 2. The driving performance data in Scenario
#3 was analyzed for the area beginning when the subject was 1250 feet before the
billboard up until the subject was 250 feet past the sign.
For Scenarios #1, 2, and 3, the eye-tracking system was used to collect data on the
participant’s eye movements throughout the entire length of the scenarios. The system
had to be manually set to collect data for each of the participants. The eye-tracking began
shortly after the participants started driving and was then stopped immediately before the
participant was instructed to stop driving. The log files produced by the eye-tracking
system and the driving simulator had time stamps which enabled researchers to match the
data from the two systems.
The driving performance measures that were collected by the driving simulator
were: speed, lane position, acceleration, deceleration, and collisions. Speed data was used
as an indication of a driver’s perceived risk when traveling along the highway. If drivers
travel at a lower speed, the driver may perceive the roadway conditions to be unsafe and
thereby reduce their speed in an attempt to feel safer. Conversely, if a subject drives at a
high speed, the driver may perceive little to no risk for that driving environment. Overall,
the billboards evaluated in the study indicated that drivers should reduce their speeds
when driving through work zones, thereby increasing safety within the work zones. Thus,
if drivers decrease their speed after seeing the billboards, this indicated that they received
the message and recognized the potential risk. Furthermore, the mean speeds and the
variance in mean speeds over the different data collection zones were used in the safety
75
evaluation of the billboards in regards to their effectiveness in communicating the work
zone safety message.
Mean speeds in the work zones were evaluated to determine if the billboards
caused drivers to display the desired behavior. Comparisons of the mean speed and speed
variance were made across the data collection zones to evaluate the different elements of
the billboard advertisements that were presented to drivers including the billboards
without advertisements. Statistical tests will be used to determine if the mean speeds and
speed variances between the different billboard advertisements presented to the drivers
were statistically significant. The null hypothesis for the speed analysis in the different
scenarios stated that there is no difference in the mean speeds between the different
elements of the billboards.
The lane position data collected was the position of the vehicle’s center of gravity
relative to the centerline of the travel lane. Values were reported as positive when the
vehicle was traveling on the right side of the lane and negative when the vehicle was
traveling on the left side of the lane. Since the travel lanes in the simulator study were 12
feet in width, the limits of the lane position data were +6 feet for the far right side of the
lane and -6 feet for the far left side of the lane. Lane position data was used as an
indication of a driver’s perceived risk when traveling through work zones. If drivers
travel farther away from the construction work area within their travel lane, the drivers
may have perceived the work zone as a safety risk. Since participants were driving on an
undivided four-lane highway with a concrete barrier separating the directions of travel
and on which the right lane was closed, drivers moving farther away from the work zone
76
would be moving toward the concrete barrier adjacent to the inside shoulder. Traveling
farther away from the work zone also increases the safety of the work zone as the
likelihood of an intrusion crash into the work area is decreased.
Additionally, the variation in lane position data was used as an indication of the
potential distraction related to the billboards. If the lane position varied greatly as a driver
approached a billboard, this could indicate that the billboard was distracting a driver from
the roadway ahead. Comparisons of the mean lane position and the variations in lane
position were made across the data collection zones to evaluate the billboard
advertisements presented to drivers. Statistical tests were used to determine if the mean
lane position and variations in lane position between the different billboard
advertisements were statistically significant. The null hypothesis for the lane position
analysis in the different scenarios stated that there was no difference in the mean lane
position and variations in lane position between the different billboard elements.
Based on the instrumentation of the driving simulator, the acceleration and
deceleration data collected were normalized values of the pressure exhibited by the driver
on the accelerator and brake pedal, respectively. When a pedal is not depressed the value
is 0.00 and when a pedal is at maximum depression the value is 1.00. While deceleration
of the vehicle without the use of the brake pedal is possible, the behavior of interest was
whether participants actively decelerated because they perceived a need to slow down
more immediately. Both acceleration and deceleration data were used as an indication of
a driver’s reaction to the billboards and their potential for being involved in a crash when
traveling along the highway. The mean accelerations and the variance in acceleration
77
were found for the different data collection zones and comparisons were made across the
zones and between the different billboards presented to the drivers. Likewise, the
deceleration data was evaluated in a similar manner. Statistical tests were used to
determine if the mean acceleration and mean deceleration between the different billboard
advertisements were statistically significant. The null hypothesis for the acceleration and
deceleration data stated that there are no differences in the mean values between the
different billboard elements.
Collision data was recorded when a participant collided with another object and
returns the name of the object that was involved in the collision. In addition, the subject’s
position in the simulated environment and the point of time during the scenario were used
in determining the details of the collision. Collision data was used as the actual risk of a
driver when traveling along the highway and the potential distraction related to the
billboards. The observed frequency of crashes served as an indication of the risk present
on the roadway environment and indicated potential aspects of the roadway environment
that were contributing to crashes. Comparisons of the crash frequency between the
different elements of the billboard advertisements were done to determine if there is a
particular aspect of the signs which may be contributing to crashes.
The eye-tracking system collected the participant’s eye movements in relation to
the screens of the driving simulator and provided the data as coordinates in the threedimensional model created by the researcher. The model was correlated to the physical
environment of the simulator screens by measuring the screen area on which the
simulated environment was presented. Specifically, as the driver approached the
78
billboard, the physical area on the screens in which the billboard appeared was measured
and coordinated to the three-dimensional model used in the eye-tracking system.
Furthermore, the eye-tracking data was coordinated with the subject’s position along the
roadway as provided by the driving simulator data by using the time stamps included
with each of the log files. The researcher was then able to determine what aspects of the
simulated environment were present on the screen in front of the driver at a particular
point of time. Thus, as the subject approached the billboard, their eye movements were
analyzed to determine when they glanced towards the billboard.
The glance behavior was analyzed from 1250 feet prior to the billboard up until
the sign was passed for every billboard presented to the participants in the study. The
participant’s gaze intersecting the area on the screen where the billboard appears was
considered a glance towards the billboard. Consecutive gaze intersections represented the
participant making a single fixation on the billboard. The minimum length of time for a
fixation was considered to be 0.10 seconds as this was believed to be the shortest amount
of time required for a participant’s cognition of the billboard. Once individual fixations
were determined for each of the billboards, several other glance behavior parameters
were then used to evaluate the billboards. These glance behavior parameters included
mean fixation, total fixation, number of fixations, and proportion of fixation duration.
The length of time that a participant fixated on a billboard advertisement indicated
how long it took the driver to comprehend the billboard and whether or not the billboard
was distracting to the driver. The glance behavior measures used in this study provide
evaluations on these two issues. Individual fixations represent all of the glances made
79
towards billboards and were used to determine if there was a billboard which had a
higher frequency of long glances. For this study long glances were considered to be those
individual fixations that are longer than 2.0 seconds as this was the limit suggested by
Rockwell (22). The frequency of long glances towards the billboards indicated if drivers
found the signs to be distracting.
Mean fixations were determined as an average of all the individual fixations a
participant made towards a billboard and provided a generalized duration for how long
participants needed to look at the billboard in order to comprehend the message. Mean
fixations also indicated if drivers find the billboards to be distracting as higher value
indicated longer glances overall. Total fixation was the total length of time that a subject
glanced at a particular billboard. This parameter was used to evaluate the length of time
needed to comprehend the billboards as larger total fixation values indicated that drivers
had to look at the billboard for longer periods of time in order to receive the message.
Number of fixations was recorded as the number of times a participant looked at a
billboard. This parameter was used to evaluate whether or not the billboard was
distracting as billboards with more frequent fixations were considered distracting.
Finally, proportion of fixation duration was a ratio of the total time a participant
glanced at a billboard compared to the total amount of time they were approaching the
sign. Alternatively, this parameter was considered the percentage of time that participants
looked at the billboard while approaching the sign. This parameter indicated if the sign
was distracting by comparing the values obtained for the test advertisements to those
values for the control advertisement. Statistical tests were conducted on all of the glance
80
behavior parameters to determine if there were statistically significant differences
between the billboard advertisements. The null hypothesis for each of the glance behavior
parameters stated that there were no differences in the mean values between the different
billboard elements being evaluated.
4.7 Sample Size Determination
For the simulator study, the number of participants required was determined based
upon the data collected, a 95 percent level of confidence, and a statistical power of 80
percent as selected for the statistical tests. The following equation was used in this study
to determine the sample size, n, that would be required to detect a certain difference level
in each of the driver behavior parameters:
(42)
Where: σ = Sample standard deviation
Zβ = Critical value for a power of 80 percent (β = 0.2) = 0.842
Zα/2 = Critical value for a 95 percent level of confidence (α = 0.05) = 1.96
ε = Detectable difference
The sample size required for each of the eight data collection parameters was
determined based upon a range of standard deviations and a range of detectable
differences. From the previous research reviewed, it was seen that neither the standard
deviations nor detectable differences were reported in any of the simulator studies.
Furthermore, the standard deviation of the data collected may be impacted by the
equipment used to collect the data. Previously, a simulator study conducted in the ORITE
Safety and Human Factors Facility reported several of the data collection parameters that
81
were used in this study (44). This previous study was similar in nature as it evaluated
how driver behavior was affected by several traffic signs. The standard deviations for
each of the parameters were reported separately and used as a guideline for the sample
size determination in this study. The detectable difference values used in the initial
sample size determination were selected based upon the practicality of measuring each of
the parameters. Once five participants had completed Scenario #1, the resulting standard
deviation for each parameter was determined and used to select the final sample size. The
description of this process as it applies to each of the parameters is discussed below.
4.7.1 Determination for Speed
From the previous simulator study conducted by ORITE, the standard deviation
for speed was reported to range from 2.87 to 12.23 mph for the different elements tested
with an average standard deviation of 5.55 mph for the study (44). Therefore, an
approximate range of standard deviations from 3.0 to 10.0 mph were used for the initial
sample size calculations. The range of values for the detectable difference in speed were
selected to range from 1.0 mph, the lowest discernible variation possible on the
speedometer in the simulator, to 5.0 mph which is the largest error possible when using a
laser speed gun to conduct a field study. The selected ranges for standard deviation and
detectable difference were used in the equation above to produce Table 4.
82
Table 4: Initial sample size determination for speed
Standard Deviation
in Speed (mph)
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
Detectable Difference in Speed (mph)
1.0
71
126
197
283
385
503
636
786
2.0
18
32
50
71
97
126
159
197
3.0
8
14
22
32
43
56
71
88
4.0
5
8
13
18
25
32
40
50
5.0
3
6
8
12
16
21
26
32
After the first five participants had completed Scenario #1, the mean speeds were
determined for each of the data collection zones. For the Pre-Work Zone, Advanced
Warning Area, and Work Zone the standard deviations in speed were found to be 4.0, 4.6,
and 5.6 mph, respectively. Examining Table 4, it can be seen that larger standard
deviation values result in a larger sample size for a particular detectable difference.
Therefore, the largest value from the five participant sample was used to select the initial
sample size. Of the detectable differences presented in Table 4, the most practical value
from a safety standpoint is likely 3.0 mph as it would likely be the smallest difference
that would be noticeable to drivers while traveling through an actual work zone.
Furthermore, a change in speed of less than 3.0 mph would be difficult to detect on actual
roadways due to the limitations of typical data collection instruments (i.e., radar guns).
Therefore, from Table 4, for a detectable difference of 3.0 mph and a standard deviation
of 5.6 mph, the required sample size would be between 22 to 32 participants.
83
4.7.2 Determination for Lane Position
Lane position was evaluated in the previous simulator study and the standard
deviation ranged between 0.46 to 0.82 feet with an overall average standard deviation of
0.60 feet (44). Therefore, an approximate range of standard deviations from 0.25 to 2.00
feet were used for the initial sample size calculations to cover potentially greater values.
The smallest detectable difference in lane position was selected to be 0.10 feet as this was
the lowest variation the driving simulator could report. The largest detectable difference
in lane position was selected to be 1.00 feet as larger values would not be useful in
assessing the billboard’s impact on driver behavior. The initial sample size calculations
using the selected ranges for standard deviation and detectable difference are shown in
Table 5.
Table 5: Initial sample size determination for lane position
Standard Deviation Detectable Difference in Lane Position (ft)
in Lane Position (ft)
0.10
0.25
0.50
0.75
1.00
0.25
50
8
2
1
1
0.50
197
32
8
4
2
0.75
442
71
18
8
5
1.00
786
126
32
14
8
1.25
1227
197
50
22
13
1.50
1767
283
71
32
18
1.75
2405
385
97
43
25
2.00
3141
503
126
56
32
After the first five participants had completed Scenario #1, the mean lane position
was determined over each of the data collection zones and the largest standard deviation
of lane position was found to be 0.66 feet. For the variation in lane position throughout
84
the area prior to the billboard the standard deviation was also found to be 0.66 feet. The
smallest and most practical difference in lane position for drivers traveling along an
actual roadway is likely 0.50 feet. The lane lines used on highways are typically three to
four inches in width and drivers are not likely to be able to visually notice changes in lane
position that are less than twice the width of a lane line. As a result, Table 5 shows that
for a difference of 0.50 feet and a standard deviation in lane position of approximately
0.75 feet, the required sample size would be 18 participants.
4.7.3 Determination for Acceleration and Deceleration
The acceleration and deceleration data are normalized values, so the initial sample
size determination was done using one of the parameters as the sample size would then
be adequate for both parameters. From the previous simulator study, the largest
acceleration value was found to be 0.38 while the maximum deceleration value was
found to be 0.16 (44). Furthermore, the mean deceleration values were all very low
(<0.01) indicating that participants did not use the brake pedal very often. Therefore, the
initial sample size was determined using acceleration data as there was more data
available to use as a guideline.
In the previous ORITE study, the standard deviations for acceleration were
reported between 0.02 and 0.07 with an average standard deviation of 0.04 (44).
Therefore, the range of standard deviations for the initial sample size calculations was
selected to be 0.01 to 0.08. The mean values for acceleration varied between 0.06 and
0.26 (44). Thus, the lowest detectable difference was selected to be 0.01 as this was the
lowest variation reported by the simulator, while the highest detectable difference was
85
selected to be 0.20 as this was approximately half of the maximum acceleration reported.
Table 4 displays the initial sample sizes calculated for the acceleration data based on the
selected ranges for standard deviation and detectable difference.
Table 6: Initial sample size determination for acceleration
Standard Deviation Detectable Difference in Acceleration
in Acceleration
0.01 0.05 0.10 0.15 0.20
0.01
8
1
1
1
1
0.02
32
2
1
1
1
0.03
71
3
1
1
1
0.04
126
6
2
1
1
0.05
197
8
2
1
1
0.06
283
12
3
2
1
0.07
385
16
4
2
1
0.08
503
21
6
3
2
After the first five participants had completed the study, the mean acceleration for
each of the data collection zones was determined and the largest standard deviation of
acceleration was found to be 0.08. Additionally, the mean acceleration across the three
data collection zones was found to be 0.18. Therefore, a detectable difference of 0.05
would be approximately a quarter of the mean value obtained from the five participant
sample. Differences in acceleration of less than 0.05 would likely not be practical on
actual roadways. From Table 6, it can be seen that for a detectable difference of 0.05 and
a standard deviation of 0.08 the sample size estimation would be 21 participants.
4.7.4 Determination for Mean Fixation and Total Fixation
The previous simulator study reported results for mean fixation but did not
include data on total fixation. Nevertheless, the standard deviations for mean fixation
86
were used as a guideline for the total fixation data as large variations between the
parameters would not be expected as total fixation is a summation of the glances which
are averaged to produce mean fixation. The standard deviation for mean fixation was
reported between 0.41 and 0.54 seconds for the four signs with an overall average of 0.47
seconds (44). Therefore, a range of standard deviations from 0.10 to 2.00 seconds was
used for the mean fixation data to account for the longer glances which were expected
towards the billboard advertisements as opposed to the traffic signs in the previous study.
For total fixation, the range was adjusted slightly higher to be from 0.25 to 2.50 seconds
to account for larger values.
The smallest detectable difference in mean fixation was selected to be 0.10
seconds as this was the lowest fixation used in the analysis of the eye-tracking data. The
largest detectable difference in mean fixation was selected to be 1.00 seconds as longer
glances that begin to approach the 2.00 second limit are seen in previous research to
negatively impact safety. The range for detectable differences for the total fixation data
was selected to be similar to that used for the mean fixation data, but slightly higher. The
initial sample size calculations for both mean and total fixation are shown in Tables 7 and
8, respectively.
87
Table 7: Initial sample size determination for mean fixation
Standard Deviation in
Mean Fixation (sec)
0.10
0.25
0.50
0.75
1.00
1.25
1.50
2.00
Detectable Difference in Mean Fixation (sec)
0.10
8
50
197
442
786
1227
1767
3141
0.25
2
8
32
71
126
197
283
503
0.50
1
2
8
18
32
50
71
126
0.75
1
1
4
8
14
22
32
56
1.00
1
1
2
5
8
13
18
32
Table 8: Initial sample size determination for total fixation
Standard Deviation in Detectable Difference in Total Fixation (sec)
Total Fixation (sec)
0.25
0.50
0.75
1.00
1.50
0.25
8
2
1
1
1
0.50
32
8
4
2
1
0.75
71
18
8
5
2
1.00
126
32
14
8
4
1.25
197
50
22
13
6
1.50
283
71
32
18
8
2.00
503
126
56
32
14
2.50
786
197
88
50
22
After the first five participants had completed the study, the mean and total
fixation for each of the billboards presented to the drivers was determined. The standard
deviation of the mean fixation data was found to be 0.55 seconds while the standard
deviation of the total fixation data was 1.51 seconds. In regards to the detectable
difference, the smallest practical value with regards to safety would likely be 0.50
seconds or a quarter of the 2.00 second limit for glances away from the roadway. This
detectable difference could also be used for the total fixation data as they are related
88
variables. However, from the sampled data the average of the total fixations was found to
be approximately twice that of the mean fixations. Consequently, a larger detectable
difference, such as one second, should likely be used in the initial sample size estimate.
From Table 7, a detectable difference in mean fixations of 0.50 seconds and a standard
deviation of 0.55 seconds results in a sample size of eight participants. From Table 8, it
can be seen that for a detectable difference of 1.00 seconds and a standard deviation of
1.51 seconds the estimated sample size is 18 participants.
4.7.5 Determination for Number of Fixations
The previous simulator study did not report the number of fixations made towards
each of the traffic signs, so no guideline was available for that data. The range of standard
deviation values was selected to be from 0.25 to 2.00 fixations as this data was not
expected to vary greatly between participants. These values were used for the initial
sample size estimate and would be reevaluated after data was collected for this study.
Since the number of fixations made by participants was recorded as integer values, the
range of detectable differences was selected to be in 0.50 intervals, or half of a fixation.
This would result in the detectable difference being to either one whole number or
another. The range of detectable differences for the initial sample size calculations was
selected to be from 0.50 to 2.50 fixations. The initial sample size calculations using the
selected ranges for standard deviation and detectable difference are shown in Table 9.
89
Table 9: Initial sample size determination for number of fixations
Standard Deviation in Detectable Difference in Number of Fixations
Number of Fixations
0.50
1.00
1.50
2.00
2.50
0.25
2
1
1
1
1
0.50
8
2
1
1
1
0.75
18
5
2
2
1
1.00
32
8
4
2
2
1.25
50
13
6
4
2
1.50
71
18
8
5
3
1.75
97
25
11
7
4
2.00
126
32
14
8
6
After the first five participants had completed the study, the number of fixations
towards each billboard presented to the driver was determined. The standard deviation in
the number of fixations made by those participants was found to be 1.60. Thus, the
selected range for the initial sample size estimate was found to be adequate. As for a
practical detectable difference, it is conceivable that a passenger could indeed detect the
number of glances a driver makes away from the roadway ahead, so a detectable
difference of a single fixation could be considered practical. Thus, using those two
values, the sample size obtained from Table 9 would be between 18 to 25 participants.
4.7.6 Determination for Proportion of Fixation Duration
The previous simulator study conducted by ORITE reported standard deviations
for proportion of fixation duration between 0.08 and 0.10 with an average standard
deviation of 0.09 for the study (44). Consequently, the range of standard deviations was
selected to be from 0.05 to 0.40 in order to account for potentially higher values due to
probable differences in glance behavior towards the billboards in contrast to the traffic
signs evaluated in the previous study. The eye-tracking data was analyzed over the Pre-
90
Work Zone section of the scenario which would take a participant approximately 13
seconds to traverse if driving at the posted speed limit. Since the lowest fixation used in
the analysis of the eye-tracking data was 0.10 seconds, the lowest proportion of fixation
duration value for a single glance towards the billboard while driving at the speed limit
would be 0.008. Thus, the smallest detectable difference was selected to be 0.01 as this
would be the minimal difference between no glances and a single glance towards the
sign. The highest detectable difference was selected to be 0.20 which was approximately
half of the maximum values which were reported in the previous simulator study. The
initial sample size calculation for proportion of fixation duration is shown in Table 10.
Table 10: Initial sample size determination for proportion of fixation duration
Detectable Difference in Proportion Fixation Duration
Standard Deviation in
Proportion Fixation Duration
0.01
0.05
0.10
0.15
0.20
0.05
197
8
2
1
1
0.10
786
32
8
4
2
0.15
1767
71
18
8
5
0.20
3141
126
32
14
8
0.25
4908
197
50
22
13
0.30
7067
283
71
32
18
0.35
9618
385
97
43
25
0.40
12562
503
126
56
32
After the first five participants had completed the study, the proportion of fixation
duration for each billboard presented to the driver was determined. The standard
deviation in the proportion of fixation duration data was found to be 0.15. During the
sample size evaluation of the total fixation parameter, the practical detectable difference
was selected to be 1.00 seconds. If a participant was driving towards the billboard at the
91
speed limit and glanced at the billboard for a total of 1.00 seconds, their proportion of
fixation duration would be 0.08. Thus, from a practicality standpoint, the smallest
detectable difference for proportion of fixation duration would likely be 0.10 in order to
account for potentially shorter travel times over the 1250 feet prior to the billboard. From
Table 10, a detectable difference of 0.10 and a standard deviation of 0.15 would result in
a sample size of 18 participants.
4.7.7 Final Sample Size Determination
After examining the standard deviations from the driving performance and glance
behavior data that was collected for the first five participants for Scenario #1, the sample
size tables above were referenced. The initial sample sizes that were determined for each
of the driver behavior parameter are summarized in Table 11 below.
Table 11: Summary of initial sample size determination
Driver Behavior Parameter
Initial Sample Size Estimate
Speed
22 to 32 Participants
Lane Position
18 Participants
Acceleration and Deceleration 21 Participants
Mean Fixation
8 Participants
Total Fixation
18 Participants
Number of Fixations
18 to 25 Participants
Proportion of Fixation Duration 18 Participants
Overall, it can be seen that the driver behavior parameter of speed controls the
determination of the sample size for the simulator study. Since the standard deviation
value from the first five participants fell between the values presented in Table 4 a range
of participants was selected for the sample size. Since the parameter of speed will control
92
the sample size determination, the sample size was calculated to be 28 participants based
on a standard deviation of 5.6 mph and a detectable difference of 3.0 mph. However, the
final sample size was selected to be slightly larger than the calculated value in order to
account for potential changes in the standard deviation values that would be obtained
from the larger sample. As a result, the final sample size was selected to be 30
participants for each of the billboard elements in order to achieve the selected detectable
differences.
The previous driving simulator study conducted by ORITE had a total sample size
of 39 participants (44). The four driving simulator studies reviewed in the Background
chapter also reported their sample sizes. Megias et al. had a sample size of 22 participants
while Edquist et al. and Young et al. each had 48 participants (33, 37, 38). Jamson and
Merat had the largest sample size of 80 participants, yet found very few statistically
significant differences at a 95 percent level of confidence despite the large sample size
(39). Given the sample sizes for the previous studies along with the findings of Jamson
and Merat, the sample size of 30 participants was considered sufficient for this study.
Due to the way that the slogans were split among the four groups in Scenario #1
and that a single participant would drive three versions of the scenario, a total of 40
participants were needed so that each slogan would be seen by 30 participants. Scenarios
#2 and #3 were run in tandem and required an additional 30 participants to complete
those scenarios. Therefore, a total of 70 participants were needed in order to complete the
simulator study, which was the number of drivers that participated in the study.
93
CHAPTER 5: STATISTICAL METHODOLOGY
In order to evaluate the different billboard advertisements, statistical analyses
were conducted on the data collected from both the billboard survey and driving
simulator studies. Statistical testing is an effective tool for assessing the differences in the
data collected and determining whether or not those variations in the data were due to
chance or the selected manipulation. Furthermore, statistical tests can indicate if the
effect being evaluated in the sample groups would truly exist in the population. This is
important to this research as the billboard survey and driving simulator studies conducted
each used sample groups from the driving population in order to assess the billboard
advertisements. By using statistical testing to analyze the data collected in each of the
studies, determinations were made on the potential impact of the various billboard
elements on driver behavior. This was done through a series of comparisons done with
statistical tests which will be described in the following sections. The descriptions are
divided between the two studies conducted and include information on the statistical tests
used in the analysis.
5.1 Statistical Tests for Public Survey Study
The billboard surveys were designed to evaluate the preferred text and
background color combination for the billboard advertisement. Based on how the surveys
were established the data collected was categorical as the participants selected one of the
four signs that were presented to them. Prior to any statistical analyses, the frequencies of
the sign selections needed to be determined. Once the frequencies were totaled, it was
seen that the sample sizes for the two sets of four billboards were unequal as the first set
94
had 156 completed surveys and the second set had 220 surveys. If the two sets had equal
sample sizes, then the statistical comparisons could be made by using the frequencies of
the responses for each of the billboards. However, to compare the two sets given the
different sample sizes, the frequencies were divided by the number of completed surveys
for that particular set to produce the proportion of responses for each of the eight
billboards.
The statistical test selected for evaluating the selection proportions was a twoproportion z-test. There are four conditions that must be met for that particular test to be
valid. The first two conditions are that the samples must be selected from the population
at random and that the samples are independent. Each of these conditions is met by the
design of the experiment since participants were selected at random from the general
driving population to complete the survey and the participants were only shown one set
of billboards. The other two conditions are that the samples must include at least five
responses in each of the groups being compared and that the population size is at least ten
times the sample size (45). These two assumptions were checked while conducting the
statistical analyses.
The proportions were used to calculate the test statistic (z) for the two-proportion
z-test by the following equation:
(45)
Where: p1 = Sample proportion from population 1
p2 = Sample proportion from population 2
SE = Standard error of the sampling distribution between two proportions
95
Calculated as:
(45)
Where: n1 = Size of sample 1
n2 = Size of sample 2
p = Pooled sample proportion
Calculated as:
(45)
For the statistical tests conducted on the survey responses, the null hypothesis
stated that the background and font color combinations used in the billboard
advertisements were selected equally. Conversely, the alternative hypothesis stated that
the two color combinations were not selected equally. The two-proportion z-test was
conducted as a two-tailed test as the null hypothesis did not indicate a directional
relationship. For this research, a 95 percent level of confidence was used throughout the
statistical analysis. Thus, the critical z-score was considered to be 1.96 for the statistical
analysis. If the calculated z-score was found to exceed the critical z-score, a significant
result was indicated and the null hypothesis was then rejected.
In order to determine which of the eight different background and font color
combinations presented in the survey was preferred by the participants, direct
comparisons were made for each of the four primary survey questions. For each of the
questions it was first determined which two color combinations had the greatest selection
proportions. The z-score was then calculated using those two proportions and, if the test
produced a significant result, the billboard with the greater proportion was found to have
96
been selected statistically more often than every other color combination. However, if the
test did not produce a significant result, there was no statistical difference between the
billboards in regards to public preference. At that point, the color combination with the
third highest proportion was tested against the billboard with the highest proportion
overall. This process would continue on in this manner until a significant result was
obtained from the test.
These comparisons were first made for the four primary survey questions for the
entire sample population. Then the demographics questions were used to separate the
participants into groups so comparisons could be made within the groups for each of the
primary questions. The participants were divided into groups based upon age, gender, and
driving experience. The comparisons conducted for the statistical analysis of the survey
results are described below:
“Which sign would make you more aware of Construction Work Zones?”

