INVESTIGATING THE
ROLE OF DRIVER DECISION
FINAL REPORT
METRANS Project 00-11
December 2003
Department of Civil and
Environmental Engineering
And
Epstein Department of Industrial
and Systems Engineering
Epstein Department of
Industrial and Systems Engineering
Michael J. Driver
Department of Management and
Organization
Marshall School of Business
The contents of
this report reflect the views of the authors, who are responsible for the facts
and the accuracy of the information presented herein. This document is disseminated under the
sponsorship of the Department of Transportation, University Transportation
Centers Program, and California Department of Transportation in the interest of
information exchange. The U.S.
Government and California Department of Transportation assume no liability for
the contents or use thereof. The
contents do not necessarily reflect the official views or policies of the State
of
This
research was designed to take a closer look at the ways by which driver
decision-making styles affect highway-rail crossing (HRC) accidents. That is, a simplistic approach of portraying
human error, as the cause of most HRC accidents, needs to be augmented with a
more complex theory of human decision-making process while performing driving
tasks before and during a highway-rail intersection. Video and still photos were taken to identify
the intersections appropriate for this study.
The intersections were among many in the
PREVIOUS RESEARCH
ON HRC SAFETY
INFORMATION
PROCESING BEHAVIOR: THE DECISION STYLE MODEL
INFORMATION
COMPLEXITY AND UNCERTAINTY
2. MTHODOLOGY AND STATISTICAL ANALYSIS
Defining Environmental Complexity
3. CONCLUSIONS AND RECOMMENDATIONS
Table 1. ANOVA Table for the Experimental Trials 7
Table 2. Objects in Intersections (Low vs. High Env. Complexity) 9
Table 3. The effect of Decision Style, LEC, and Recall 10
Table 4. The effect of Decision Style, LEC, and Recognition 10
Table 5. The effect of Decision Style, HEC, and Recall 11
Table 6. The effect of Decision Style, HEC, and Recognition 11
Table 7. Chi-Square Test Table (Time Pressure Vs. Env. Complexity) 12
Table 8. Environmental Complexity Objects and Their Assigned Weights 13
Table 9. The Results of Chi-Square Tests for LEC 14
Table 10. The Results of Chi-Square Tests for HEC 14
Funding Source:
Names of principal investigators:
Michael J. Driver
Total Dollar Amount:
$88,062
Title of Project:
Investigating the Role of Driver Decision Styles in
Highway-Rail Crossing Accidents
Relationship to report:
The
Subcontract statement:
This project was funded in
entirety under this contract to California Department of Transportation
This research was entirely
supported by USC/CSULB Metrans Center, its funding agencies, the Department of
Transportation and California Department of Transportation.
The research team thanks the
Los Angeles Metropolitan Transit Authority (MTA) and its associated agencies
for providing valuable information throughout this research. We would especially like to express our
appreciation to Abdul K. Zohbi, Vijay Khawani, Essam Aly, Anthony Loui, Paul J.
Lennon, and Roger Snoble for taking time from their
busy schedules to meet with us, provide information and answers to our
queries. The support of MTA contractor,
Mr. Jim Curry of Parsons, Brinckerhoff Quade & Douglas is also appreciated.
We express our gratitude to
the following individuals for explicitly expressing their support of our research
on grade crossing safety: Ms. Yvonne Brathwaite
Burke, Supervisor of the Second District, County of Los Angeles, and her
assistant, Mr. Michael Bohlke; Congresswoman Diane E.
Watson and her assistant, Mr. Jim B. Clarke; Mr. Angelo J. Bellomo,
Director of the Office of Environmental Health and Safety, Los Angeles Unified
School District; and Dr. Thomas G. Raslear of the
Office of Research and Development, Federal Railroad Administration.