Responses for the total sample population

Responses within the six age groups

Responses by gender

Responses within the five groups by driving experience

Responses based upon reported knowledge of rules for driving in a work zone
“Which sign would make you more cautious in the Construction Work Zone?”

Responses for the total sample population

Responses within the six age groups

Responses by gender
97

Responses within the five groups by driving experience

Responses based upon reported knowledge of rules for driving in a work zone

Responses based upon reported number of crashes in work zones
“Which sign would you be able to see best during the day?”

Responses for the total sample population

Responses within the six age groups

Responses by gender

Responses within the five groups by driving experience
“Which sign would you be able to see best during the night?”

Responses for the total sample population

Responses within the six age groups

Responses by gender

Responses within the five groups by driving experience
5.2 Statistical Tests for Driving Simulator Study
The driving simulator study was designed to evaluate the impact of the billboards
on driver behavior. For the statistical analysis of the data collected in the simulator study,
it was determined that several comparisons needed to be made between the behavior
observed by the different billboard advertisements. As a result, the data was labeled by
the type of billboard presented to the driver so that comparisons could be made between
the signs to determine if there were any differences detected in driver behavior.
Since each of the three scenarios comprising the simulator study include at least
four billboard advertisements, the statistical tests used in the analysis needed to be able to
98
compare multiple groups. The t-test is a statistical test that is commonly used to compare
the means of two groups (41). If multiple t-tests were conducted between each of the
groups included in the different scenarios in the study, Type I errors would be
compounded. Type I errors are the detection of an effect in the sample population when
there is no effect present (41). This is represented by a value referred to as the α-level.
To minimize the chance of having a Type I error occur, the statistical tests were carried
out at a 95 percent level of confidence where the α-level is 0.05. Conducting multiple ttests would increase the probability for a Type 1 error to occur as the familywise error
rate would increase. The familywise error rate is calculated using the following equation:
(41)
Where: n = The number of tests being conducted on the data
For Scenario #2 there are four groups being tested while Scenarios #1 and #3 each
have 12 groups. If each of the groups were compared to one another, the number of tests
to be conducted would be six for Scenario #2 and 66 for Scenarios #1 and #3.
Subsequently, if all of those tests were conducted, the familywise error rate would be
0.26 for Scenario #2 and 0.97 for Scenarios #1 and #3. These rates are considered to be
extremely high and the results from the multiple tests would not be useful. As a result, the
one-way analysis of variance (ANOVA) was selected for the statistical analysis of the
data. The one-way ANOVA is able to compare the variances between several groups and
still control the Type I error by using the Bonferroni α, which is the α-level divided by
the number of tests (41).
99
The one-way ANOVA is a parametric test and, as such, the data used in the test
must meet four assumptions for the test to produce valid results. The assumptions are that
the data should be measured independently, the collected data must be interval data, there
should be homogeneity of variances in the data between the groups, and the data must be
normally distributed. For the first two assumptions, the design of the driving simulator
study ensures that those assumptions are met. Each participant was scheduled to
participate at separate times and only one person could drive the simulator at a time, so
the behavior of one participant would not have any effect on the behavior of another
participant. The data parameters collected in the study were collected over separate zones
with data logged at a rate of 60 times per second and recorded in units that quantified the
observed behaviors.
The homogeneity of variances for the data was evaluated by conducting Levene’s
test. The null hypothesis for this test stated that the sample variances between the groups
were equal. If the test produced a significant result, the variances were not homogeneous
and that assumption of the parametric tests would be violated. The final assumption that
the data should follow a normal distribution can be evaluated by plotting the data in a
histogram. This provides a visual evaluation of the data; however, statistical tests were
also desired to provide verification on conformance to normality. To check normality of
the data, the Kolmogorov-Smirnov and Shapiro-Wilk tests were conducted. These tests
each compare the sampled data to a normally distributed dataset that has an equivalent
mean and standard deviation (41). If the test produced a significant result, the sampled
data would not follow the normal distribution.
100
After Levene’s test was conducted on the data for each of the scenarios in the
study, it was determined that the data parameters of acceleration in the area prior to the
billboards, total fixation, and proportion of fixation duration did not have homogeneous
variances for Scenarios #1 and #2. Additionally for Scenario #1, acceleration in the
advanced warning zone and mean fixation did not have homogeneous variances.
However, all of the data parameters in Scenario #3 were determined to have
homogeneous variances.
The results for the Kolmogorov-Smirnov and Shapiro-Wilk tests indicated only
two parameters in Scenario #2 and one parameter in Scenario #3 had data that followed
the normal distribution. For Scenario #2 those parameters were acceleration in the
advanced warning zone and lane position in the area prior to the billboard. For Scenario
#3, the single parameter with normally distributed data was lane position.
Upon reviewing the histograms of the data, it was seen that the distribution of the
data was not drastically different from the normal distribution. Most of the distributions
had the issue of positive skew where the data tended to fall on the left side of the
distribution. This issue was most prominent in the glance behavior parameters across all
three scenarios. Those parameters will have an inherent positive skew due to how the
data is recorded and the nature of the data itself. As discussed in the review of previous
research on glance behavior, drivers tend to make short glances away from the roadway
as long glances are found to correspond to decreases in safety. Accordingly, the glance
behavior data in this research indicated that drivers most often made either no glances or
short glances, both of which corresponded to low values. Hence, the distributions of the
101
glance behavior data are practically centered on a value of zero and thereby sit on the
left-side of a normal distribution.
There are three approaches suggested by Field to correct problems with nonnormal distributions which are to remove outliers, to change the value of the outliers, or
to transform the data (41). The option of changing the values of the raw data was not
selected because it did not seem ethical to alter the values recorded even if they were
extreme values. If cases were determined to be outliers, they would be reviewed to
determine if they were truly unrepresentative of actual driving behaviors and thus needed
to be removed from the analysis. To find if there were outliers in the data, the values
were standardized to z-scores by the following equation:
(41)
Where: X = Each individual data points
= The mean of all scores
s = The standard deviation of all scores
After the data was converted to z-scores, it was analyzed to determine the extent
of outliers present in the data. Based on the z-scores, a normal distribution would have no
more than 5 percent with absolute values greater than 1.96, 1 percent with absolute values
greater than 2.58, and no absolute values greater than 3.29 (41). For Scenario #1, the
parameters of speed in the area prior to the billboard and mean fixation were found to
have the largest problem with outliers as the data points exceeding a z-score of 3.29 were
1.9 percent and 1.1 percent, respectively, of the 480 data points collected. Further review
of the data found that the outliers for speed were eight data points ranging from
102
approximately 81 to 90 mph and a single data point of approximately 45 mph. The five
outliers in mean fixation were found to range from 3.62 seconds to 5.08 seconds. These
speeds and mean fixations could be considered extreme values that would not likely
occur in actual driving conditions. However, the study was designed to isolate the driver
on the roadway to see the extent of their behavior. Hence, the magnitude of the variation
in the data above what would be expected to occur on actual roadways could be
considered logical as the actual conditions are not replicated perfectly. As a result, it was
determined that the outliers would be conditionally removed from the speed data to see if
it resulted in a normal distribution as the speed data had the most outliers. If the
normality of the dataset did not improve, the outliers would be left in the dataset.
After removing the identified outliers in the speed data for Scenario #1, the
Kolmogorov-Smirnov and Shapiro-Wilk tests were conducted on the new data sets.
These tests still indicated that the speed data did not follow a normal distribution despite
removing the nine outliers. To see if outliers may still be an issue the z-scores were
recalculated for the new data set and it was found that five data points exceeded the
critical z-score of 3.29. Those five data points were speeds ranging between 79 and 81
mph. Since the removal of the outliers did not improve the normality of the data, it was
selected that no data would be removed for any of the parameters.
Given the previous evaluation that most of the collected data does not following a
normal distribution, the method that was used to determine outliers may not be applicable
as it is based on a normal distribution. Furthermore, the purpose of this study is to
evaluate the potential safety benefits of the billboards based on the observed driver
103
behavior and any removal of the outliers would result in removing observed instances of
extreme driver behavior. These extreme behaviors are the actions that are recognized in
previous research to correspond with decreases in safety. Consequently, further attempts
to resolve the non-normal data were done through transforming the data.
In the review of previous research it was seen that Horrey and Wickens
implemented a log transformation to the glance behavior data in their study to address the
violations of normality that had occurred (25). This type of transformation was selected
by Horrey and Wickens to address the positive skew in their data. That same positive
skew problem was also seen in the glance behavior for this research. There are other
possible transformations that can be applied to datasets including taking either the square
root or reciprocal of the data points. However, for some of the parameters being studied
(i.e., lane position) there are data points that are either zero or negative values, so the
square root and reciprocal transformations would be inappropriate. The log
transformation did require adding a constant to the data points to account for the negative
values and zeroes.
To apply the log transformation to all of the datasets, a constant of six was added
to the lane position data while a constant of one was added to the other parameters. The
constant of six was selected for the lane position data as the smallest value that could be
recorded for lane position would be -6 as that would correspond to the vehicle being
centered on the far left edge of the travel lane. The log transformations were done using
the following equations:
For lane position:
(41)
104
For other parameters:
(41)
Where: x = Each of the data points
xtransformed = Transformed data point
After transforming all of the data sets, the Kolmogorov-Smirnov and ShapiroWilk tests were conducted to check normality and Levene’s test was also conducted to
check for homogeneous variances. For Scenario #1, only lane position in the pre-work
zone and advanced warning zones followed a normal distribution while the other
parameters still did not. Also three parameters still did not have homogeneous variances
in Scenario #1 as compared to five parameters in the untransformed data. For Scenario
#2, three parameters had normally distributed data in the transformed datasets, an
improvement of one parameter over the untransformed datasets. Similarly, the
transformed datasets only corrected one of the parameters which previously did not have
homogeneous variances. No improvements in normality of the data were seen for
Scenario #3 after applying the log transformations.
Overall, the log transformations were considered to be ineffective at correcting
the normality and homogeneous variance problems that occurred in the data sets. While
some improvements could be made, the collected data sets still violated the assumptions
of the parametric tests. Although Field does indicate that when the Kolmogorov-Smirnov
and Shapiro-Wilk tests are used on large sample sizes there is a possibility that small
deviations from normality can produce a significant result (41). Likewise, Levene’s test
can produce a false significant result when used on large sample sizes. Given that the
number of data points analyzed in these tests were 480, 120, and 348 for Scenarios #1,
105
#2, and #3, respectively, there is a possibility that the tests are indeed producing a false
significant result. As discussed previously, the visual check of the histograms across the
three scenarios did not show large variations from the normal distribution, so the data
may not be in violation of the assumptions for parametric tests.
Furthermore, Field states that the one-way ANOVA test is considered a robust
statistical test because it is able to withstand violations of the assumptions (41). If the
homogeneity of variances assumption is violated, ANOVA is considered to still be robust
for equal sample sizes. If sample sizes are not equal then Welch’s F-ratio can be used to
mitigate the issue. Glass, Peckham, and Sanders reviewed the robustness of the one-way
ANOVA with regards to violations of normality and found it to be very robust for
particular conditions (46). According to their findings, skewed populations are found to
have very little effect on the level of significance or power of the test. Upon consideration
of the issues surrounding the normality and homogenous variances assumptions and the
robustness of the one-way ANOVA test, it was selected to proceed with the one-way
ANOVA to conduct the statistical analysis of the data. After the one-way ANOVA was
conducted and the results of the test were obtained, questions regarding how the potential
violations had impacted the results were reviewed and addressed in Chapter 7.
The one-way ANOVA functions by comparing the amount of variability due to
the experimental manipulation, which would be the billboard advertisements in this
research, to the amount of variability due to the unsystematic variation, which can be
considered to be other factors within the experiment. The comparison is made by
calculating the mean squares of the model, MSM, and the mean squares of the residual,
106
MSR. The F-ratio is then calculated as the ratio of the mean squares of the model to the
mean squares of the residual. The equations used to calculate the F-ratio are as follows:
(41)
Where: SST = Total sum of squares
xi = Observed data point
xgrand = Grand mean
(41)
Where: SSM = Model sum of squares
nk = Number of participants within the group
xk = Mean of the group
With degrees of freedom:
(41)
Where: dfM = Degrees of freedom of the model
k = The number of groups
(41)
Where: SSR = Residual sum of squares
xik = Observed data point within the group
With degrees of freedom:
(41)
Where: dfR = Degrees of freedom of the residual
(41)
Where: MSM = Mean squares of the model
107
(41)
Where: MSR = Mean squares of the residual
(41)
Where: F = Ratio of the systematic variation and the unsystematic variation
After calculating the F-ratio it is compared to a critical F-ratio for the confidence
level (α) and the degrees of freedom (dfM and dfR) being used to conduct the test. If the
calculated F-ratio is found to exceed the critical F-ratio this is a significant result
indicating that the null hypothesis is rejected. In the one-way ANOVA test, the null
hypothesis stated that the groups being evaluated have equal means. The null hypothesis
is rejected when at least one of the groups has a mean which does not equal the means of
the other groups. The extent to which groups have means that are unequal to the other
groups is unknown from simply conducting the one-way ANOVA test. Thus, to conclude
which groups are different either a planned comparison or a post hoc test must be
conducted. Planned comparisons and post hoc tests are similar to whether a one- or twotailed test is conducted (41). If a specific hypothesis is formulated prior to conducting the
experiment, as for a one-tailed test, then planned comparisons are performed. However, if
no specific hypothesis is formulated, as for a two-tailed test, post hoc tests are used.
In this research there were no specific hypotheses pre-determined on how the
billboards would affect driver behavior prior to conducting the study. Accordingly, post
hoc tests will be used to check for differences between the groups. Several post hoc tests
can be used after conducting a one-way ANOVA test, but the tests each perform better
under certain conditions. Based on the potential issues with normality and homogeneous
108
variances in the datasets, a post hoc test which can withstand these issues is preferred.
The Games-Howell test performs well under those two issues along with performing well
when sample sizes are unequal (41).
Once the significant differences are identified, the effect size can be calculated to
indicate the impact of the experimental manipulation. Effect size, r, is defined by Cohen
to correlate to either a small (r = 0.10), medium (r = 0.30), or large effect (r = 0.50) (41).
As the effect size increases, the experimental manipulation is considered to have caused
more of the variation seen between the groups. For the results of the one-way ANOVA
tests, effect size was calculated using the following equation:
(41)
While conducting the one-way ANOVA test, the Welch’s F-ratio was also
determined to correct for violations of the homogeneous variances assumption. Welch’s
F-ratio is able to account for those violations as it adjusts the residual degrees of freedom
to produce a revised F-ratio (41). Also the model sum of squares and mean squares of the
model are adjusted due to a weighting factor for each group that is included in the
equations. The following equations were used to calculate the Welch’s F-Ratio:
(41)
Where: wk = Weighting factor for the group
sk = Variance of the group
(41)
Where: xgrand(Welch) = Weighted grand mean
(41)
109
Where: SSM(Welch) = Weighted model sum of squares
(41)
Where: MSM(Welch) = Adjusted mean squares of the model
(41)
Where: Λ = Weighting factor
With degrees of freedom:
(41)
(41)
Where: FW = Welch’s F-ratio
As for the statistical analysis of the billboard survey data, all of the statistical tests
on the data from the driving simulator study were conducted at a 95 percent level of
confidence (α = 0.05). For the statistical analysis of the data collected from the driving
simulator study, the following comparisons were conducted using the one-way ANOVA
test:
Scenario #1:

Compared the 12 billboard slogans with the control billboard in the Pre-Work
Zone for the following parameters: speed, lane position, variation in lane position,
acceleration, mean fixation, total fixation, number of fixations, and proportion of
fixation duration

Compared the 12 billboard slogans with the control billboard in the Advanced
Warning Area for the following parameters: speed, lane position, and acceleration
110

Compared the 12 billboard slogans with the control billboard in the Work Zone
for the following parameters: speed, lane position, and acceleration
Scenario #2:

Compared the 3 billboard graphic combinations with the control billboard in the
Pre-Work Zone for the following parameters: speed, lane position, variation in
lane position, acceleration, mean fixation, total fixation, number of fixations, and
proportion of fixation duration

Compared the 3 billboard graphic combinations with the control billboard in the
Advanced Warning Area for the following parameters: speed, lane position, and
acceleration

Compared the 3 billboard graphic combinations with the control billboard in the
Work Zone for the following parameters: speed, lane position, and acceleration
Scenario #3:

Compared the 12 billboard placements in the area leading up to the billboard for
the following parameters: speed, lane position, variation in lane position,
acceleration, mean fixation, total fixation, number of fixations, and proportion of
fixation duration
Impact of Long Glances on Safety in All Three Scenarios:

Compared the total variation in lane position as participants were fixating on the
billboards for both long glances (≥ 2.0 seconds) and short glances (< 2.0 seconds)
111
CHAPTER 6: RESULTS
The billboard advertisements being evaluated in this research were designed to be
the main component of a work zone safety campaign. Different elements of those
billboard advertisements were evaluated by conducting a public survey study and a
driving simulator study. In each of the studies, the primary objective was to determine to
what extent the billboards were able to have an effect on driver behavior. For the public
survey study, the evaluation was to ascertain which background and text color
combination was preferred by the public in regards to communicating the safety message.
During the driving simulator study, driving performance and glance behaviors of the
participants were observed in a controlled environment. Those observed behaviors were
then analyzed to compare the different billboards shown to drivers. When examining the
effects on driver behavior in both studies, considerations were given to how the specific
behaviors would affect safety on the roadway. The results of the public survey and
driving simulator studies are described in the following sections.
6.1 Public Survey Study
Upon completion of the public survey study, a total of 376 responses had been
obtained. Completed survey responses were obtained from 175 females (47 percent) and
197 males (52 percent). There were four surveys where the gender question had not been
answered. Survey responses were also totaled for the following age groups: 16-24, 25-34,
35-44, 45-54, 55-64, and 65 years and older. These age groups were selected due to their
use by the National Highway Traffic Safety Administration in analyzing crash data by
demographic groups (1). The number of responses collected in each of the six age groups
112
is shown in Figure 6. When conducting the survey study, researchers attempted to match
the demographics of the survey sample population to that of the general driving
population in terms of gender and age. However, since participants were recruited on a
voluntary basis and due to time constraints, an exact match of the demographics was not
feasible.
Number of Responses
250
200
150
100
50
0
16-24
25-34
35-44
45-54
55-64
65+
Age Group
Figure 6. Public survey responses by age group.
As seen in Figure 6, the sample population had a majority of participants (53
percent) in the 16-24 age group. From the 2010 Census, the age group of 15-24 years old
constituted 17.1 percent of Ohio’s population eligible for a driver’s license (47).
Comparatively, in the 2010 Census the 45-64 age group made up a plurality of Ohio’s
population eligible for a driver’s license at 34.4 percent, whereas that same age group
comprised 16.8 percent of the population in this study (47). Based on this information,
younger drivers were considered to be overrepresented in the public survey study. As
seen in previous research, younger drivers are identified as a high risk group since they
113
are involved in crashes at a higher rate than other age groups. In 1996, the crash
involvement rate for drivers 16 to 24 years old was 1,222 crashes per 100 million VMT
which more than doubled the other age groups as those 25-44, 45-64 and 65 years and
older had rates of 532, 405, and 529 crashes per 100 million VMT, respectively (48).
Thus, the overrepresentation of younger drivers in this study was considered to be
beneficial to this research. Any positive impacts to safety found would be based largely
on a high risk group of drivers and could possibly have a direct impact on addressing
roadway safety.
When the survey was developed, there was a question included which asked
participants how many crashes they had been involved in while driving in a construction
work zone. During the analysis of the survey responses, it was determined that there were
24 participants that responded they had been involved in at least one work zone crash. Of
those participants, 18 had reportedly been involved in at least one crash, four had been in
two crashes, and one participant had been in three crashes. Therefore, there were not
adequate responses for an analysis to be conducted on the number of crashes in which a
driver had been involved. The number of responses were too small to meet the
assumptions of the two-proportion z-test as outlined in Chapter 5. As a result, statistical
analyses of the survey responses were not conducted for the intended comparisons on
number of crashes.
One of the goals of the public survey study was to determine if drivers considered
the billboard originally used in the work zone safety campaign to be cluttered. The survey
responses showed that approximately 60 percent of the participants thought the billboards
114
used in the survey were too cluttered. The other goal of the study was to determine which
background and text color combination would be preferred by drivers for communicating
the work zone safety campaign message. This determination was made using the four
primary questions included in the survey about the color combinations and comparing the
responses within individual demographic groups.
6.1.1 Increase Awareness of Work Zones
The first primary question was in regards to increasing a driver’s awareness of the
presence of construction work zones on the roadways. Participants were asked to select
the sign which they believed would make them more aware of work zones. The selection
proportions for the eight billboards are shown in Figure 7 and the results of the two-
Color Combination
proportion z-test for the total sample population are shown in Table 12.
Red with Black
Blue with Black
Red with White
Yellow-Green with Black
Yellow with Black
Orange with Black
Green with White
Black with Orange
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70
Proportion of Responses
Figure 7. Proportions for the total population on increased awareness of work zones.
115
Table 12: Results of z-test for total population on increased awareness of work zones
Color Combinations
Group
P
SE
z-statistic
Result
Compared
Reject Null:
Total
Yellow-Green and Yellow 0.620 0.025
2.68
pyellow-green ≠ pyellow
The statistical test produced a significant result which means that participants
selected the fluorescent yellow-green background with black text statistically more often
than the yellow background with black text as the proportions were 0.65 and 0.58,
respectively. Comparisons were then made for the responses within the different
demographic groups to see if there were any variations from the billboard preferred by
the total population. Those z-test results are summarized in Tables 13, 14, and 15.
Table 13: Results of z-test by age group for increased awareness of work zones
Age
Color Combinations
P
SE
z-statistic
Result
Group
Compared
Accept Null:
Yellow-Green and Yellow 0.619 0.035
1.87
pyellow-green = pyellow
16-24
Reject Null:
Yellow-Green and Orange 0.500 0.036
10.24
pyellow-green ≠ porange
Reject Null:
25-34 Yellow-Green and Yellow 0.636 0.065
3.29
pyellow-green ≠ pyellow
Reject Null:
35-44 Yellow-Green and Yellow 0.587 0.073
2.02
pyellow-green ≠ pyellow
Accept Null:
45-54 Yellow-Green and Yellow 0.586 0.091
0.72
pyellow-green = pyellow
Reject Null:
55-64 Yellow and Yellow-Green 0.588 0.084
3.55
pyellow ≠ pyellow-green
Reject Null:
65+
Yellow-Green and Yellow 0.846 0.100
2.50
pyellow-green ≠ pyellow
116
Table 14: Results of z-test by gender for increased awareness of work zones
Color Combinations
Gender
P
SE
z-statistic
Result
Compared
Accept Null:
Yellow-Green and Yellow 0.607 0.035
0.84
pyellow-green = pyellow
Male
Yellow-Green and Red
Reject Null:
0.444 0.034
10.41
with White
pyellow-green ≠ pred(wh.)
Reject Null:
Female Yellow-Green and Yellow 0.628 0.037
3.78
pyellow-green ≠ pyellow
Table 15: Results of z-test by driving experience for increased awareness of work zones
Years of
Color Combinations
zP
SE
Result
Experience
Compared
statistic
Yellow-Green and
Accept Null:
0.615 0.042
1.32
Yellow
pyellow-green = pyellow
1-5 years
Yellow-Green and
Reject Null:
0.504 0.043
8.26
Orange
pyellow-green ≠ porange
Yellow-Green and
Reject Null:
6-10 years
0.597 0.060
3.69
Yellow
pyellow-green ≠ pyellow
Yellow-Green and
Reject Null:
11-15 years
0.636 0.084
3.58
Yellow
pyellow-green ≠ pyellow
Yellow-Green and
Accept Null:
0.500 0.112
0.90
Yellow
pyellow-green = pyellow
16-20 years
Yellow-Green and
Accept Null:
0.500 0.112
0.90
Orange
pyellow-green = porange
Yellow and YellowAccept Null:
0.642 0.047
1.48
Green
pyellow = pyellow-green
20+ years
Yellow and Red with
Reject Null:
0.403 0.048
10.278
White
pyellow ≠ pred(wh.)
From Table 13, the age groups of 16-24 and 45-54 were found to produce a
different color combination preference than that of the overall population. Each of these
groups selected the fluorescent yellow-green and yellow backgrounds with black text at
equal proportions and statistically more often than the other six color combinations.
Participants in the 55-64 age group also contrasted from the overall population as they
117
were found to prefer the yellow background with black text over the fluorescent yellowgreen background with black text. Further statistical analyses could not be done for the
45-54 age group after the first two color combinations as no other combination garnered
at least five responses as required by the assumptions of the test.
From Table 14, males were found to have selected the color combinations with
yellow and fluorescent yellow-green backgrounds at statistically equal rates while
females displayed the same preference as that of the overall population. In Table 15, there
were a few variations seen from the overall population as participants with 1-5 and 20+
years of driving experience selected the yellow-green and yellow backgrounds with black
text equally. Moreover, participants with 16-20 years of driving experience equally
preferred the color combinations using the fluorescent yellow-green, yellow, and orange
backgrounds. No additional color combinations could be tested as they did not elicit at
least five responses.
When analyzing which billboard was preferred for increasing awareness of work
zones, the results were also evaluated by the participant’s self-reported knowledge of the
rules and regulations for driving in a work zone. This comparison was considered to be
important as those individuals who do not have knowledge of the rules were more likely
to violate those rules. If the billboard was able to communicate the safety message to
these drivers and make them more aware of construction work zones, they may be more
likely to learn and comply with the rules.
For the question on knowledge of rules, participants selected a response on a scale
from strongly agree to strongly disagree. The participants who answered either neutral or
118
disagree made up 11 percent of the overall population while the other 89 percent
answered either strongly agree or agree. The two-proportion z-test was carried out for
those who did not claim to have knowledge of the rules (neutral or disagree) to determine
their preferred color combination. Results of the statistical analysis, shown in Table 16,
indicate that those drivers who reported not having knowledge of the rules for driving in
work zones selected the fluorescent yellow-green and yellow backgrounds with black text
statistically more often than the other combinations.
Table 16: Results of z-test by knowledge of rules for increased awareness of work zones
Color Combinations
zGroup
P
SE
Result
Compared
statistic
Accept Null:
Yellow-Green and Yellow
0.750 0.068
0.74
Neutral
pyellow-green = pyellow
and
Reject Null:
Disagree Yellow-Green and Orange
0.550 0.079
6.29
pyellow-green ≠ porange
6.1.2 Increase Caution in Work Zones
The second primary question asked participants to select the sign which they
believed would make them more cautious in work zones. Figure 8 displays the selection
proportions for the eight color combinations while the results of the two-proportion z-test
for the total sample population are shown in Table 17.
Color Combination
119
Red with Black
Blue with Black
Red with White
Yellow-Green with Black
Yellow with Black
Orange with Black
Green with White
Black with Orange
0.00
0.10
0.20
0.30
0.40
0.50
0.60
Proportion of Responses
Figure 8. Proportions for the total population on increased caution in work zones.
Table 17: Results of z-test for total population on increased caution in work zones
Color Combinations
Group
P
SE
z-statistic
Result
Compared
Accept Null:
Yellow-Green and Yellow 0.519 0.026
1.46
pyellow-green = pyellow
Total
Reject Null:
Yellow-Green and Orange 0.497 0.026
3.45
pyellow-green ≠ porange
The results in Table 17 indicate that the participants selected the fluorescent
yellow-green and yellow backgrounds with black text statistically more often than the
other color combinations for making them more cautious in work zones. From Figure 8,
the proportions were 0.53 and 0.50 for those two combinations whereas the orange
background with black text had a proportion of 0.45. Additional statistical analyses were
then conducted based upon the demographic groups to see if there were any variations
from the color combination preferred by the total population. The results of the z-tests by
age, gender, and driving experience are summarized in Tables 18, 19, and 20,
respectively.
120
Table 18: Results of z-test by age group for increased caution in work zones
Age
Color Combinations
P
SE
z-statistic
Result
Group
Compared
Accept Null:
Yellow-Green and Orange 0.513 0.036
1.19
pyellow-green = porange
Accept Null:
16-24 Yellow-Green and Yellow 0.508 0.036
1.56
pyellow-green = pyellow
Yellow-Green and Red
Reject Null:
0.416 0.032
7.10
with White
pyellow-green ≠ pred(wh)
Accept Null:
Yellow-Green and Yellow 0.545 0.067
0.70
pyellow-green = pyellow
25-34
Reject Null:
Yellow-Green and Orange 0.491 0.067
2.47
pyellow-green ≠ porange
Accept Null:
Yellow-Green and Yellow 0.500 0.074
1.78
pyellow-green = pyellow
35-44
Reject Null:
Yellow-Green and Orange 0.478 0.074
2.33
pyellow-green ≠ porange
Reject Null:
45-54 Orange and Yellow-Green 0.586 0.091
2.49
porange ≠ pyellow-green
Reject Null:
55-64 Yellow and Yellow-Green 0.529 0.086
4.48
pyellow ≠ pyellow-green
Reject Null:
65+
Yellow-Green and Yellow 0.846 0.100
2.50
pyellow-green ≠ pyellow
Table 19: Results of z-test by gender for increased caution in work zones
Color Combinations
Gender
P
SE
z-statistic
Result
Compared
Accept Null:
Yellow-Green and Yellow 0.508 0.036
0.96
pyellow-green = pyellow
Male
Reject Null:
Yellow-Green and Orange 0.487 0.036
2.24
pyellow-green ≠ porange
Accept Null:
Yellow-Green and Yellow 0.529 0.038
1.53
pyellow-green = pyellow
Female
Reject Null:
Yellow-Green and Orange 0.517 0.038
2.37
pyellow-green ≠ porange
121
Table 20: Results of z-test by driving experience for increased caution in work zones
Years of
Color Combinations
zP
SE
Result
Experience
Compared
statistic
Yellow-Green and
Accept Null:
0.537 0.043
1.36
Orange
pyellow-green = porange
Yellow-Green and
Accept Null:
1-5 years
0.529 0.043
1.83
Yellow
pyellow-green = pyellow
Yellow-Green and Red
Reject Null:
0.436 0.038
6.46
with White
pyellow-green ≠ pred(wh.)
Orange and YellowAccept Null:
0.493 0.061
1.23
Green
porange = pyellow-green
6-10 years
Reject Null:
Orange and Yellow
0.442 0.069
2.79
porange ≠ pyellow
Yellow-Green and
Accept Null:
0.576 0.086
0.52
Yellow
pyellow-green = pyellow
11-15 years
Yellow-Green and
Reject Null:
0.455 0.087
3.08
Orange
pyellow-green ≠ porange
Yellow-Green and
Accept Null:
0.600 0.110
1.11
Yellow
pyellow-green = pyellow
16-20 years
Yellow-Green and
Accept Null:
0.550 0.111
1.91
Orange
pyellow-green = porange
Yellow and YellowAccept Null:
0.547 0.048
0.74
Green
pyellow = pyellow-green
20+ years
Reject Null:
Yellow and Orange
0.466 0.053
3.85
pyellow ≠ porange
Table 18 shows that the 16-24, 25-34, and 35-44 age groups displayed the same
preferences as the overall population in selecting the fluorescent yellow-green and yellow
backgrounds with black text equally. However, the 16-24 age group also selected the
orange background with black text as statistically often as the aforementioned color
combinations. In contrast to the overall population, the age groups of 45-54, 55-64, and
65 and older preferred just one color combination. The 45-54 age group differed the most
from the overall population as they selected the orange background with black text
statistically more often. Those participants in the 55-64 age group preferred the yellow
122
background with black text whereas those in the 65 and older group preferred the
fluorescent yellow-green background.
From Table 19, neither gender varied from the two color combinations preferred
by the overall population. Also those participants with 11-15 and 20+ years of driving
experience selected the same combinations as the overall population as seen in Table 20.
However, that was not the case for participants with 6-10 years of driving experience.
Those drivers selected the orange and fluorescent yellow-green backgrounds with black
text statistically more often than the other billboards. Participants with 1-5 and 16-20
years of experience preferred the orange background with black text along with the color
combinations preferred by the overall population. As in the analysis of the previous
question, no further tests could be conducted in the 16-20 age group as the remaining
color combinations did not meet the minimum of five responses.
During the analysis on which billboard would increase caution, it was also
selected to evaluate the responses by self-reported knowledge of the rules for driving in a
work zone as done for the previous question. Those participants who did not report
knowing the rules for driving in a work zone may not be driving cautiously throughout
work zones. The two-proportion z-test was conducted on the responses for those
answering either neutral or disagree and the results are shown in Table 21. Those
participants were found to have equally selected the yellow, orange, and fluorescent
yellow-green backgrounds with black text and the red background with white text. No
further tests could be done within that group due to violating the minimum responses.
123
Table 21: Results of z-test by knowledge of rules for increased awareness of work zones
Color Combinations
zGroup
P
SE
Result
Compared
statistic
Accept Null:
Yellow and Orange
0.500 0.083
0.00
pyellow = porange
Neutral
Accept Null:
and
Yellow and Yellow-Green
0.450 0.079
1.16
pyellow = pyellow-green
Disagree
Accept Null:
Yellow and Red with White
0.450 0.079
1.16
pyellow = pred(wh.)
6.1.3 Billboard Seen Best in Daylight
The third primary question asked participants to select the sign which they would
be able to see the best in daylight while driving along a freeway at 65 mph. Figure 9
displays the selection proportions for the eight color combinations and the results of the
Color Combination
two-proportion z-test for the total sample population are shown in Table 22.
Red with Black
Blue with Black
Red with White
Yellow-Green with Black
Yellow with Black
Orange with Black
Green with White
Black with Orange
0.00
0.10
0.20
0.30
0.40
0.50
0.60
Proportion of Responses
Figure 9. Proportions for the total population on billboard seen best in daylight.
124
Table 22: Results of z-test for total population on billboard seen best in daylight
Color Combinations
Group
P
SE
z-statistic
Result
Compared
Reject Null:
Total
Yellow-Green and Yellow 0.504 0.026
2.61
pyellow-green ≠ pyellow
The results in Figure 9 and Table 22 indicate that at a proportion of 0.53 the
participants selected the fluorescent yellow-green color combination statistically more
often than the yellow color combination (proportion of 0.46) at a 95 percent level of
confidence. Thus, the public perception is that a billboard using a fluorescent yellowgreen background and black text would be seen the best during the day. To determine if
there were any variations in the preferred color combination from that of the total
population, additional statistical analyses were conducted by age, gender, and driving
experience as summarized in Tables 23, 24, and 25, respectively.
Table 23: Results of z-test by age group for billboard seen best in daylight
Age
Color Combinations
zP
SE
Result
Group
Compared
statistic
Reject Null:
16-24 Yellow-Green and Yellow
0.480 0.036
2.08
pyellow-green ≠ pyellow
Accept Null:
Yellow and Yellow-Green
0.509 0.067
1.38
pyellow = pyellow-green
25-34
Reject Null:
Yellow and Red with White 0.455 0.067
2.88
pyellow ≠ pred(wh.)
Reject Null:
35-44 Yellow-Green and Yellow
0.478 0.074
2.33
pyellow-green ≠ pyellow
Reject Null:
45-54 Yellow-Green and Orange
0.586 0.091
2.32
pyellow-green ≠ porange
Accept Null:
Yellow and Yellow-Green
0.588 0.084
1.88
pyellow = pyellow-green
55-64
Reject Null:
Yellow and Red with Black 0.353 0.082
6.00
pyellow ≠ pred(bl.)
125
Table 24: Results of z-test by gender for billboard seen best in daylight
Color Combinations
Gender
P
SE
z-statistic
Result
Compared
Reject Null:
Male
Yellow-Green and Yellow 0.503 0.036
2.43
pyellow-green ≠ pyellow
Accept Null:
Yellow-Green and Yellow 0.509 0.038
1.34
pyellow-green = pyellow
Female
Reject Null:
Yellow-Green and Orange 0.451 0.038
5.55
pyellow-green ≠ porange
Table 25: Results of z-test by driving experience for billboard seen best in daylight
Years of
Color Combinations
zP
SE
Result
Experience
Compared
statistic
Yellow-Green and
Reject Null:
1-5 years
0.449 0.043
2.54
Yellow
pyellow-green ≠ pyellow
Yellow-Green and
Accept Null:
0.537 0.061
1.00
Yellow
pyellow-green = pyellow
6-10 years
Yellow-Green and
Reject Null:
0.463 0.061
4.16
Orange
pyellow-green ≠ porange
Yellow and Red with
Accept Null:
0.515 0.087
1.02
White
pyellow = pred(wh.)
11-15 years
Yellow and YellowReject Null:
0.455 0.087
2.56
Green
pyellow ≠ pyellow-green
Yellow-Green and
Accept Null:
16-20 years
0.550 0.111
1.91
Yellow
pyellow-green = pyellow
Yellow-Green and
Reject Null:
20+ years
0.557 0.048
2.01
Yellow
pyellow-green ≠ pyellow
In Table 23 it can be seen that the age groups of 16-24, 35-44, and 45-54 matched
the color combination preference of the total population. Differing from the total
population were the age groups of 25-34 and 55-64 as they equally selected the yellow
and fluorescent yellow-green backgrounds with black text for the billboard seen best
126
during the daylight. Statistical analyses could not be conducted on the responses of the 65
years and older group as only one color combination received at least five responses for
being seen best during the day.
From Table 24, males were found to have the same preference as the overall
population while females preferred both the fluorescent yellow-green and yellow
backgrounds with black text at a 95 percent level of confidence. Variations from the color
combination preferred by the total population were noticed for three groups in Table 25.
Drivers with 6-10 and 16-20 years of driving experience equally selected the fluorescent
yellow-green and yellow backgrounds with black text as seen best during the day.
However, no further comparisons could be conducted within the group with 16-20 years
of experience as none of the other color combinations received at least five responses. At
a 95 percent level of confidence, participants with 11-15 years of driving experience were
found to prefer both the yellow background with black text and the red background with
white text.
6.1.3 Billboard Seen Best at Nighttime
The fourth primary question asked participants to select the sign which they
would be able to see the best at nighttime while driving along a freeway at 65 mph. The
selection proportions for each of the eight color combinations are displayed in Figure 10.
Results of the two-proportion z-test for the total sample population are shown in Table
26.
Color Combination
127
Red with Black
Blue with Black
Red with White
Yellow-Green with Black
Yellow with Black
Orange with Black
Green with White
Black with Orange
0.00
0.20
0.40
0.60
0.80
1.00
Proportion of Responses
Figure 10. Proportions for the total population on billboard seen best at nighttime.
Table 26: Results of z-test for total population on billboard seen best at nighttime
Color Combinations
Group
P
SE
z-statistic
Result
Compared
Accept Null:
Yellow and Yellow-Green 0.810 0.020
0.92
pyellow = pyellow-green
Total
Yellow and Red with
Reject Null:
0.421 0.026
26.87
White
pyellow ≠ pred(wh.)
Figure 10 illustrates that participants selected the fluorescent yellow-green and
yellow backgrounds with black text at proportions of 0.80 and 0.82, respectively, from
the two sets of color combinations. Those color combinations were selected more often
than the other color combinations at a 95 percent level of confidence as shown in Table
26. Thus, the public perception is that the fluorescent yellow-green and yellow
backgrounds would be seen the best during the night. Additional statistical analyses were
then conducted by classifying the responses within different demographic groups. This
was done to determine if there were any variations in the preferred color combinations
128
from those of the total population. The results of the two proportion z-tests by age,
gender, and driving experience are summarized in Tables 27, 28, and 29, respectively.
Table 27: Results of z-test by age group for billboard seen best at nighttime
Age
Color Combinations
zP
SE
Result
Group
Compared
statistic
Accept Null:
Yellow and Yellow-Green
0.811 0.028
1.94
pyellow = pyellow-green
16-24
Reject Null:
Yellow and Red with White 0.434 0.035
19.10
pyellow ≠ pred(wh.)
Reject Null:
25-34 Yellow-Green and Yellow
0.818 0.052
2.05
pyellow-green ≠ pyellow
Accept Null:
35-44 Yellow-Green and Yellow
0.761 0.063
1.42
pyellow-green = pyellow
Reject Null:
45-54 Yellow and Yellow-Green
0.897 0.057
2.95
pyellow ≠ pyellow-green
Accept Null:
55-64 Yellow and Yellow-Green
0.765 0.073
0.69
pyellow = pyellow-green
Table 28: Results of z-test by gender for billboard seen best at nighttime
Color Combinations
Gender
P
SE
z-statistic
Result
Compared
Accept Null:
Yellow and Yellow-Green 0.812 0.028
1.87
pyellow = pyellow-green
Male
Yellow and Red with
Reject Null:
0.462 0.036
19.28
White
pyellow ≠ pred(wh.)
Accept Null:
Yellow-Green and Yellow 0.803 0.030
1.17
pyellow-green = pyellow
Female
Yellow-Green and Red
Reject Null:
0.358 0.036
18.41
with White
pyellow-green ≠ pred(wh.)
129
Table 29: Results of z-test by driving experience for billboard seen best at nighttime
Years of
Color Combinations
zP
SE
Result
Experience
Compared
statistic
Yellow and YellowAccept Null:
0.809 0.034
0.70
Green
pyellow = pyellow-green
1-5 years
Yellow and Red with
Reject Null:
0.412 0.042
15.61
White
pyellow ≠ pred(wh.)
Yellow and YellowReject Null:
6-10 years
0.821 0.047
2.22
Green
pyellow ≠ pyellow-green
Yellow-Green and
Reject Null:
11-15 years
0.727 0.078
3.30
Yellow
pyellow-green ≠ pyellow
Yellow-Green and
Reject Null:
16-20 years
0.850 0.080
3.42
Yellow
pyellow-green ≠ pyellow
Yellow and YellowReject Null:
20+ years
0.811 0.038
2.35
Green
pyellow ≠ pyellow-green
Table 27 shows that the 16-24 age group has the same color combination
preference as the overall population at a 95 percent level of confidence. The 35-44 and
55-64 age groups also selected the fluorescent yellow-green and yellow backgrounds
with black text more than the other color combinations, but further tests could not be
conducted on the remaining color combinations as they did not garner enough responses.
Participants in the 25-34 age group selected the fluorescent yellow-green background
with black text statistically more often while those in the 45-54 selected the yellow
background with black text statistically more often. Statistical analyses could not be
conducted on the responses of the 65 years and older group as only one color
combination received at least five responses for being seen best during at night.
From Table 28, there were no variations from the overall population found in the
color combinations preferred by both males and females. Additionally, no variations from
the overall population were found for drivers with 1-5 years of driving experience based
130
on the results in Table 29. However, the other four driving experience groups each
preferred one of the two color combinations preferred by the overall population.
Participants with 6-10 years and those with over 20 years of experience each selected the
yellow background with black text more often at a 95 percent level of confidence. Those
participants with either 11-15 or 16-20 years of driving experience were found to select
the fluorescent yellow-green background with black text statistically more often than the
other color combinations.
6.1.4 Overview of the Statistical Analyses of Survey Responses
The statistical analyses conducted on the survey responses largely indicated that
the public preferred two color combinations over the rest for each of the four primary
questions. The preferred combinations were the fluorescent yellow-green and yellow
backgrounds with black text. These two color combinations were preferred by the overall
population for increasing caution in work zones and for being seen best during the night.
For increasing awareness of work zones and being seen best during the day, the
fluorescent yellow-green background with black text was preferred. During the analyses
by different demographic groups, the two color combinations were selected equally in
over half of the cases analyzed. Based on these results and considering the meanings
assigned to those colors as defined in the MUTCD, the yellow background with black
text was selected for use in the billboard advertisements developed for the driving
simulator study. The yellow background is believed to be more consistent with the
purpose of the safety campaign and better maintains driver expectancy. Additional
discussion on this decision can be found in Section 7.1.
131
6.2 Driving Simulator Study
Based on the demographics of the sample population from the public survey
study, it was decided that the participants recruited for the driving simulator study would
be focused on the younger driver population. The overrepresentation of younger drivers
would be beneficial to the simulator study as the driving behaviors typically displayed by
this group are not as developed as those of more experienced drivers. This can often
result in younger drivers exhibiting behaviors that are considered to be unsafe. If
participants are observed to drive in a safer manner through the work zones after having
seen the billboard advertisements, then the signs can be considered to potentially have an
effect on the general driving population.
Participants were recruited primarily on the college campus in order to meet the
criterion of using younger drivers in the study. The resulting population recruited for the
driving simulator study was a total of 71 participants. This is larger than predetermined
sample size of 70 because one participant was unable to complete the study due to
technical difficulties with the driving simulator, so an additional driver was recruited.
Within the sample population, 48 participants were between the ages of 18-20 (67
percent) whereas 22 drivers were between the ages of 21-25 (31 percent). One participant
did not provide a response to the age question on the pre-test questionnaire. Additionally,
the population consisted of 47 males (66 percent) and 24 females (34 percent). The
plurality of participants drove a total of 0.5-1.0 hours and between 5-10 miles in one
direction during a typical day (42 and 32 percent, respectively). As seen in the public
survey study, the majority of drivers reported having knowledge of the rules and
132
regulations for driving in a construction work zone as 66 percent agreed with the
statement and 24 percent strongly agreed.
The pre-test questionnaires given to participants also included a couple of
questions on safety campaigns to assess the experiences and beliefs of the participants.
The first question was whether the driver knew any safety campaigns and if they could
recall the slogan that was used. This question was included to examine whether previous
safety campaigns had been able to reach the drivers in regards to exposure. Most of the
participants (62 percent) reported not knowing any safety campaigns. Of the 27 drivers
who did report knowing a safety campaign, 18 listed slogans for seat belt campaigns with
the primary response being the “Click it or Ticket it” slogan. There were six participants
who listed slogans used in work zone safety campaigns including the slogans of “Give
‘em a Brake” and “Slow Down, My Daddy Works Here” that were used in this study.
The other question included on the pre-test questionnaire asked drivers if a safety
campaign informing them on the hazards in a work zone would cause them to alter their
driving behavior and, if so, presented them with multiple options on how their behavior
would change. Approximately 83 percent of participants responded that the campaign
would cause a change in their behavior. For those participants who would change their
behavior, 88 percent indicated that they would reduce their speed, 85 percent would pay
closer attention, and 69 percent would drive farther away from the work zone. There were
no participants who indicated they would drive faster or drive closer to the work zones.
The driving simulator study was designed to have participants see a variety of
billboard advertisements and collect data on the driving performance and glance
133
behaviors observed. The driving performance measures collected were speed, lane
position, acceleration, deceleration, and collisions. After processing the collected data it
was seen that no collisions occurred during the entire study. Additionally, the data
collected on deceleration showed that participants very rarely used the brake pedal across
the three data collection zones. Overall, drivers were seen in these areas to reduce their
speed by reducing the amount of pressure applied to the accelerator pedal.
Since the acceleration data is a measure of how much pressure is applied to the
pedal, in terms of safety, this data is best suited for observing a driver’s response to a
particular event. For example, acceleration data would be useful if a researcher was
interested in how long it takes a driver to start accelerating once a traffic signal turns
green or the time it takes a driver to respond to another vehicle suddenly pulling out in
front of them. In each of these cases the data can be examined against time to evaluate the
driver’s response. For this study acceleration data was analyzed for a small sample of
participants as they approached the billboards to determine if their acceleration data
varied in response to seeing the signs. However no trends were seen across the sampled
data, so no further attempts were made to analyze the data. Based upon this finding, it
was decided to remove acceleration data from the statistical analyses of the billboards.
Consequently, the only driving performance parameters reported here are speed,
lane position, and variation in lane position. The glance behavior parameters reported in
the following sections are mean fixation, total fixation, number of fixations, and
proportion of fixation duration. Multiple comparisons of the billboard advertisements
were made based on these parameters to determine their effect on driver behavior. The
134
results for the three scenarios in the driving simulator study are presented in separate
sections below as each of the scenarios was designed to evaluate a different component of
the billboards. Following those sections is an analysis of driving performance while
participants looked at the signs in all three scenarios.
6.2.1 Scenario #1 – Slogans
The first scenario in the driving simulator study was designed to test the slogans
used in the billboard advertisements. Participants were presented with either a control
billboard or a work zone safety campaign billboard prior to driving through a
construction work zone. The campaign billboards used in this scenario had one of the 12
campaign slogans being tested, while the graphics and color combination of the sign as
well as the placement and orientation with respect to the driver were kept constant. As
participants drove through the simulated environment data was continuously collected on
the driving performance and glance behavior parameters outlined previously.
During the processing of the data, mean values for the parameters were taken
across the area prior to the billboards (Pre-Work Zone), the area directly prior to the work
zone (Advanced Warning Area), and the work zone area denoted by the construction
barrels (Work Zone). The only parameter which was not processed as a mean value was
variation in lane position as this is measured by the total lateral distance the vehicle
travels throughout the Pre-Work Zone. The four glance parameters were only processed
with respect to the participant’s glances towards the billboards and hence are only
included for the Pre-Work Zone. Mean speed and mean lane position were the only
parameters processed for all three data collection zones.
135
To determine to what extent the slogans in the billboards had an effect on driver
behavior, comparisons were done by conducting one-way ANOVA tests on the data
collected from the sample population. Each of the seven driver behavior parameters were
analyzed to compare the behaviors exhibited as drivers approached either the control or
test billboards as well as the behaviors after having seen those signs. For the statistical
analyses, the one-way ANOVA tests were conducted at a 95 percent level of confidence
(α = 0.05). The null hypothesis for each of the parameters stated that the means were
equal for each type of billboard encountered. This would indicate that the observed
behaviors were the same between each of the signs. The alternative hypothesis stated that
there was a difference in the mean values between the types of billboards encountered
which would equate to a difference in driver behavior by those signs. The results of the
one-way ANOVA tests conducted on the data in the Pre-Work Zone are shown in Table
30. The mean fixation, total fixation, and proportion of fixation were found to have
violated the homogeneous variances assumption so the F-ratio and degrees of freedom
values reported in Table 30 are the results of the Welch’s F-test.
Table 30: Results of the one-way ANOVA tests on driver behavior in Pre-Work Zone of
Scenario #1
Degrees
Source of Sum of
Mean
FCritical
Parameter
of
Result
Variation Squares
Square Ratio F-ratio
Freedom
Speed
Lane
Position
Model
Residual
Model
Residual
54.21
9222.44
11.73
219.00
12
467
12
467
4.52
19.75
0.98
0.47
0.23
1.77 Accept Null
2.08
1.77
Reject Null;
E.S. = 0.23
136
Table 30: Continued
Parameter
Variation
in Lane
Position
Mean
Fixation
Total
Fixation
Number of
Fixations
Proportion
Fixation
Duration
Source of Sum of
Variation Squares
Model
Residual
Model
Residual
Model
Residual
Model
Residual
Model
Residual
Degrees
Mean
FCritical
of
Square Ratio F-ratio
Freedom
11.64
12
0.97
494.71
38.02
222.32
230.17
1195.62
62.98
1064.54
467
12
144.25
12
144.17
12
455
1.06
3.17
0.49
19.18
2.63
5.17
2.34
2.21
12
0.18
11.24
144.19
0.03
Result
0.92
1.77 Accept Null
9.22
1.82
Reject Null;
E.S. = 0.38
10.13
1.82
Reject Null;
E.S. = 0.40
2.21
1.77
Reject Null;
E.S. = 0.23
10.33
1.82
Reject Null;
E.S. = 0.41
Based on the results of the one-way ANOVA tests, there were no significant
differences found in either speed or variation in lane position at a 95 percent level of
confidence. This indicates that the mean speeds and mean variations in lane position as
participants drove up to the billboards were equivalent between the test and control signs.
Conversely, the results of the one-way ANOVA for the other parameters produced Fratios which were greater than the respective critical F-ratios, indicating that the null
hypothesis should be rejected for those parameters.
Since the alternative hypothesis was accepted, the results of the Games-Howell
tests were referenced for those parameters to determine where the differences existed
between the mean values for the different signs. Since the Games-Howell test compares
each billboard individually, the resulting tabulated data is quite lengthy for this particular
scenario. In attempts to limit the length of this chapter, the Games-Howell test results can
137
be found in Appendix D. Mean fixation, total fixation, and proportion of fixation duration
were all found to have significant differences between the control and test billboards at a
95 percent level of confidence. The control billboards were found to have significantly
shorter mean and total fixations in comparison to the test billboards, as expected. Drivers
were also found to look at the control billboards for a smaller proportion of the time they
were approaching the signs when compared to the test billboards. Based on these results
and the effect sizes listed in Table 30, the campaign billboards were found to have a
medium-sized practical effect on those glance behaviors. No statistically significant
differences were found between any of the test billboards for these parameters.
Additionally, the Games-Howell tests did not produce any significant results for
either lane position or number of fixations despite the results of the one-way ANOVA.
This could be an indication that the violations of the normality assumption may have
produced false significant results for these two parameters. To verify the results of the
statistical tests, the mean values for each billboard were plotted in Figures 11 and 12.
Lane Position (ft.)
-1.50
-1.25
-1.00
-0.75
-0.50
-0.25
Control
Slogan 12
Slogan 11
Slogan 10
Billboard Content
Slogan 9
Slogan 8
Slogan 7
Slogan 6
Slogan 5
Slogan 4
Slogan 3
Slogan 2
Slogan 1
0.00
Figure 11. Mean values for lane position in the Pre-Work Zone of Scenario #1.
138
2.50
2.00
1.50
1.00
0.50
Control
Slogan 12
Slogan 11
Slogan 10
Billboard Content
Slogan 9
Slogan 8
Slogan 7
Slogan 6
Slogan 5
Slogan 4
Slogan 3
Slogan 2
0.00
Slogan 1
Number of Fixations
3.00
Figure 12. Mean values for number of fixations in the Pre-Work Zone of Scenario #1.
The mean values for lane position were found to vary between 0.87 feet and 1.33
feet left of the center of the lane for the 12 test billboards and control billboard as seen in
Figure 11. The “My Daddy Works Here” billboard (Slogan #8) was found to have drivers
traveling closer to the center of the lane while drivers were furthest away from center as
they approached the control billboard. However, this was a difference of approximately
six inches across 30 participants having seen each of the test billboards while the control
billboard was seen 120 times. Thus, the Games-Howell test results are likely correct in
indicating no significant differences in lane position as drivers encountered the signs.
Figure 12 shows that the mean number of fixations ranged between 1.62 fixations
for the control billboard and 2.76 fixations for the “Give ‘em a Brake” billboard (Slogan
#7). The test billboard with the smallest mean was the sign using the “Slow Down Save
Lives” slogan (Slogan #10) at 2.21 fixations. Upon examining the means, it seems
plausible that the results of the one-way ANOVA are correct in indicating a difference in
139
the means for the number of fixations data. As for the other glance behavior parameters,
the number of fixations towards the control billboard appears to be different than for the
test billboards whereas no differences exist between the test billboards. However, the
interval nature of this parameter and the non-normality of the data may be combining to
produce a false significant result. In attempts to remain conservative with the analysis,
the null hypothesis was accepted. Thus, the control and test billboards have no significant
differences in the number of fixations made towards the signs.
One-way ANOVA tests were also conducted on the mean values for speed and
lane position in the Advanced Warning Area and Work Zone. Results of the statistical
analyses for these parameters are shown in Tables 31 and 32. Based on the results of the
one-way ANOVA tests, the calculated F-ratios were found to be less than the critical Fratio in all four tests. Hence, the null hypothesis of equivalent means for the test and
control billboards is accepted for both parameters across both areas. At a 95 percent level
of confidence, there were no significant differences in the observed mean speeds and
mean lane positions of the participants after they had passed either the safety campaign
billboards or the blank control billboard.
Table 31: Results of the one-way ANOVA tests on driver behavior in Advanced Warning
Area of Scenario #1
Degrees
Source of Sum of
Mean
FCritical
Parameter
of
Result
Variation Squares
Square Ratio F-ratio
Freedom
Speed
Lane
Position
Model
Residual
Model
Residual
67.08
9737.73
3.49
161.20
12
467
12
467
5.59
20.85
0.29
0.35
0.27
1.77 Accept Null
0.84
1.77 Accept Null
140
Table 32: Results of the one-way ANOVA tests on driver behavior in Work Zone of
Scenario #1
Degrees
Source of Sum of
Mean
FCritical
Parameter
of
Result
Variation Squares
Square Ratio F-ratio
Freedom
Speed
Lane
Position
Model
Residual
Model
Residual
123.37
7522.19
5.71
250.77
12
467
12
467
10.28
16.11
0.48
0.54
0.64
1.77 Accept Null
0.89
1.77 Accept Null
After participants had completed the driving portion of Scenario #1, they were
provided a brief post-test questionnaire to complete. The questionnaire included two
questions which asked participants if they considered the signs to be cluttered and asked
them to write down any of the slogans they could recall from the billboards. The first
question was used to compare the simplified billboards used in the driving simulator
study with those used in the public survey study. Meanwhile, the second question was
used to see which of the slogans were considered to be memorable to the participants.
Upon completion of the driving simulator study, the questionnaire responses were
compiled and there was a total of 47 responses collected from participants who drove
through at least one version of Scenario #1. Of those 47 participants, only four considered
the billboards too cluttered. Thus, the percentage of participants who thought the signs
were cluttered went from approximately 60 percent in the public survey study to 8.5
percent in the driving simulator study. The only changes made to the billboards were to
141
reduce the amount of text present by just having a single slogan over two lines of text and
reducing the size of the graphics in relation to the font size of the slogan.
For the second question, all 47 participants had written down at least one slogan
they believed they could recall. There were a total of 87 slogans that participants wrote
down, but only 53 of those (60.9 percent) were exact matches to the slogans used in the
scenario. Typically, the other 34 responses were either portions of a slogan that was used
or a combination of two or more slogans. Figure 13 summarizes the exact matches for
each of the slogans that participants could recall. Within the exact matches, the slogan
with the highest frequency was “My Daddy Works Here” at 18 responses. “Slow Down
in Work Zones” was the slogan with the next most responses at a total of nine. There
were only three slogans that did not receive a response which was an exact match. Those
slogans were as follows: “Slow for the Cone Zone,” “Slow Down Respect the Barrel,”
Slogan Used
and “See Orange Drive Slow.”
My Daddy Works Here
Slow Down in Work Zones
Give 'Em a Brake
Drive Slow Drive Safe
Let 'Em Work, Let 'Em Live
Slow Down, Save Lives
Brake for Barrels
Go Slow for Safety
Go Slow Thru the Zone
0
5
10
15
Number of Exact Matches
Figure 13. Results of post-test questionnaire on recalling the slogans used.
20
142
6.2.2 Scenario #2 – Graphics
The second scenario in the driving simulator study was designed to test the
graphics used in the billboard advertisements. As in Scenario #1, participants were
presented with either a control billboard or a work zone safety campaign billboard prior
to driving through a construction work zone. The campaign billboards used in Scenario
#2 had one of three graphics combinations while the slogan, color combination, and the
placement and orientation of the sign with respect to the driver were all kept constant. As
participants drove through the simulated environment, data was collected on the driving
performance and glance behavior parameters outlined previously. The data was processed
over the same three areas as for Scenario #1. The four glance parameters were processed
with respect to the participant’s glances towards the billboards and, hence, are only
included for the Pre-Work Zone. Mean speed and mean lane position were processed for
all three zones whereas variation in lane position was processed only in the Pre-Work
Zone.
To determine to what extent the graphics in the billboards had an effect on driver
behavior, comparisons of the mean values were done by conducting one-way ANOVA
tests on the data collected from the sample population. Each of the seven driver behavior
parameters were analyzed to compare the behaviors exhibited when the driver was
approaching the control and test billboards as well as after having passed the signs. For
the statistical analyses, the one-way ANOVA tests were conducted at a 95 percent level
of confidence (α = 0.05). The null hypothesis for each of the parameters stated that the
means were equal for each type of billboard encountered. This would indicate that the
143
observed behaviors were the same between each of the signs. The alternative hypothesis
stated that there was a difference in the mean values between the types of billboards
encountered, which would equate to a difference in driver behavior between the signs.
The results of the one-way ANOVA tests conducted on the data in the Pre-Work Zone
are shown in Table 33. The total fixation and proportion of fixation data were each found
to have violated the homogeneous variances assumption for the one-way ANOVA. To
address that issue the F-ratio and degrees of freedom values reported in Table 33 are the
results of the Welch’s F-test.
Table 33: Results of the one-way ANOVA tests on driver behavior in Pre-Work Zone of
Scenario #2
Degrees
Source of Sum of
Mean
FCritical
Parameter
of
Result
Variation Squares
Square Ratio F-ratio
Freedom
Speed
Lane
Position
Variation
in Lane
Position
Mean
Fixation
Total
Fixation
Number of
Fixations
Proportion
Fixation
Duration
Model
Residual
Model
Residual
21.74
1609.78
0.62
58.51
3
116
3
116
7.25
13.88
0.21
0.50
0.78
3
0.26
Residual
Model
Residual
Model
Residual
72.88
7.76
43.24
64.10
287.78
116
3
116
3
60.19
0.63
2.56
0.37
21.37
2.48
Model
Residual
38.00
318.00
3
116
12.67
2.74
Model
0.57
3
0.19
Residual
2.55
60.43
0.02
Model
0.52
2.69 Accept Null
0.41
2.69 Accept Null
0.42
2.69 Accept Null
6.86
2.69
Reject Null;
E.S. = 0.39
16.21
2.77
Reject Null;
E.S. = 0.43
4.62
2.69
Reject Null;
E.S. = 0.33
15.88
2.77
Reject Null;
E.S. = 0.43
144
Based on the results of the one-way ANOVA tests reported in Table 33, there
were no significant differences found in any of the three driving performance measures at
a 95 percent level of confidence. The null hypothesis was accepted for speed, lane
position, and variation in lane position which correlates to the driver behavior being
equivalent for these measures as participants drove up to the control and test billboards.
However, for the glance behavior measures, the null hypothesis was rejected in each of
the one-way ANOVA tests signifying that a difference in the mean values exists between
the four types of billboards being compared. The Games-Howell test was conducted on
each of the four glance behavior parameters in order to determine where the differences
existed between the mean values. Once again the results of the Games-Howell tests are
presented in Appendix D. Mean fixation, total fixation, number of fixations, and
proportion of fixation duration were all found to have significant differences between the
control and test billboards at a 95 percent level of confidence. However, no statistically
significant differences were found between any of the test billboards for these
parameters.
Mean fixation and total fixation for the control billboards were found to be of
significantly shorter duration in comparison to those for the campaign billboards. Each of
the three test billboards had mean fixations that were approximately 0.90 seconds, yet
were almost three times larger than the mean fixation of 0.32 seconds for the control
billboard. The difference was even larger in magnitude for the total fixation data. The test
billboard using only the Laborer’s Local logo had the shortest total fixation at a mean of
2.22 seconds while the control billboard had a mean of 0.61 seconds for total fixation.
145
The number of fixations per sign were also significantly different as the control billboard
averaged 1.03 fixations and the test billboards averaged between 2.23 and 2.43 fixations.
Finally, participants were found to look at the control billboards for a significantly shorter
proportion of the time they were approaching the signs at 5.8 percent in comparison to
approximately 21 percent for the three campaign billboards. Combining these results with
the effect sizes listed in Table 33, it was found that the presence of an advertisement on
the sign had a medium-sized practical effect on each of the glance behavior parameters.
One-way ANOVA tests were also conducted on the mean values for speed and
lane position for both the Advanced Warning Area and Work Zone. Results of the
statistical analyses for these driver behavior parameters are shown in Tables 31 and 32.
Based on the results of the one-way ANOVA tests, the calculated F-ratios were found to
be less than the critical F-ratio in all four tests. Hence, the null hypothesis of equivalent
means for the test and control billboards is accepted for speed and lane position across
both areas. At a 95 percent level of confidence, there were no significant differences in
the observed speeds and lane positions of the participants after they had passed either the
safety campaign billboards or the blank control billboard.
Table 34: Results of the one-way ANOVA tests on driver behavior in Advanced Warning
Area of Scenario #2
Degrees
Source of Sum of
Mean
FCritical
Parameter
of
Result
Variation Squares
Square Ratio F-ratio
Freedom
Speed
Lane
Position
Model
Residual
Model
Residual
18.72
1750.63
0.20
39.47
3
116
3
116
6.24
15.09
0.07
0.34
0.41
2.69 Accept Null
0.20
2.69 Accept Null
146
Table 35: Results of the one-way ANOVA tests on driver behavior in Work Zone of
Scenario #2
Degrees
Source of Sum of
Mean
FCritical
Parameter
of
Result
Variation Squares
Square Ratio F-ratio
Freedom
Speed
Lane
Position
Model
Residual
Model
Residual
2.09
548.63
0.09
77.89
3
116
3
116
0.70
4.73
0.03
0.67
0.15
2.69 Accept Null
0.05
2.69 Accept Null
6.2.3 Scenario #3 – Placement and Orientation
The third scenario in the driving simulator study was designed to test the
placement and orientation of the billboard advertisement with respect to the driver to
examine what impact it had on driver behavior. Unlike the previous scenarios, there were
no work zones placed on the freeway in Scenario #3. Participants were presented with the
same work zone safety campaign billboard multiple times as they drove along the fourlane undivided freeway in an urban setting. The placement and orientation of the sign
with respect to the driver was varied throughout the scenario to see how the location of
the billboard in the physical world affected the participants. Billboards were placed at
varying heights (high or low), on either side of the freeway, and at three different angles.
This resulted in a total of 12 different combinations for the placement and orientation of
the signs. As participants drove through the simulated environment, data was collected on
the three driver performance and four glance behavior parameters outlined previously.
147
The data was processed in the same manner as the Pre-Work Zone in Scenarios #1 and
#2; starting at a distance of 1250 feet prior to the billboard and ending at the sign itself.
To determine to what extent the placement and orientation of the billboard had an
effect on driver behavior, the mean values for the various parameters were compared by
conducting one-way ANOVA tests on the data collected from the sample population.
Each of the seven driver behavior parameters were analyzed to compare the behaviors
exhibited as the driver was approaching each of the billboard locations. For the statistical
analyses, the one-way ANOVA tests were conducted at a 95 percent level of confidence
(α = 0.05). The null hypothesis for each of the parameters stated that the means were
equal for each billboard location encountered, which would indicate that the observed
behaviors were the same between each of the physical locations. The alternative
hypothesis stated that there was a difference in the mean values between the billboard
locations encountered, which would equate to a difference in driver behavior between the
signs. The results of the one-way ANOVA tests conducted on the data as participants
approached the various billboard placements and orientations are shown in Table 36.
Table 36: Results of the one-way ANOVA tests on driver behavior in approaching the
billboards in Scenario #3
Parameter
Speed
Lane
Position
Variation
in Lane
Position
Source of Sum of Degrees of
Variation Squares Freedom
Model
Residual
Model
Residual
Model
Residual
Mean
Square
58.68
3660.20
5.50
180.18
11
336
11
336
5.33
10.89
0.50
0.54
10.70
11
0.97
473.52
336
1.41
FRatio
Critical
F-ratio
Result
0.49
1.82
Accept
Null
0.93
1.82
Accept
Null
0.69
1.82
Accept
Null
148
Table 36: Continued
Parameter
Mean
Fixation
Total
Fixation
Number of
Fixations
Proportion
Fixation
Duration
Source of Sum of Degrees of
Variation Squares Freedom
Model
Residual
Model
Residual
Model
Residual
Mean
Square
5.73
134.68
39.70
1311.41
13.47
1855.59
11
336
11
336
11
336
0.52
0.40
3.61
3.90
1.23
5.52
Model
0.24
11
0.02
Residual
8.22
336
0.02
FRatio
Critical
F-ratio
Result
1.30
1.82
Accept
Null
0.93
1.82
Accept
Null
0.22
1.82
Accept
Null
0.91
1.82
Accept
Null
From Table 36, the one-way ANOVA tests resulted in acceptance of the null
hypothesis for all three driving performance measures and all four glance behavior
measures. Thus, there were no significant differences found in any of the driver behavior
measures at a 95 percent level of confidence. Overall, driver behavior was found to be
equivalent as participants encountered the twelve billboards set at various placements and
orientations.
6.2.4 Impact of Long Glances on Safety
As discussed in Chapter 2, when drivers make glances away from the forward
roadway for more than two seconds, their ability to perform driving tasks decreases and
their risk of being involved in a crash increases. Thus, long glances away from the
roadway are recognized to negatively impact safety. During the previous analyses of the
different billboard elements the impacts on driving performance and glance behavior
were examined separately. However, the two elements are interrelated and are recognized
to contribute to a driver’s ability to safely travel on the roadways.
149
When the eye-tracking data was processed, the individual fixations made by a
participant were found in order to determine the glance behavior parameters used in this
study. In Scenario #1 there were a total of 1,025 individual fixations toward the
billboards. Approximately 11 percent of those fixations, or a total of 110, were two
seconds or longer in duration. During Scenario #2, long glances to campaign billboards
comprised 9.2 percent of the 240 fixations recorded while, for Scenario #3, long glances
comprised 5.5 percent of the 896 fixations recorded. The percentages of participants
making at least one long glance were 64 percent, 45 percent, and 54 percent of the sample
populations for Scenarios #1, #2, and #3, respectively.
During the processing of the eye-tracking data, the data for the observed driving
performance over the same time span were also determined so the direct relationship
between the two aspects of driver behavior could be studied. For the purposes of
analyzing a participant’s driving performance while they were glancing at the billboards
in this study, it was decided that the most important parameter in terms of safety was the
total variation in lane position. Since the fixations on the billboards were mainly short in
duration, the amount of variation in speed would likely be minimal. Furthermore, lane
position was more likely to be affected by the glances made towards the billboards as
drivers would be less likely to realize they were drifting within the lane since their
attention was away from the forward roadway. As a result, data processing efforts were
focused on determining the total variation in lane position as drivers were looking
towards the billboards.
150
For the comparison of long glances and short glances, one-way ANOVA tests
were conducted on the total amount of variation in lane position during those glances for
all three scenarios in the driving simulator study. The one-way ANOVA tests were each
conducted at a 95 percent level of confidence (α = 0.05). The null hypothesis stated that
the means were equal for long and short glances which would indicate that the observed
movement within the lane was equivalent between the types of glances. The alternative
hypothesis stated that there was a difference in the mean values between long and short
glances which would equate to a difference in driver behavior between the two. The
results of the one-way ANOVA tests conducted on the total variation in lane position as
participants glanced at the billboards are shown in Table 37. In all three scenarios, the
total variation in lane position data was found to have violated the homogeneous
variances assumption for the one-way ANOVA test. To correct those issues the F-ratio
and degrees of freedom values reported in Table 37 are the results of the Welch’s F-test.
Table 37: Results of the one-way ANOVA tests on total variation in lane position while
participants were glancing at the billboards by scenario
Degrees
Scenario Source of Sum of
Mean
FCritical
of
Result
#
Variation Squares
Square Ratio F-ratio
Freedom
Reject
Model
28.77
1
28.77
1
88.60
3.92
Null;
Residual
85.78
112.78
0.08
E.S.= 0.50
Reject
Model
10.29
1
10.29 17.46
2
4.32
Null;
Residual
23.31
21.30
0.10
E.S.= 0.55
Reject
Model
15.76
1
15.76
3
31.35
4.04
Null;
Residual
58.34
48.41
0.07
E.S.= 0.46
151
In Table 37, it can be seen that for all three scenarios the null hypothesis was
rejected for the total variation in lane position while participants were glancing at the
billboards. Thus, there were significant differences found in the driving behavior
displayed during the long and short glances at a 95 percent level of confidence. For
Scenario #1, the total variation in lane position was an average of 0.74 feet during long
glances as opposed to 0.19 feet during short glances. For Scenario #2, the means for total
variation were 0.89 feet during long glances and 0.17 feet during short glances. Finally,
participants making long glances in Scenario #3 varied 0.73 feet within the lane on
average whereas participants making short glances varied 0.15 feet. Additionally, the
effect sizes listed in Table 37 indicate that glance length had a medium-sized practical
effect on the total variation in lane position across the three scenarios.
152
CHAPTER 7: CONCLUSIONS
This research was conducted to evaluate the impacts on driver behavior from
billboard advertisements used in a work zone safety campaign. Two studies were
conducted in order to evaluate the different elements of the billboards and assess their
ability in communicating the campaign message. The following sections include
discussion of the results of the respective studies and how those results correspond to
using billboards in safety campaigns. The final section presents several recommendations
for future research into this field.
7.1 Public Survey Study
The primary goal of the public survey study was to determine the background and
text color combination that was preferred by the public for use in a work zone safety
campaign advertisement. Participants were presented with a set of four billboards and
asked to select the sign which best met the conditions described in that particular
question. Based on the statistical analyses of the survey responses, the fluorescent
yellow-green with black text was preferred by the overall population for increasing
awareness of work zones and for being seen during the daytime. That same color
combination was also preferred by the overall population along with the combination
featuring a yellow background with black text for increasing caution in work zones and
for being seen during the nighttime.
These two color combinations were also largely preferred when examining the
responses by age group, gender, years of driving experience, and knowledge of rules for
driving in work zones. When examining the four conditions by each of those different
153
groups, there were a total of 52 cases where a billboard preference was determined
through statistical analysis. The fluorescent yellow-green background with black text was
found to be a preferred color combination in 45 cases whereas the yellow background
with black text was a preferred color combination in 34 cases. There were a total of 28
cases where those two color combinations were equally selected.
As seen in the national data from 2010 regarding speeding-related fatal crashes in
work zones, the population with the highest risk was drivers in the 16-25 age group with
the majority of those drivers being males (1). This trend is evident in previous data as the
overall crash involvement rates on the roadways nationwide in 1996 were highest for
drivers in the 16-24 age group (48). Also from 1992 to 2004, Li and Bai found male
drivers to be at fault in the majority of injury and fatal crashes classified as speedingrelated and having occurred in highway work zones in Kansas (6). When examining the
responses from those participants in the 16-24 age group, the fluorescent yellow-green
and yellow backgrounds were equally selected for three of the four conditions being
evaluated. Only the fluorescent yellow-green background with black text was preferred
for the sign best seen during the day. The same color combination preferences were seen
for male drivers in the corresponding conditions.
Since the color combinations using the fluorescent yellow-green and yellow
backgrounds were equally selected in most cases, the MUTCD sign color meanings were
once again referenced to determine which of the two should be used. This was done so
that consistency was maintained with how traffic signs communicate their message to
drivers to avoid potential driver confusion with the intent of the billboards. From the
154
MUTCD, fluorescent yellow-green is to be used as the background color for school,
pedestrian, and bicycle warning signs whereas yellow is the background for general
warning signs (40). While each of these colors are consistent with the campaign message
warning drivers to be more cautious in work zones, the yellow background is more
appropriate for the conditions of the campaign. Fluorescent yellow-green is intended for
use in particular conditions that do not associate with the intent of the work zone safety
campaign.
The background color for traffic signs that best matches the purpose of the
campaign is orange as it is used for temporary traffic control signs that are intended to
guide drivers safely through work zones (40). However, the orange background was
directly compared with the yellow background in the same set of billboards and was only
found to be selected more frequently than the yellow background for two cases. Thus, the
yellow background with black text was determined to be the color combination that
should be used for billboard advertisements in a work zone safety campaign.
Another goal of the public survey study was to determine if the original billboard
advertisement used by Laborer’s Local 860 was considered to be too cluttered. The
researchers believed that the sign contained too much content for drivers to be able to
process the campaign message while traveling along the freeway. Accordingly,
participants were asked whether or not the recreated billboards were cluttered. Since 60
percent of participants indicated that the billboards are too cluttered, they are likely
ineffective at communicating the safety message. Drivers seeing the original billboards
were probably not able to easily receive the message given the cluttered format.
155
7.2 Driving Simulator Study
The driving simulator study was designed to evaluate three different elements of
the billboard advertisements and assess the impacts of those elements on driver behavior.
The elements studied were the slogans and graphics used in the advertisement as well as
the placement and orientation of the sign with respect to the driver. Each of these
elements were evaluated in separate scenarios by varying that particular element while
keeping the other elements consistent throughout the scenario. While in the simulator,
participants drove along an undivided four-lane highway in an urban environment with
no other traffic present. During Scenarios #1 and #2 participants saw a billboard prior to
driving through a work zone in which the right lane was closed. For Scenario #3,
participants saw several signs that were placed at varying locations on the side of the
highway and were oriented at various angles, but there were no work zones present.
Throughout the three scenarios, data was continuously collected on multiple
driving performance and glance behavior parameters to observe participant’s behavior as
they approached the billboards and after having passed the signs. While data was
collected on additional driving performance parameters, there were only three parameters
which were included in the statistical analyses. Mean values for speed and lane position
were analyzed as participants approached the billboards and after having passed the
signs. Variation in lane position was only analyzed as drivers approached the billboards.
The four glance behavior parameters included in the statistical analyses were mean
fixation, total fixation, number of fixations, and proportion of fixation duration. Each of
156
these parameters was analyzed by the glances towards the billboards while drivers were
approaching the signs.
Across each of the three scenarios, overall driving performance was found to not
be impacted by the billboard elements. Mean speeds were found to be equivalent between
the billboards being tested within the individual scenarios. Despite the campaign message
telling drivers to slow down in work zones, there were no differences seen in mean
speeds through the work zones if drivers had seen a campaign billboard or the blank
control billboard. This is in direct contrast from the pre-test questionnaire results where
the majority of drivers had reported that a work zone safety campaign would cause them
to reduce their speeds. Similar results were found by Kane et al. who found the majority
of drivers surveyed had responded that they slow down in work zones, yet spot speed
studies showed large percentages of drivers exceeding the speed limit in work zones (12).
For the work zones in the driving simulator study, mean speeds for each of the
billboards tested were between 56-58 mph in Scenario #1 and just over 55 mph in
Scenario #2. While these mean speeds do represent a decrease from the speeds seen in the
Pre-Work Zone and Advanced Warning Area, they are still exceeding the posted speed
limit. Thus, drivers are indeed reducing their speeds in the work zones when compared to
their prior speeds, although not to a level that is below the posted speed limit. However,
the reduction in speeds cannot be attributed to the campaign billboards as speeds were
determined to be equivalent between billboard types. Consequently, the campaign
message in the billboards was not found to cause drivers to alter their speeds when
driving in the work zones. This finding is consistent with Jamson and Merat who found
157
that displaying safety campaign messages about speed on variable message signs did not
impact the observed speeds (39).
Mean lane position was also found to not be impacted by the billboards
encountered in the three scenarios. In the first two scenarios, participants were observed
to have driven farther to the left-side of the travel lane, or away from the work zone,
when compared to their mean lane position as they drove in the Pre-Work Zone and
Advanced Warning Area. However, there were no significant differences found between
the test and control billboards in mean lane position throughout the work zone in either
scenario. This indicates that while participants had seen the work zone safety message,
the billboards did not cause them to travel farther away from the work zone. Participants
had responded during the pre-test questionnaire that safety campaigns would cause them
to drive farther away from work zones, yet that behavior was not observed.
During each of the scenarios, as participants were approaching the billboards,
there were no significant differences found in mean lane position or variation in lane
position. Variation in lane position can indicate driver distraction as increased lateral
movement within the lane potentially corresponds to the driver unknowingly allowing the
vehicle to drift laterally. None of the elements studied can be considered to be more
distracting than their respective counterparts as the means were found to be equivalent
within each scenario. This contrasts with the findings of Young et al. where lateral
control of the vehicle was hindered when driving past billboards (38). The results of this
study are more consistent with Lee, McElheny, and Gibbons who found no differences in
lane position as drivers passed billboards located along an interstate highway (31).
158
Glance behavior was found to be impacted by the billboard content in Scenarios
#1 and #2. For these scenarios, the glance behavior toward the control billboard was
determined to be significantly different from the behavior towards the campaign
billboards. For the control billboards, mean fixation and total fixation were found to be
significantly shorter than those to the billboards containing the various slogans and
graphics. Accordingly, the significant differences in proportion of fixation duration
showed that participants looked at the control billboards for a smaller proportion of time
while driving up to the signs in the first two scenarios. In Scenario #2, the number of
fixations toward the campaign billboards were significantly greater than those towards
the control billboards. Conversely, in Scenario #3 there were no significant differences
found in the glance behavior towards the signs despite their varied locations on the side
of the freeway and being at different orientation angles with respect to the driver.
The results of the four glance behavior parameters across the scenarios in the
study show conclusively that drivers were observed to glance more frequently and for
longer periods at billboards containing advertisements than towards those that contained
no content. However, there were no significant differences observed between the
campaign billboards presented in each scenario. Overall, the glance behavior results of
this research are primarily in contrast with those seen in previous research. Crundall,
Loon, and Underwood found that mean fixation and number of fixation were affected by
the height of advertisements (32). However, no differences were seen in glance behavior
as the billboard height was varied during Scenario #3. Megias et al. found that billboards
with negative content had greater mean fixations, total fixations, and number of fixations
159
(33). The slogans in Scenario #1 each communicated the need to reduce speeds, but the
emotions tied with how that message was portrayed differed as some used personal
connections such as “My Daddy Works Here.” Yet there were no differences seen in the
glance behavior between the slogans used in Scenario #1.
Additionally, differences were seen in the magnitude of the observed glance
behaviors in this research. During an on-road study Smiley, Smahel, and Eizenmann
found the longest glance to a video billboard to be 1.47 seconds (29). The individual
fixations to billboards in this study were observed to be longer overall, with
approximately 10 percent of all fixations longer than two seconds in duration. Perez and
Bertola also found short glances towards billboards in their study and the resulting
proportion of fixation duration to the billboards were low at roughly 0.03 (30). On
average, the proportion of fixation duration for each of the billboards in this study ranged
between 0.05 and 0.29. The larger values for the glance behavior parameters in this study
may be due to there being no traffic on the highways. Without any traffic present,
participants were more likely to look away from the forward roadway as there were no
concerns about what was going on around the vehicle. Thus, the magnitude of glance
behavior for this study is likely greater than that seen on actual roadways.
Despite these differences, there were two studies that had findings which were
similar to that of this research. In previous studies of conventional billboards and signs
with active components, such as digital billboards, glance behavior was found to be
impacted. Beijer found the number of fixations to be greater to billboards with active
content while Lee, McElheny, and Gibbons found mean fixations to be longer for digital
160
billboards (28, 31). These findings are consistent with the differences in glance behavior
seen between the campaign billboards and those that were blank.
The results of the statistical analyses on the lane position and number of fixations
data from Scenario #1 illustrated the potential issues with having violated the assumption
of normality. For these two parameters, the one-way ANOVA results indicated that there
were differences between the groups, but the Games-Howell tests indicated there were no
differences. After examining the data to see what may have occurred, there were no large
differences seen in lane position so the null hypothesis was accepted. It did appear that
there may be a difference in number of fixations between the control and test billboards;
however the null hypothesis was rejected to remain conservative in the analysis. The nonnormality of the data may have produced the differing results in those statistical tests, but
the overall results of the statistical analyses are likely valid. Examining the mean values
for the groups in each of the analyses conducted for the study it is evident that the
differences found are valid. Throughout the three scenarios there were only minor
differences seen in the driving performance parameters between the groups, but not to an
extent that would be practical on actual roadways. Similarly, the glance behaviors
observed between the test billboards did not differ to a practical extent for the three
scenarios. Therefore, the violations of the normality assumption are not believed to have
impacted the results of the statistical analyses conducted in this research.
Analysis of the variation in lane position observed during the glances to the
billboards was done in each of the scenarios to examine the impacts to safety from long
glances. Results of the statistical analyses indicated there were significant differences in
161
the variation in lane position between long and short glances. Long glances were defined
to be those longer than two seconds as proposed by Rockwell (22). Throughout the study,
participants were found to have larger variations in lane position when making long
glances than during short glances. Thus, as drivers made long glances to the billboards
their ability to maintain their lane position was negatively impacted.
This confirms previous research which indicated that long glances away from the
forward roadway are a safety concern. Klauer et al. found that the chance of being
involved in a crash more than doubles when long glances are made away from the
forward roadway (26). If drivers are unable to keep their vehicle within the lane while
glancing away from the roadway than they may leave their travel lane and potentially
come into contact with another vehicle. Consequently, the use of billboards in a work
zone safety campaign should balance how long participants need to look away from the
roadway to comprehend the sign with the benefits observed in work zone safety.
Based on the observed behaviors throughout the work zones in this study, the
campaign billboards were not found to impact driver behavior in a way that would
improve safety. While participants on average drove through the work zones at reduced
speeds and drove further away from the construction barrels than they did prior to the
work zones, the same behavior was seen for whether they had passed a campaign
billboard or a control billboard. Thus, the campaign billboards were not found to impact
driver behavior in a way that improved safety. As a result, billboard advertisements
should not be the only component of an effective work zone safety campaign. The signs
should be used as part of a larger campaign effort which includes advertising to promote
162
awareness and enforcement of that message within work zones. Using personal appeals in
the campaign would also be beneficial as the participants in this study best recalled the
slogan “My Daddy Works Here” after having seen several billboards. The ability to draw
on personal connections may be more likely to increase awareness of work zone safety
rather than by just using a message which communicates the desired behavioral change.
Development of work zone safety campaigns should give consideration to the
elements involved and how it will be carried out. With regards to the billboards evaluated
in this research, using the yellow background with black text along with streamlined
content, in terms of slogans and graphics, is more likely to communicate the message to
drivers. However, communication of that message alone will not produce the desired
behavioral changes associated with increased caution in work zones. Thus, the billboard
advertisements should be used as a method of promoting the work zone safety campaign
and should be supplemented with other elements to maximize their effectiveness.
7.3 Recommendations for Future Research
Based on the results of this research, there are additional areas in which further
evaluation of billboard advertisements in safety campaigns could be conducted. First,
introducing traffic onto the highway with participants would likely produce a better
evaluation of glance behavior to the signs. As discussed previously, the glance behaviors
observed in this research were found to be greater than that seen in previous studies. One
of the primary factors is likely the lack of traffic on the highways throughout the driving
simulator study. This probably contributed to participants making longer glances to the
billboards than they typically might in an actual driving environment. From personal
163
experience with the driving simulator used in this study, there are some limitations in
being able to accurately replicate traffic driving through a work zone. Conducting a
naturalistic driving study could potentially result in a more accurate assessment, however
the scope of billboards evaluated may need to be limited.
Expanding the participant pool to include additional age groups is another area to
consider for future studies. Including participants of different age groups could possibly
produce differences in the observed driving performance and glance behavior. Younger
drivers are recognized to have behaviors that differ from those of more experienced
drivers which contributes to why younger drivers are found to be involved in more
crashes. Furthermore, drivers in the other age groups are those that travel through work
zones more often, so those individuals should also be targeted in safety campaigns.
Evaluating the driver behavior of individuals in those groups may produce different
results than from focusing specifically on younger drivers.
Finally, evaluation of additional advertising media should be done to compare
those methods with billboards. Most safety campaigns use multiple media outlets, such as
television and newspaper advertisements, to promote their message in order to reach as
many people as possible. Comparisons of these outlets have been done primarily through
survey responses in previous research, so the results are based on self-reported behaviors
which tend to differ from actual behaviors. These types of media can be evaluated in a
driving simulator as done in this research. For example, drivers can watch a television
campaign commercial prior to driving in the simulator or can be played a radio
commercial while driving up to a work zone.
164
REFERENCES
1. National Center for Statistics and Analysis. Fatality Analysis Reporting System
Encyclopedia. National Highway Traffic Safety Administration, U.S. Department of
Transportation. http://wwwfars.nhtsa.dot.gov//QueryTool/QuerySection/SelectYear.aspx. Accessed February 23,
2013.
2. Antonucci, N. D., K. K. Hardy, J. E. Bryden, T. R. Neuman, R. Pfefer, and K. Slack.
NCHRP Report 500: Guidance for Implementation of the AASHTO Strategic
Highway Plan. Vol. 17, A Guide for Reducing Work Zone Collisions. Transportation
Research Board of the National Academies, Washington, D.C., 2005.
3. Campaign Programs. The National Work Zone Safety Information Clearinghouse.
http://www.workzonesafety.org/public_awareness/. Accessed December 9, 2010.
4. National Work Zone Awareness Week – FHWA Work Zone. Federal Highway
Administration, U.S. Department of Transportation.
http://www.ops.fhwa.dot.gov/wz/outreach/wz_awareness.htm. Accessed February 23,
2013.
5. National Center for Statistics and Analysis. Traffic Safety Fact Sheet: 2010 Speeding.
Publication DOT-HS-811-636. National Highway Traffic Safety Administration, U.S.
Department of Transportation, 2012.
6. Li, Y., and Y. Bai. Highway Work Zone Risk Factors and Their Impact on Crash
Severity. Journal of Transportation Engineering, Vol. 135, No. 10, 2009, pp. 694701.
7. National Center for Statistics and Analysis. Traffic Safety Facts 2009. Publication
DOT-HS-811-402. National Highway Traffic Safety Administration, U.S.
Department of Transportation, 2011.
8. Migletz, J., J. L. Graham, I. B. Anderson, D. W. Harwood, and K. M. Bauer. Work
Zone Speed Limit Procedure. In Transportation Research Record: Journal of the
Transportation Research Board, No. 1657, Transportation Research Board of the
National Academies, Washington, D.C., 1999, pp. 24-30.
9. Solomon, D. Accidents on Main Rural Highways Related to Speed, Driver, and
Vehicle. Bureau of Public Roads, U.S. Department of Commerce, 1964.
10. Hauer, E. Speed and Safety. In Transportation Research Record: Journal of the
Transportation Research Board, No. 2103, Transportation Research Board of the
National Academies, Washington, D.C., 2009, pp. 10-17.
165
11. Harsha, B., and J. Hedlund. Changing America’s Culture of Speed on the Roads.
Improving Traffic Safety Culture in the United States – The Journey Forward, AAA
Foundation for Traffic Safety, Washington, D.C., 2007, pp. 257-272.
12. Kane, M. R., L. E. King, K. A. Buch, and M. L. Carpenter. Motorists Perception of
Work Zone Safety. Publication FHWA/NC/99-006. The University of North Carolina
at Charlotte, Charlotte, NC, 1999.
13. California Department of Transportation. The Cone Zone Goes Statewide. California
Department of Transportation Journal, Vol. 4, No. 1, 2003, pp. 37-39.
14. Phillips, R., and R. Torquato. A Review of 45 Anti-speeding Campaigns. Publication
TOI Report 1003/2009. Institute of Transport Economics, Oslo, Norway, 2009.
15. Adamos, G., and E. G. Nathanail. An Experimental Approach Towards the
Evaluation of Mass Media Road Safety Campaigns – A Comparative Analysis of
Two Case Studies. Presented at 90th Annual Meeting of the Transportation Research
Board, Washington, D.C., 2011.
16. Wundersitz, L. N., T. P. Hutchinson, and J. E. Woolley. Best Practice in Road Safety
Mass Media Campaigns: A Literature Review. Publication CASR074. Centre for
Automotive Safety Research, The University of Adelaide, Australia, 2010.
17. Hoekstra, T., and F. Wegman. Improving the Effectiveness of Road Safety
Campaigns: Current and New Practices. IATSS Research, Vol. 34, No. 2, 2011, pp.
80-86.
18. Lee, C., M. Saxena, P. Lin, E. Gonzalez-Velez, and J. W. Rouse. Aggressive Driving
and Safety Campaigns: Lessons Learned from the Better Driver Campaign in Florida.
Presented at 89th Annual Meeting of the Transportation Research Board, Washington,
D.C., 2010.
19. Gantz, W., M. Fitzmaurice, and E. Yoo. Seat Belt Campaigns and Buckling Up: Do
the Media Make a Difference? Health Communication, Vol. 2, No. 1, 1990, pp. 1-12.
20. Eby, D. W., L. P. Kotsyniuk, K. Sudharsan, J. M. Vivoda, and H. K. Spradlin. The
Effects of Paid Media and Enforcement on Safety Belt Use in Michigan. Publication
UMTRI-2002-27. The University of Michigan Transportation Research Institute, Ann
Arbor, MI, 2002.
21. Phillips, R. O., P. Ulleberg, and T. Vaa. Meta-analysis of the Effect of Road Safety
Campaigns on Accidents. Accident Analysis and Prevention, Vol. 43, No. 3, 2011, pp.
1204-1218.
166
22. Rockwell, T. H. Spare Visual Capacity in Driving – Revisited: New Empirical
Results for an Old Idea. Vision in Vehicles II: Proceedings of the Second
International Conference on Vision in Vehicles, Nottingham, U.K., 1988, pp.317-324.
23. Zwahlen, H. T., C. C. Adams Jr., and P. J. Schwartz. Safety Aspects of Cellular
Telephones in Automobiles. Presented at the 18th International Symposium on
Automotive Technology and Automation, Florence, Italy, 1988.
24. Wierwille, W. W. Visual and Manual Demands of In-car Controls and Displays.
Automotive Ergonomics, Taylor and Francis, U.K., 1993, pp.299-320.
25. Horrey, W. J., and C. D. Wickens. In-Vehicle Glance Duration: Distributions, Tails,
and Model of Crash Risk. In Transportation Research Record: Journal of the
Transportation Research Board, No. 2018, Transportation Research Board of the
National Academies, Washington, D.C., 2007, pp. 22-28.
26. Klauer, S. G., T. A. Dingus, V. L. Neale, J. D. Sudweeks, and D. J. Ramsey. The
Impact of Driver Inattention on Near-Crash/Crash Risk: An Analysis Using the 100Car Naturalistic Driving Study Data. Publication DOT HS 810 594. Virginia Tech
Transportation Institute, Blacksburg, VA, 2006.
27. Sodhi, M., B. Reimer, and I. Llamazares. Glance Analysis of Driver Eye Movements
to Evaluate Distraction. Behavior Research Methods, Instruments, & Computers, Vol.
34, No. 4, 2002, pp. 529-538.
28. Beijer, D. Observed Driver Glance Behaviour at Roadside Advertising. Presented at
83rd Annual Meeting of the Transportation Research Board, Washington, D.C., 2004.
29. Smiley, A., T. Smahel, and M. Eizenmann. The Impact of Video Advertising on
Driver Fixation Patterns. Presented at 83rd Annual Meeting of the Transportation
Research Board, Washington, D.C., 2004.
30. Perez, W., and M. A. Bertola. The Effect of Visual Clutter on Driver Eye Glance
Behavior. Presented at Sixth International Driving Symposium on Human Factors in
Driver Assessment, Training and Vehicle Design, Lake Tahoe, CA, 2011.
31. Lee, S. E., M. J. McElheny, and R. Gibbons. Driving Performance and Digital
Billboards. Center for Automotive Safety Research, Virginia Tech Transportation
Institute, Blacksburg, VA, 2007.
32. Crundall, D., E. Van Loon, and G. Underwood. Attraction and Distraction of
Attention with Roadside Advertisements. Accident Analysis and Prevention, Vol. 38,
No. 4, 2006, pp. 671-677.
33. Megias, A., A. Maldonado, A. Catena, L. L. Di Stasi, J. Serrano, and A. Candido.
Modulation of Attention and Urgent Decisions by Affect-laden Roadside
167
Advertisement in Risky Driving Scenarios. Safety Science, Vol. 49, No. 10, 2011, pp.
1388-1393.
34. Owens, D. A., J. M. Stevenson, A. Osborn, and J. Geer. Differences in the Perception
of Potential Risk by Novice and Experienced Drivers. Presented at 89th Annual
Meeting of the Transportation Research Board, Washington, D.C., 2010.
35. Tantala, M. W., and A. M. Tantala. An Examination of the Relationship between
Digital Billboards and Traffic Safety in Reading, Pennsylvania Using Empirical
Bayes Analyses. Presented at Moving Toward Zero: 2011 ITE Technical Conference
and Exhibit, Lake Buena Vista, FL, 2011.
36. Smiley, A., B. Persaud, G. Bahar, C. Mollett, C. Lyon, T. Smahel, and W. L. Kelman.
Traffic Safety Evaluation of Video Advertising Signs. In Transportation Research
Record: Journal of the Transportation Research Board, No. 1937, Transportation
Research Board of the National Academies, Washington, D.C., 2005, pp. 105-112.
37. Edquist, J., T. Horberry, S. Hosking, and I. Johnston. Effects of Advertising
Billboards During Simulated Driving. Applied Ergonomics, Vol. 42, No. 4, 2011, pp.
619-626.
38. Young, M. S., J. M. Mahfoud, N. A. Stanton, P. M. Salmon, D. P. Jenkins, and G. H.
Walker. Conflicts of Interest: The Implications of Roadside Advertising for Driver
Attention. Transportation Research Part F: Traffic Psychology and Behaviour, Vol.
12, No. 5, 2009, pp. 381-388.
39. Jamson, A. H., and N. Merat. The Effectiveness of Safety Campaign Messages – A
Driving Simulator Investigation. Presented at Fourth International Driving
Symposium on Human Factors in Driver Assessment, Training and Vehicle Design,
Stevenson, WA, 2007.
40. Manual of Uniform Traffic Control Devices. 2009 Edition including Revisions 1 and
2. Federal Highway Administration, U.S. Department of Transportation, 2012.
41. Field, A. Discovering Statistics Using SPSS. 2nd Edition. Sage Publications Ltd.,
London, 2005.
42. Hinkle, D. E., W. Wiersma, and S. G. Jurs. Applied Statistics for the Behavioral
Sciences. 5th Edition. Houghton Mifflin Company, Boston, MA, 2003.
43. Molino, J. A., J. Wachtel, J. E. Farbry, M. B. Hermosillo, and T. M. Granda. The
Effects of Commercial Electronic Variable Message Signs (CEVMS) on Driver
Attention and Distraction: An Update. Publication FHWA-HRT-09-018. Human
Centered Systems Team, Office of Safety Research and Development, Federal
Highway Administration, U.S. Department of Transportation, 2009.
168
44. McAvoy, D. S. Work Zone Speed Reduction Utilizing Dynamic Speed Signs. Ohio
University, Athens, OH, 2011.
http://ntl.bts.gov/lib/42000/42200/42221/Dynamic_Speed_Sign_Final_Report_OU1.p
df. Accessed August 14, 2012.
45. AP Statistics Tutorial: Hypothesis Test for Difference Between Proportions. Stat
Trek. http://stattrek.com/hypothesis-test/difference-in-proportions.aspx?tutorial=ap.
Accessed January 13, 2011.
46. Glass, G. V., P. D. Peckham, and J. R. Sanders. Consequences of Failure to Meet
Assumptions Underlying the Fixed Effects Analyses of Variance and Covariance.
Review of Educational Research, Vol. 42, No. 3, 1972, pp. 237-288.
47. Profile of General Population and Housing Characteristics: 2010 Demographic
Profile Data. American FactFinder. United States Census Bureau.
http://factfinder2.census.gov/faces/tableservices/jsf/pages/productview.xhtml?fpt=tab
le. Accessed January 3, 2012.
48. Cerrelli, E. C. Crash Data and Rates for Age-Sex Groups of Drivers, 1996. National
Center for Statistics and Analysis, National Highway Traffic Safety Administration,
U.S. Department of Transportation, 1998. http://wwwnrd.nhtsa.dot.gov/Pubs/98.010.PDF. Accessed January 3, 2012.
169
APPENDIX A: PUBLIC SURVEY STUDY MATERIALS
Survey Questionnaire
1) Age:
______________ Years
2) Gender:
Female
Male
3) I am a licensed driver.
Yes
No
4) How long have you been driving?
1-5 Years
6-10 Years
11-15 Years
16-20 Years
20+
Years
5) Do you have any of the following driving limitations? (Check all that apply)
Visual impairments (Night time, color blindness, etc.)
Corrective lenses
Medical concerns
Drug use for medical condition(s)
Physical Limitations
6) I have been in ____ crashes while driving in Construction Work Zones.
No Crashes
1 Crash
2 Crashes
3 Crashes
4+
Crashes
7) I know the rules and regulations for driving in a Construction Work Zone.
Strongly
Agree
Neither agree
Disagree
agree
nor disagree
Strongly
disagree
8) Regarding the speed limit, I typically drive:
The limit to
6-10 MPH
11-15 MPH
5 MPH over
over
over
Below
the limit
16-20 MPH
over
9) Which sign would make you more aware of Construction Work Zones?
A
B
C
D
10) Which sign would make you more cautious in the Construction Work Zone?
A
B
C
D
11) If you were driving along a freeway at 65 mph, which sign would be able to see best
during the day?
A
B
C
D
170
12) If you were driving along a freeway at 65 mph, which sign would be able to see best
during the night?
A
B
C
D
13) Are the signs too cluttered with information?
Yes
No
171
A:
B:
C:
D:
172
A:
B:
C:
D:
173
APPENDIX B: IRB DOCUMENTS
OHIO UNIVERSITY
INSTITUTIONAL REVIEW BOARD (IRB)
PROJECT OUTLINE FORM
Proposal Title
1.
Safety Evaluation of the Configuration and Placement of Billboard
Advertisements on Driver Behavior in Work Zones
Investigator(s) Information
Primary Investigator Name
First
Patrick
Department
Address
Middle J.
Fry
Civil Engineering
Stocker 308
(If off-campus, include city, state and zip code)
Email
[email protected]
Training Module Completed?
Co-investigators
Name
Julie Owens
Address 243 Porter Hall
Email
[email protected]
Training Module Completed?
Advisor Information (if applicable)
Name
Last
Deborah McAvoy, Ph.D., P.E.,
P.T.O.E.
Address Stocker 118
Email
[email protected]
Training Module Completed?
Phone
614-419-5734
Yes
No
X
Department
Phone
740-593-1074
Yes
No
X
Department
Phone
Psychology
Civil Engineering
740-593-1468
Yes
No
X
174
2.
Study Timeline
a. Anticipated Starting Date (Study, including recruitment, cannot begin prior
to IRB approval. This date should never precede the submission date)
b. Duration of Study
3.
Years
Funding Status
a. Is the researcher receiving support or applying for funding?
If YES
List Source
February 1, 2011
Months 2
Yes
X
No
Laborers Union – 860 (Cleveland)
Describe any consulting or other
relationships with this sponsor.
Funding will be used for:
X
X
4.
Paying Participants (Provide further details in compensation section)
Researcher Expenses (Postage, Equipment, Travel, etc.)
Other
__Salaries_______________________________________________________
Review Level
Based on the definition in the guidelines, do you believe your research qualifies for?
Exempt Review – See description of categories at:
Category
X
http://www.ohio.edu/research/compliance/Exemption-Categories.cfm
Expedited Review - See description of categories at:
http://www.hhs.gov/ohrp/humansubjects/guidance/expedited98.htm
Full Board Review
Category
7
175
5.
Recruitment/Selection of Subjects
a. Maximum Number of Participants to be Enrolled – If screening occurs,
include number that will need to be screened in order to get the N necessary for
statistical significance.
b. Characteristics of subjects (check as many boxes as appropriate).
Minors
Disabled (Physically or Mentally)
Legally Incompetent
X Adults
Prisoners
Cognitively Impaired
Pregnant
Non-English Speaking
X
400
Elementary School Students
Middle School Students
High School Students
University Students
c. Briefly describe the criteria for selection of subjects (inclusion/exclusion). Include such
information as age range, health status, etc. Attach additional pages if necessary.
The criteria for selection of subjects is 18 years or older, two years of driving experience, and a
valid driver’s license. The subjects are limited to 18 years and older with two years of driving
experience due to the inexperience levels of new drivers.
d. Please describe how you will identify and recruit prospective participants.
Participants for the study will be solicited on voluntary basis by the PI working on the research
study through direct person to person contact or by phone from the community-at-large. Flyers
will be posted throughout the university requesting participants. Individuals will also be recruited
at businesses and community centers through the use of a pamphlet that will be left at the front
desk or with the business administrator. Participants will also be recruited from the Psych Pool.
The flyer/pamphlet has been provided in Appendix B.
It is anticipated approximately 50 individuals will be recruited per month.
e. Records
Are you accessing private, i.e. medical, educational, or employment
Yes
No
X
records?
If YES, Describe process for obtaining approval for the use of the records or for securing consent
from the subjects. Attach a letter of support from the holder or custodian of the records i.e.
primary physician, therapist, public school official.)
f. Please describe your relationship to the potential participants, i.e. instructor of class, co-worker,
etc. If no relationship, state no relationship.
No Relationship
Attach copies of all recruitment tools (advertisements, posters, etc.), label as APPENDIX B
g. Performance Sites/Location of Research
X
Ohio University Facility
Public Location
Other – Describe below and provide letters of cooperation and/or support
176
6.
Project Description
a. Please provide a brief summary of this project, using non-technical terms that would be
understood by a non-scientific reader. Please limit this description to no more than one
page, and provide details in the methodology section.
Traffic safety along the nation’s roadways is a major concern for traffic and safety
professionals. Engineering, educational and enforcement efforts have reduced the fatality
rate from 5.5 deaths per 100 million miles of travel in 1966 to 1.13 deaths in 2009. In
recent years there has been a change in roadway work zone projects from new roadway
construction to maintenance and rehabilitation of the existing infrastructure. This change
in construction activity has required work zone locations in which traffic flow must be
maintained. Operating work zones on existing roadways while maintaining traffic flow
places construction workers at a higher exposure to vehicle intrusion as well as exposing
the driving public to unexpected roadway conditions. The result is a decrease in safety for
both the work zone personnel and drivers. Advances in technology have helped to reduce
the number of fatalities occurring due to the maintenance of traffic through active work
zones. However, there are still over 700 fatalities in work zones every year.
Each year various organizations spend money on advertisement campaigns with the aim
of drawing public awareness to the issue of work zone safety. In order to ensure that an
advertisement campaign will be likely to increase public awareness of work zone safety,
a driving simulator study is proposed to test different billboard advertisements. Driving
simulators have been utilized for human factors and transportation engineering research
where they have proven to be a valuable and valid tool for understanding human
performance for a variety of applications. They are an especially effective tool when the
simulation model is able to represent, as closely as possible, real-world driving conditions
while allowing for accurate driver performance measurements. It is proposed to use a
high-fidelity driving simulator to determine the particular billboard advertisement and
placement which most effectively conveys the message of work zone safety to drivers.
b. Please describe the specific scientific objectives (aims) of this research and any previous
relevant research.
The objective of this research is to compare driver performance, and therefore work zone safety,
in work zones which have a billboard advertisement placed prior to the work zone. In order to
fulfill the objectives of the research needs, the project has been subdivided into individual tasks
culminating with a report outlining the results of the research conducted and direction for the
research. The tasks to fulfill the objectives of this research are as follows:
Conduct a simulator study to examine driver behavior in work zones when different billboard
slogans are utilized and compare the results to determine which most effectively increases safety.
Conduct a simulator study to examine driver behavior in work zones when different billboard
graphics are utilized and compare the results to determine which most effectively increases
safety.
Conduct a simulator study to examine driver behavior in a downtown setting and on an urban
freeway when different billboard placements are utilized along the roadway. An eye-tracking
system will be used to collect the driver’s gaze during the experiment in order to determine
which placement most effectively catches the driver’s attention.
177
c. Methodology: please describe the procedures (sequentially) that will be performed/followed
with human participants.
After recruitment, the participants will be introduced to the project using the following
narrative paragraph.
You have volunteered to be in a research study to compare the relative driving performance of
drivers in a controlled laboratory environment to understand the impact of various billboard
advertisements on driver behavior. The driving simulator has been built to represent the interior
of a standard size passenger automobile, including dashboard, steering wheel, gas pedal, and
brake pedal. As you drive the simulator, operate it as if you were driving an actual automobile.
The computer screens are intended to represent the actual images you would encounter while
looking through the windshield, door windows, and rear view mirrors. The scenes have been
programmed with images to replicate typical features encountered on the roadway, including
other vehicles present on the road, pedestrians, lane markings, traffic signs, traffic signals, etc.
You will drive through four scenarios. The first one is merely to get you acclimated to the
vehicle controls and computer-generated images. The first scenario should take about 15
minutes. Altogether, you will be driving for about 45 minutes. Your total time commitment for
this study is approximately 60 minutes; 45 minutes for driving and 15 minutes to answer the
questionnaire. Your driving performance will be monitored and statistics will be recorded on the
computer and quantified. Your name will remain confidential in the study and your performance
will be identified by a subject number only. Your participation in this study will be greatly
appreciated and may be a valuable benefit to society. Through your participation the safety
consequences along roadways will be quantified and the results may serve to improve safety
campaigns in the State of Ohio as well as across the nation, which ultimately aim to reduce traffic
crashes on road and highways as well as reduce taxes.
The participants will then be asked to fill out the informed consent form and the pre-test
questionnaire.
The participants will drive in the simulator for approximately 15 minutes in order to
become accustomed to the motion platform and simulated world.
If the participants desire to continue they will drive in the simulator for an additional 30
minutes. If the participants do not desire to continue in the study, they will be thanked
for their time and leave the study.
After completing the simulation participants will be asked to fill out a post-test
questionnaire.
The participants will be informed of the scope of the project as well as where and when
they will be able to find the results of the study.
178
d. Describe any potential risks or discomforts of participation and the steps that will be taken to
minimize them.
The risks involved in this experiment are minimal. Some participants may experience simulator
sickness after driving in the simulator for extended periods of time; mainly for driving periods
greater than one hour. Based upon previous studies conducted by the PI, it is anticipated that less
than one percent of the participants will experience slight to moderate simulator sickness. If a
participant indicates at any time that they are too uncomfortable to continue, they will be released
from the experiment. Also, crackers and water will be available for the participants to help with
any simulator sickness.
e. Describe the anticipated benefits to the individual participants. If none, state that. (Note that
compensation is not a benefit, but should be listed in the compensation section on the next page.)
The benefits for the individual participants will be the same as those for society.
f. Describe the anticipated benefits to society and/or the scientific community in lay language.
There must be some benefit to justify the use of human subjects.
Engineering, educational and enforcement efforts have reduced the fatality rate along the nation’s
highway from 5.5 deaths per 100 million miles of travel in 1966 to 1.13 deaths in 2009;
however, there are still more than 700 fatalities each year in work zones across the nation. The
trend of roadway construction toward rehabilitation of existing infrastructure and away from
building new roadways has placed work zone personnel and the driving public at an increased
risk. The aim of this research is to determine which billboard configuration and placement most
effectively increases work zone safety. These findings will present state departments of
transportation and other organizations a basis from which they can draw on when developing
their work zone safety public awareness campaigns. This in turn will increase the safety of
roadway work zones for both the driving public and work zone personnel and, ultimately, reduce
work zone crashes and save lives.
179
7.
Confidentiality
a. Check all that apply
Data is collected anonymously
Data will be recorded without possibility of identification (Computerized Data)
X
Data will be recorded with a code replacing identifiers and a master list connecting
the code and the identifier exists for some period of time
Data will be recorded with identifying information, e.g. name, SSN, oak id, etc.
Nature of data makes it potentially identifiable (e.g. material with DNA,
photographs)
b. If master code list is used (3rd option); please provide detail, such as how/where code
list is securely stored, when it will be destroyed, etc.).
N/A
c. If data is stored with identifiers, please provide details of how data will be stored
securely (i.e. locked cabinet, password protected, etc.) as well as timeframe of when data
will be de-identified.
N/A
d. Data Sharing
Will identifiable data be shared with anyone outside the immediate
research team?
If YES, please describe
Yes
e. Recording
Will participants be
No
X
Audio recorded?
Yes
No
X
Video recorded?
Yes
No
X
If YES, please describe how/where recordings will be stored, who will have access to them,
and an estimate of the date (month/year) that they will be destroyed.
f. Additional Details (if needed)
180
8.
Compensation
a. Will participants receive a gift or token of appreciation?
If YES, list the item and its approximate value.
Yes
No
X
b. Will participants receive services, treatment or supplies that have a
monetary value?
If YES, please describe and provide the approximate value.
Yes
No
X
c. Will participants receive course credit?
Yes X No
If YES, please describe non-research alternatives to earn the credit, the number of points awarded
and what percentage of total points for the course it represents. If you are using the Psychology
Pool, which has already established guidelines that provide these details to the IRB, simply write
Psych Pool.
Psych Pool, Non-Psych Pool participants will not receive compensation
d. Will participants receive monetary compensation (including gift
Yes
No
X
cards)?
If YES, please detail the amount per session and total compensation possible. Additionally,
describe what compensation amount is paid to participants who discontinue participation prior to
completion.*
*
If University funds are used to compensate participants, minimally, the name and address of
participants will need to be provided to the Finance Office at OU. If participants will be paid $100 or more
in a calendar year, participant social security numbers must be provided to Finance. The consent form must
reflect this.
9.
Instruments
a. List all questionnaires, instruments, standardized tests below, with a brief description,
and provide copies of each, labeled as APPENDIX C.
Experiment Questionnaire
The experiment questionnaire will be used as a pre-test survey to ask participants about their
driving experiences in work zones. The participants will also be asked to supply demographic
information to determine the representative nature of the participants with the population in terms
of age, gender, etc. Also the experiment questionnaire will be used as a post-test survey to ask
participants about the slogans utilized in the billboard advertisements that they drove past.
181
10.
Data Analysis
How will the data be analyzed? What statistical procedures will be used to test hypotheses; if
qualitative, how will data be coded, etc.
Demographic Data
This analysis will allow the researchers the ability to determine the extent to which the sample
population is representative of the overall driving population. The observed frequency
distribution or percentage of participant demographics will be compared with the corresponding
values of the expected distribution of the nation’s demographics. The intent of the comparison
will be to test whether the discrepancies between the observed and expected frequencies or
percentages were attributable to chance or were significantly different. If the discrepancies were
attributable to chance, then the differences between the two percentages can be deemed
statistically insignificant. The statistical analysis to determine if the demographic data (gender or
age) in the sample population was significantly different than the population will be the test of
goodness-of-fit or the chi-square test. The null hypothesis for the chi-square test will be as
follows: There was no difference between the demographic data (gender or age) of the simulator
experiment sample and the nation’s population.
Speed Data
The analysis of speed data will be used as an indication of a motorist’s perceived risk of traveling
along a roadway. The speed data will be analyzed using the overall mean speed, the mean speeds
prior to and after the presence of the billboards, the mean speed with a billboard present and
without a billboard present, as well as the variance of the speeds will be calculated. Statistical
tests will be used to determine if the mean speeds and speed variances between the billboard
scenarios are statistically significant. The null hypothesis for the scenarios speed analysis will be
that there is no difference between the mean speeds of the scenarios. The Paired t-test will be
utilized to determine the effectiveness of the billboards based upon the comparison of the mean
speeds and speed variances. If the data collected violates normality for the speed data, the mean
speeds will be compared with the Paired t-test’s nonparametric counterpart, the Mann-Whitney U
Test.
Acceleration and Deceleration Data
The acceleration and deceleration data will be used as an indication of a motorist’s potential for a
crash as well as their reaction to the billboards. Acceleration and deceleration are recorded as
normalized values with a range between 0.00 and 1.00. A value of 0.00 indicates that the pedal is
not depressed while a value of 1.00 indicates that the pedal is at maximum depression. The
acceleration data will be analyzed using the overall mean acceleration, the mean acceleration
prior to and after the presence of the billboards, the mean acceleration with a billboard present
and without a billboard present, as well as the variance of the accelerations will be calculated.
Likewise, the deceleration data will be analyzed and calculated in a similar manner. Statistical
tests will be conducted to determine if the mean acceleration and mean deceleration between the
billboard scenarios are statistically significant. The null hypothesis for the acceleration and
deceleration is that there is no difference between the acceleration and deceleration of vehicles.
The Paired t-test will be utilized to determine the effectiveness of the billboard advertisement
based upon the comparison of the overall mean acceleration and deceleration and the
corresponding variances. If the data collected violates normality for the acceleration and
deceleration data, the means will be compared with the Paired t-test’s nonparametric counterpart,
the Mann-Whitney U Test.
182
Collision Data
The analysis of collision data will be used as an indication of a motorist’s actual risk of traveling
along a roadway and potential distraction related to the billboards. Collision data is recorded on
the simulator control station if the participant collides with another object. From this data it can
be determined when the collision occurred, where in the work zone the collision occurred, and
with what object the subject vehicle collided. The observed frequency distribution or percentage
of collisions will be compared among the billboard scenarios to the test scenario without a
billboard. The intent of the comparison will be to test whether the discrepancies between the
billboard scenario crashes and non-billboard scenario crashes were attributable to chance or were
significantly different. If the discrepancies were attributable to chance, then the differences
between the two percentages can be deemed statistically insignificant. The statistical analysis to
determine if the collision frequency data in the sample population was significantly different will
be the test of goodness-of-fit or the chi-square test. The null hypothesis for the chi-square test
will be as follows: There was no difference between the collision frequency data of the simulator
experiment sample.
Detection Distance Data
Detection distance will be used as an indication of the participant’s ability to see the billboard.
The participant’s gaze relative to the simulator screens is recorded on the eye-tracking control
station. Where their gaze intersects the billboard for a significant period of time represents the
participant detecting the billboard ahead. This data will be coordinated with the vehicle position
data recorded on the simulator control station to determine at what distance from the billboard
can the participant detect the sign. Statistical tests will be conducted to determine if the detection
distance between the billboard scenarios are statistically significant. The null hypothesis is that
there is no difference between the detection distance for the billboard scenarios. The Paired t-test
will be utilized to determine the effectiveness of the billboard advertisement based upon the
comparison of the detection distance. If the data collected violates normality for detection
distance, the lateral placement data will be compared with the Paired t-test’s nonparametric
counterpart, the Mann-Whitney U Test.
Time of Gaze Data
Time of gaze will be used as an indication of the time required for the participant’s cognition of
the billboard. The length of time that the participant’s gaze is fixated on the billboard is recorded
on the eye-tracking control station. The time of gaze for each billboard will be analyzed.
Statistical tests will be conducted to determine if the time of gaze among the various billboards
are statistically significant. The null hypothesis is that there is no difference between the time of
gaze for the billboards. The Paired t-test will be utilized to determine the effectiveness of the
billboard advertisement based upon the comparison of the time of gaze for the billboards. If the
data collected violates normality for time of gaze, the time of gaze data will be compared with the
Paired t-test’s nonparametric counterpart, the Mann-Whitney U Test.
183
11.
Informed Consent Process
Select One of the Following Options
I am obtaining signed consent for this study (Attach copies of all consent documents as
X
Appendix A, using the template provided at the end of this document.
I am requesting a waiver or alteration of Informed Consent (provide details below and
attach information that will be provided to participants regarding the study (email, cover
letter) as Appendix A.
Waiver of signature
___ Exempt study
___ Waiver needed to protect the privacy of participants
___ Waiver needed due to cultural norms (e.g. wary of forms needing
signatures)
___ Impracticable (online or phone study)
___ Other
________________________________________________________________
Deception (incomplete disclosure)
___ Necessary to avoid participants altering behavior (e.g. not informing of 2
way mirror; providing cover story )
___
Complete waiver of consent
Provide additional information regarding the waiver, if needed.
Attach copies of all consent documents or text and label as APPENDIX A. Please use the
template provided at the end of this document.
b. How and where will the consent process occur? Will participants have an opportunity to ask
questions and have them answered? What steps will be taken to avoid coercion or undue
influence?
The potential participants in the study will be debriefed regarding the study as outlined above and
in the Appendices. They will be given the informed consent form to read, review and signed, if so
desired. Prior to signing the document, they will be allowed to ask any questions regarding the
experiment with appropriate, honest answers provided. The participants will not be forced or
asked to proceed with the experiment unless they are completely comfortable with the task.
The consent process will occur in the Safety and Human Factors Facility (Stocker 308) after the
participant has initially agreed to participate in the study. The participant will be given a verbal
briefing regarding the project and experiment. At that point, the participant will be provided the
informed consent form to read, review and sign. The participants will be encouraged to ask any
questions that they may have regarding the experiment and the task they may be completing
before they sign the consent form and throughout the study. The participants will also be made
aware that at any time they may withdraw from participating in the project without ramifications.
184
c. Will the investigator(s) be obtaining all of the informed consents?
Yes
X
No
X
No
If NO, identify by name and training who will be describing the research to
subjects/representatives and inviting their participation?
d. Will all adult participants have the capacity to give informed
consent?
Yes. This is demonstrated by their ability to obtain a driver’s license.
If NO, explain procedures to be followed.
Yes
e. Will any participants be minors?
Yes
No
If YES, include procedures/form for parental consent and for the assent from the minor.
f. Will participants be deceived or incompletely informed regarding
any aspect of the study?
Yes
No
If YES, provide rationale for use of deception.
If YES, attach copies of post-study debriefing information and label as APPENDIX D.
Additionally, complete the questions related to a consent form waiver or alteration on page 11.
X
X
185
Investigator Assurance
I certify that the information provided in this outline form is complete and correct.
I understand that as Principal Investigator, I have ultimate responsibility for the
protection of the rights and welfare of human subjects, conduct of the study and the
ethical performance of the project.
I agree to comply with Ohio University policies on research and investigation involving
human subjects (O.U. Policy # 19.052), as well as with all applicable federal, state and
local laws regarding the protection of human subjects in research, including, but not
limited to the following:
The project will be performed by qualified personnel, according to the OU approved protocol.
No changes will be made in the protocol or consent form until approved by the OU IRB.
Legally effective informed consent will be obtained from human subjects if applicable, and
documentation of informed consent will be retained, in a secure environment, for three years after
termination of the project.
Adverse/Unexpected events will be reported to the OU IRB promptly.
All protocols are approved for a maximum period of one year. Research must stop at the end of
that approval period unless the protocol is re-approved for another term.
I further certify that the proposed research is not currently underway and will not begin
until approval has been obtained. A signed approval form, on Office of Research
Compliance letterhead, communicates IRB approval.
Primary Investigator Signature
Date
(Please print name)
Co-Investigator Signature
(Please print name)
Date
186
Faculty Advisor/Sponsor Assurance
By my signature as sponsor on this research application, I certify that the student(s) or guest
investigator is knowledgeable about the regulations and policies governing research with human
subjects and has sufficient training and experience to conduct this particular study in accord with
the approved protocol. In addition:
I agree to meet with the investigator(s) on a regular basis to monitor study progress.
Should problems arise during the course of the study, I agree to be available, personally, to
supervise the investigator in solving them.
I assure that the investigator will report adverse/unexpected events to the IRB in writing
promptly.
If I will be unavailable, as when on sabbatical or vacation, I will arrange for an alternate faculty
sponsor to assume responsibility during my absence.
I further certify that the proposed research is not currently underway and will not begin until
approval has been obtained. A signed approval form, on Office of Research Compliance
letterhead, communicates IRB approval.
Advisor/Faculty Sponsor Signature
Date
(Please print name)
*The faculty advisor/sponsor must be a member of the OU faculty. The faculty member is
considered the responsible party for legal and ethical performance of the project.
187
Ohio University Consent Form
Non-Psych Pool Participants
Title of Research: Safety Evaluation of the Configuration and Placement of Billboard
Advertisements on Driver Behavior in Work Zones
Researchers: Patrick Fry, Graduate Research Assistant; Deborah McAvoy, Ph.D., P.E.,
P.T.O.E., Assistant Professor
You are being asked to participate in research. For you to be able to decide whether you
want to participate in this project, you should understand what the project is about, as
well as the possible risks and benefits in order to make an informed decision. This
process is known as informed consent. This form describes the purpose, procedures,
possible benefits, and risks. It also explains how your personal information will be used
and protected. Once you have read this form and your questions about the study are
answered, you will be asked to sign it. This will allow your participation in this study.
You should receive a copy of this document to take with you.
Explanation of Study
You have volunteered to be in a research study to compare the relative driving
performance of drivers in a controlled laboratory environment to understand the impact
of various billboard advertisements on driver behavior. The driving simulator has been
built to represent the interior of a standard size passenger automobile, including
dashboard, steering wheel, gas pedal, and brake pedal. As you drive the simulator,
operate it as if you were driving an actual automobile. The computer screens are
intended to represent the actual images you would encounter while looking through the
windshield, door windows, and rear view mirrors. The scenes have been programmed
with images to replicate typical features encountered on the roadway, including other
vehicles present on the road, pedestrians, lane markings, traffic signs, traffic signals, etc.
You will drive through four scenarios. The first one is merely to get you acclimated to the
vehicle controls and computer-generated images. The first scenario should take about 15
minutes. Altogether, you will be driving for about 45 minutes. Your driving performance
will be monitored and statistics will be recorded on the computer and quantified. Your
name will remain confidential in the study and your performance will be identified by a
subject number only. Your participation in this study will be greatly appreciated and may
be a valuable benefit to society. Through your participation the safety consequences
along roadways will be quantified and the results may serve to improve safety campaigns
in the State of Ohio as well as across the nation, which ultimately aim to reduce traffic
crashes on road and highways as well as reduce taxes.
We are also requesting that you fill out a questionnaire regarding your demographics
and past driving habits. We will use the demographic data to determine if the results from
this study can be applied to the nation’s population. If the demographic data for the
survey participants are significantly different than the nation’s demographic data, we
may not be able to generalize the results of this project. We are attempting to assure we
188
have an adequate representation so that we can draw significant conclusions from this
project. Your total time commitment for this study is approximately 60 minutes; 45
minutes for driving, 10 minutes to answer the pre-test questionnaire and 5 minutes to
answer the post-test questionnaire.
Risks and Discomforts
The risks to which you will be exposed by participating in the experiment are minimal.
The risks are as follows:
1. Simulator sickness due to driving in a simulator; generally less than one percent
of participants experience nausea and a headache at the onset of driving or after
driving for an hour or more.
2. Discomfort while sitting in the simulator for an extended period of time.
While these risks generally occur in less than one percent of participants, the following
precautions will be taken to ensure minimal risk to you:
1. You have the right to withdraw from the experiment at any time.
2. You will be allowed to take up to a two-minute break in between driving sessions
to alleviate any discomfort you may experience due to sitting for an extended
period of time.
3. The length of the driving simulation has been kept to one hour.
4. Crackers and water will be available to participants who experience motion
sickness.
Benefits
The current trend in roadway construction is to repair existing infrastructure, thus
placing work zone personnel and the driving public at an increased risk. The aim of this
research is to determine which billboard configuration and placement most effectively
increases work zone safety. These findings will present organizations with a resource
they can use when developing their work zone safety public awareness campaigns.
Therefore this project will assist in reducing the number of work zone crashes and saving
the lives of work zone personnel and the driving public.
Confidentiality and Records
The data collected from the experiment will be identified by a time stamp including date
and time of travel run. Your completed questionnaires will be also be linked by a time
stamp that will correspond to the collected data. Your name will not appear in any
document or tape related to this research. Participation in this study is completely
confidential.
Additionally, while every effort will be made to keep your study-related information
confidential, there may be circumstances where this information must be shared with:
* Federal agencies, for example the Office of Human Research Protections,
whose responsibility is to protect human subjects in research;
189
* Representatives of Ohio University (OU), including the Institutional Review
Board, a committee that oversees the research at OU.
Compensation
No compensation will be provided.
Contact Information
If you have any questions regarding this study, please contact:
Pat Fry, Graduate Research Assistant, By Email: [email protected]
Deborah McAvoy, Ph.D., P.E., P.T.O.E., By Email: [email protected]
If you have any questions regarding your rights as a research participant, please contact
Jo Ellen Sherow, Director of Research Compliance, Ohio University, (740)593-0664.
By signing below, you are agreeing that:
 you have read this consent form (or it has been read to you) and have been given
the opportunity to ask questions and have them answered
 you have been informed of potential risks and they have been explained to your
satisfaction.
 you understand Ohio University has no funds set aside for any injuries you might
receive as a result of participating in this study
 you are 18 years of age or older
 your participation in this research is completely voluntary
 you may leave the study at any time. If you decide to stop participating in the
study, there will be no penalty to you and you will not lose any benefits to which
you are otherwise entitled.
Signature
Date
Printed Name
Version Date: 01/18/10
190
Ohio University Consent Form
Psych Pool Participants
Title of Research: Safety Evaluation of the Configuration and Placement of Billboard
Advertisements on Driver Behavior in Work Zones
Researchers: Patrick Fry, Graduate Research Assistant; Deborah McAvoy, Ph.D., P.E.,
P.T.O.E., Assistant Professor
You are being asked to participate in research. For you to be able to decide whether you
want to participate in this project, you should understand what the project is about, as
well as the possible risks and benefits in order to make an informed decision. This
process is known as informed consent. This form describes the purpose, procedures,
possible benefits, and risks. It also explains how your personal information will be used
and protected. Once you have read this form and your questions about the study are
answered, you will be asked to sign it. This will allow your participation in this study.
You should receive a copy of this document to take with you.
Explanation of Study
You have volunteered to be in a research study to compare the relative driving
performance of drivers in a controlled laboratory environment to understand the impact
of various billboard advertisements on driver behavior. The driving simulator has been
built to represent the interior of a standard size passenger automobile, including
dashboard, steering wheel, gas pedal, and brake pedal. As you drive the simulator,
operate it as if you were driving an actual automobile. The computer screens are
intended to represent the actual images you would encounter while looking through the
windshield, door windows, and rear view mirrors. The scenes have been programmed
with images to replicate typical features encountered on the roadway, including other
vehicles present on the road, pedestrians, lane markings, traffic signs, traffic signals, etc.
You will drive through four scenarios. The first one is merely to get you acclimated to the
vehicle controls and computer-generated images. The first scenario should take about 15
minutes. Altogether, you will be driving for about 45 minutes. Your driving performance
will be monitored and statistics will be recorded on the computer and quantified. Your
name will remain confidential in the study and your performance will be identified by a
subject number only. Your participation in this study will be greatly appreciated and may
be a valuable benefit to society. Through your participation the safety consequences
along roadways will be quantified and the results may serve to improve safety campaigns
in the State of Ohio as well as across the nation, which ultimately aim to reduce traffic
crashes on road and highways as well as reduce taxes.
We are also requesting that you fill out a questionnaire regarding your demographics
and past driving habits. We will use the demographic data to determine if the results from
this study can be applied to the nation’s population. If the demographic data for the
survey participants are significantly different than the nation’s demographic data, we
may not be able to generalize the results of this project. We are attempting to assure we
191
have an adequate representation so that we can draw significant conclusions from this
project. Your total time commitment for this study is approximately 60 minutes; 45
minutes for driving, 10 minutes to answer the pre-test questionnaire and 5 minutes to
answer the post-test questionnaire.
Risks and Discomforts
The risks to which you will be exposed by participating in the experiment are minimal.
The risks are as follows:
1. Simulator sickness due to driving in a simulator; generally less than one percent
of participants experience nausea and a headache at the onset of driving or after
driving for an hour or more.
2. Discomfort while sitting in the simulator for an extended period of time.
While these risks generally occur in less than one percent of participants, the following
precautions will be taken to ensure minimal risk to you:
1. You have the right to withdraw from the experiment at any time.
2. You will be allowed to take up to a two-minute break in between driving sessions
to alleviate any discomfort you may experience due to sitting for an extended
period of time.
3. The length of the driving simulation has been kept to one hour.
4. Crackers and water will be available to participants who experience motion
sickness.
Benefits
The current trend in roadway construction is to repair existing infrastructure, thus
placing work zone personnel and the driving public at an increased risk. The aim of this
research is to determine which billboard configuration and placement most effectively
increases work zone safety. These findings will present organizations with a resource
they can use when developing their work zone safety public awareness campaigns.
Therefore this project will assist in reducing the number of work zone crashes and saving
the lives of work zone personnel and the driving public.
Confidentiality and Records
The data collected from the experiment will be identified by a time stamp including date
and time of travel run. Your completed questionnaires will be also be linked by a time
stamp that will correspond to the collected data. Your name will not appear in any
document or tape related to this research. Participation in this study is completely
confidential.
Additionally, while every effort will be made to keep your study-related information
confidential, there may be circumstances where this information must be shared with:
* Federal agencies, for example the Office of Human Research Protections,
whose responsibility is to protect human subjects in research;
192
* Representatives of Ohio University (OU), including the Institutional Review
Board, a committee that oversees the research at OU.
Compensation
For participation in this study you will receive one research credit. This credit will be
received even if you withdraw from the study prior to its completion.
Contact Information
If you have any questions regarding this study, please contact:
Pat Fry, Graduate Research Assistant, By Email: [email protected]
Deborah McAvoy, Ph.D., P.E., P.T.O.E., By Email: [email protected]
If you have any questions regarding your rights as a research participant, please contact
Jo Ellen Sherow, Director of Research Compliance, Ohio University, (740)593-0664.
By signing below, you are agreeing that:
 you have read this consent form (or it has been read to you) and have been given
the opportunity to ask questions and have them answered
 you have been informed of potential risks and they have been explained to your
satisfaction.
 you understand Ohio University has no funds set aside for any injuries you might
receive as a result of participating in this study
 you are 18 years of age or older
 your participation in this research is completely voluntary
 you may leave the study at any time. If you decide to stop participating in the
study, there will be no penalty to you and you will not lose any benefits to which
you are otherwise entitled.
Signature
Date
Printed Name
Version Date: 01/18/10
193
Date: _________________
Pre-Test Questionnaire
1) Gender:
Female
Male
2) Age:
16-20
21-25
26-35
36-45
46-60
61-70
3) Do you have any of the following driving limitations? (Check all that apply)
Visual impairments (Night time, color blindness, etc.)
Corrective lenses
Medical concerns
Drug use for medical condition(s)
Physical Limitations
4) How much total time do you spend driving on a typical day on your way to and from
work/school?
0.0-0.5 hours
0.5-1.0 hours
1.0-1.5 hours
1.5-2.0 hours
2.0-2.5 hours
Over 2.5 hours
5) How many miles do you drive in one direction on a typical day to arrive at
work/school?
0-5 miles
5-10 miles
11-20 miles
21-30 miles
31-40 miles
Over 41 miles
6) I know the rules and regulations for driving in a Construction Work Zone.
Strongly
Agree
Neither agree
Disagree
agree
nor disagree
Strongly
disagree
7) Do you know of any safety campaigns conducted by the State of Ohio or any road
agencies?
Yes
No
Do you recall the slogan? (Any words are appropriate)
__________________________________________
8) Would a safety campaign informing you of the hazards in a work zone alter your
driving behavior?
Yes
No
If yes, check all that apply:
Reduce speed
Increase speed
Pay closer attention to workers
Drive farther away from the work zone or barrels
Drive closer to the work zone or barrels
Other: _________________________________
194
Post-Test Questionnaire
Date: _________________
1) Are the signs too cluttered with information?
Yes
No
2) Which of the slogans do you remember best?
195
196
Ohio University
Institutional Review Board
Periodic Review Form
Proposal #
11X007
Proposal Title
Funding Source
Safety Evaluation of the Configuration and Placement of Billboard
Advertisements on Driver Behavior in Work Zones
Principal Investigator Information
Name
Patrick Fry
Address
Email
Laborers Union – 860
(Cleveland)
Department
Civil Engineering
Stocker 308
[email protected]
Phone
614-419-5734
Indicate Study Status:
X
Open to continuing enrollment of new participants
Enrollment closed, plan to re-open enrollment once approved
Enrollment closed, participants still receiving treatment/intervention
Enrollment closed, only data analysis occurring on identifiable data
Completed (no enrollment, no treatment/intervention, data has no identifiers)
Provide the total number of participants enrolled in the study, to date:
70
1. Summarize all amendment submissions approved by IRB (after original approval):
None
2. Are there any revisions to be considered in this review?
If yes, please respond to items a – d below.
YES
a. Describe the proposed changes and why they are being made.
b. Describe how, if at all, the proposed changes affect the risks of the study.
NO
X
197
c. Describe how, if at all, the proposed changes affect the benefits of the study.
d. Does the proposed change affect the consent/assent document(s)?
YES
If yes:
NO
Will any participants need to be re-consented as a result of the changes? If so,
please describe process to be used. Include two copies of the revised
consent/assent documents, one with changes highlighted, and one without
highlighting.
3. Provide a synopsis of the results to date (include the progress of the study as compared to the
hypothesis). If the risk/benefit assessment has been altered based on the results obtained
from the study thus far, describe.
The simulator study has been conducted with 70 participants and the prescribed data
was collected. The results for each of the 70 participants have been determined on an
individual basis; however, statistical analysis of the entire dataset still needs to be
conducted at this time.
It is anticipated that more participants will need to be enrolled in the study in order to
expand the age demographics of the sample population. Currently, only participants in
the 18-24 age range have participated in the study at this time. In order for the study to
be more representative of the actual population, further recruitment of participants in
other age groups is necessary.
4. Have there been any:
Adverse events or unanticipated results?
Withdrawal of subjects from research?
Complaints about the research?
Enrollment Problems?
Literature, findings, or other information that has become
available since starting study that indicates a need to amend
the study?
Changes to funding status?
Yes
Yes
Yes
Yes
Yes
No
No
No
No
No
X
X
X
X
X
Yes
No
X
a. If you answered ‘yes’ to any of the above questions, please explain below or attach
explanatory material.
198
5. Provide a copy of all currently approved informed consent documents, assent documents,
and a copy of any debriefing information, if applicable. Please do not submit any
document that contains participant signatures.
Principal Investigator Signature
Date
Advisor Signature
Date
Please return this form to: Office of Research Compliance
117 Research & Technology Center
Ohio University
Athens, OH 45701-2979
You may also scan the signed form and email it to [email protected]
199
200
APPENDIX C: BILLBOARD ADVERTISEMENTS USED IN SIMULATOR STUDY
Scenario #1 Billboards
Version A
201
Version B
Version C
202
Version C (Continued)
Version D
203
Version D (Continued)
Scenario #2 Billboards
*Also used for Scenario #3
204
APPENDIX D: RESULTS OF THE GAMES-HOWELL TESTS ON THE DATA
COLLECTED IN THE DRIVING SIMULATOR STUDY
Scenario #1 – Slogans
Results of Games-Howell Test on Lane Position in Pre-Work Zone
95 percent LOC
Interval
Mean
Standard
Comparison
Difference
Error
Lower
Upper
Bound
Bound
Slogan 1
-0.383
0.165
-0.963
0.197
Slogan 2
-0.289
0.128
-0.736
0.158
Slogan 3
-0.382
0.143
-0.881
0.118
Slogan 4
-0.144
0.139
-0.632
0.344
Slogan 5
-0.382
0.138
-0.865
0.101
Slogan 6
-0.154
0.143
-0.654
0.345
Control
Slogan 7
-0.102
0.132
-0.564
0.360
Slogan 8
-0.460
0.143
-0.962
0.041
Slogan 9
-0.342
0.121
-0.762
0.079
Slogan 10
-0.199
0.148
-0.718
0.320
Slogan 11
-0.061
0.135
-0.532
0.409
Slogan 12
-0.239
0.142
-0.736
0.258
Results of Games-Howell Test on Mean Fixation in Pre-Work Zone
95 percent LOC
Interval
Mean
Standard
Comparison
Difference
Error
Lower
Upper
Bound
Bound
Slogan 1
-0.622
0.150
-1.154
-0.089
Slogan 2
-0.855
0.198
-1.566
-0.144
Slogan 3
-0.639
0.123
-1.074
-0.204
Slogan 4
-0.465
0.136
-0.949
0.019
Slogan 5
-0.540
0.133
-1.010
-0.070
Slogan 6
-0.673
0.150
-1.207
-0.139
Control
Slogan 7
-0.491
0.133
-0.963
-0.020
Slogan 8
-0.821
0.239
-1.681
0.039
Slogan 9
-0.654
0.117
-1.067
-0.240
Slogan 10
-0.603
0.126
-1.049
-0.157
Slogan 11
-0.475
0.121
-0.905
-0.045
Slogan 12
-0.457
0.088
-0.766
-0.148
Result
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Result
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
205
Scenario #1 – Slogans (Continued)
Results of Games-Howell Test on Total Fixation in Pre-Work Zone
95 percent LOC
Interval
Mean
Standard
Comparison
Difference
Error
Lower
Upper
Bound
Bound
Slogan 1
-1.589
0.357
-2.855
-0.322
Slogan 2
-1.610
0.335
-2.801
-0.419
Slogan 3
-1.856
0.381
-3.215
-0.496
Slogan 4
-1.680
0.361
-2.965
-0.394
Slogan 5
-1.473
0.355
-2.735
-0.212
Slogan 6
-1.539
0.344
-2.764
-0.315
Control
Slogan 7
-1.646
0.321
-2.784
-0.509
Slogan 8
-1.633
0.342
-2.850
-0.415
Slogan 9
-1.863
0.336
-3.055
-0.671
Slogan 10
-1.432
0.312
-2.539
-0.325
Slogan 11
-1.419
0.334
-2.605
-0.233
Slogan 12
-1.295
0.306
-2.378
-0.213
Results of Games-Howell Test on Number of Fixations in Pre-Work Zone
95 percent LOC
Interval
Mean
Standard
Comparison
Difference
Error
Lower
Upper
Bound
Bound
Slogan 1
-0.685
0.313
-1.780
0.410
Slogan 2
-0.660
0.306
-1.730
0.410
Slogan 3
-0.695
0.298
-1.740
0.350
Slogan 4
-0.971
0.377
-2.300
0.360
Slogan 5
-0.651
0.290
-1.660
0.360
Slogan 6
-0.695
0.294
-1.730
0.340
Control
Slogan 7
-1.143
0.364
-2.430
0.140
Slogan 8
-0.833
0.324
-1.970
0.310
Slogan 9
-0.951
0.336
-2.140
0.230
Slogan 10
-0.592
0.269
-1.530
0.350
Slogan 11
-0.660
0.306
-1.730
0.410
Slogan 12
-0.798
0.320
-1.920
0.330
Result
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Reject Null
Result
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
Accept Null
206
Scenario #1 – Slogans (Continued)
Results of Games-Howell Test on Proportion Fixation Duration in Pre-Work Zone
95 percent LOC
Interval
Mean
Standard
Comparison
Result
Difference
Error
Lower
Upper
Bound
Bound
Slogan 1
-0.153
0.035
-0.276
-0.030 Reject Null
Slogan 2
-0.156
0.032
-0.268
-0.043 Reject Null
Slogan 3
-0.183
0.036
-0.313
-0.054 Reject Null
Slogan 4
-0.165
0.036
-0.292
-0.037 Reject Null
Slogan 5
-0.147
0.035
-0.272
-0.022 Reject Null
Slogan 6
-0.151
0.033
-0.270
-0.032 Reject Null
Control
Slogan 7
-0.160
0.032
-0.272
-0.048 Reject Null
Slogan 8
-0.160
0.033
-0.275
-0.044 Reject Null
Slogan 9
-0.183
0.033
-0.299
-0.067 Reject Null
Slogan 10
-0.142
0.031
-0.250
-0.033 Reject Null
Slogan 11
-0.136
0.033
-0.251
-0.020 Reject Null
Slogan 12
-0.129
0.029
-0.232
-0.027 Reject Null
Scenario #2 – Graphics
Results of Games-Howell Test on Mean Fixation in Pre-Work Zone
95 percent LOC
Interval
Mean
Standard
Comparison
Difference
Error
Lower
Upper
Bound
Bound
Logo &
Barrel
-0.575
0.137
-0.940
-0.210
Control
Logo
-0.590
0.157
-1.010
-0.170
Barrels
-0.586
0.125
-0.920
-0.250
Result
Reject Null
Reject Null
Reject Null
207
Scenario #2 – Graphics (Continued)
Results of Games-Howell Test on Total Fixation in Pre-Work Zone
95 percent LOC
Interval
Mean
Standard
Comparison
Difference
Error
Lower
Upper
Bound
Bound
Logo &
Barrel
-1.696
0.357
-2.650
-0.740
Control
Logo
-1.613
0.361
-2.580
-0.650
Barrels
-1.745
0.346
-2.670
-0.820
Results of Games-Howell Test on Number of Fixations in Pre-Work Zone
95 percent LOC
Interval
Mean
Standard
Comparison
Difference
Error
Lower
Upper
Bound
Bound
Logo &
Barrel
-1.200
0.390
-2.230
-0.170
Control
Logo
-1.267
0.446
-2.450
-0.080
Barrels
-1.400
0.375
-2.390
-0.410
Result
Reject Null
Reject Null
Reject Null
Result
Reject Null
Reject Null
Reject Null
Results of Games-Howell Test on Proportion Fixation Duration in Pre-Work Zone
95 percent LOC
Interval
Mean
Standard
Comparison
Result
Difference
Error
Lower
Upper
Bound
Bound
Logo &
Barrel
-0.159
0.034
-0.249
-0.069 Reject Null
Control
Logo
-0.151
0.034
-0.242
-0.061 Reject Null
Barrels
-0.165
0.033
-0.253
-0.076 Reject Null
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
Thesis and Dissertation Services
!