We would like to express our
gratitude to our USC colleagues, including professor Kurt Palmer for his advise
concerning statistical analysis; professors Genevieve Giuliano,
Lastly, we would like to
thank our graduate assistants, Ms. Sahar Shayesteh-Mehr and Mr. Kidra Shadroo, and our undergraduate
student assistants, Ms. Laurel Chen, Mr. William Murphy, Mr. Joseph Deato, and Mr. Faysal Al-Hashem for their help in this research. They were instrumental in conducting the
field data collection and videotapes, experimental setup and data collection in
our laboratory, and reducing the data for further analysis.
“Every grade crossing is an accident waiting to happen.”
David Solow, Executive Director of the
Railroad crossing accidents pose a serious safety problem
in the
There
is on average a collision between a train and a vehicle every 90 minutes in the
The number of
fatalities on street-level light rail crossings in the Southland has reached an
alarmingly epidemic proportion.
Sixty-five people have died on the Metropolitan Transportation
Authority’s (MTA) 60 miles of light railways in
The 22-mile
L.A.-Long Beach Blue Line has had more than 620 accidents (95% of the total MTA
accidents) and had “the highest light rail accident rate” in the state during
the 1990s. According to the California
Public Utilities Commission (CPUC) and the Federal Transit Administration it is
among the deadliest railways in the nation.
These are particularly troublesome because currently the Blue line is
the only significant above ground rail line in the MTA system with heavy street
intersections. (The Green Line is the
other major above ground light rail line, but it passes down the center of the
105 Freeway and does not cross over arterial streets.) With all of this tragedy, what will happen as
The same concerns
loom over the Exposition Light Rail that the MTA is planning to build on the
vacant Right of Way median of the
Railroad
safety engineers strive to provide grade crossing systems in which lights,
bells, and gates are activated with sufficient lead-time to accommodate the
fastest train, the slowest driver, and the worst environmental condition. Twenty years ago, only about 50,000 of the
country’s 225,000 public grade crossings were protected by flashing lights,
bells and/or gates that drop down when a train is about to pass. Today, there are about 62,000 of these active
crossings out of 154,000. Such active
warning systems are expensive -- even simple ones can cost $150,000 to
install. And active crossings have
human-factors problems of their own, which may help explain why they account
for half of all grade-crossing collisions (Los
Angeles Times,
However, these designs fail to consider the inherent limitations of human
judgment and decision-making in a potentially hazardous crossing event. Over 50% of crashes at public crossings occur
where active warning devices such as gates, lights, and bells exist and function
properly. More surprisingly, about 70%
of these collisions occur when the train is traveling less than 40 mph
(Highway-Rail Crossing and Trespasser Facts,
An in-depth look
at the literature in this area has revealed a number of influential factors in
HRC cause-effect relationships. A HRC
accident report of the Los Angeles County Blue Line has detailed sixteen
“contributing factors” for their accidents (LAC/MTA, 1999). Among these factors, the ones least studied
appear to pertain to the “human factors” of HRC accidents. The significance of these causal factors was
shown by Wilner (1998) who reported that over
one-third of rail accidents and 80% of train collisions are caused by human
error. Moreover, a study in Japan looked
at HRC accidents in a comprehensive and multi-factorial approach. (Anandarao and Martland,
1998). Based on their
classification, it was found that the leading cause of the crossing accidents
was the driver’s “Ignorance of Warning.”
It is plausible that they did not notice the warnings in time to
stop. Nevertheless, in many cases, a
major factor could also have been the driver’s choice to ignore the warning and
voluntarily enter the crossing.
Therefore, it is
critical to analyze driver behavior and decision-making as well as the
interaction of these variables with the engineering design of the crossing
(Rahimi and Meshkati, 2001). By adding these to a generic model of HRC
accidents (suggested by Tustin et al., 1986), we see a large number of
influential factors classified as: vehicle (size, maintenance), highway
(surface, geometry, traffic load, environmental conditions), driver (risk
perception, reaction time), train (speed, flagging, brake time), and crossing
feature (warning systems, visibility, trespassing, enforcement).
The current safety technologies seem to have been saturated and are ineffective in significantly reducing HRC accidents. In the mid-1980s, we began to see strong connections between the theories of cognition and human information processing and transportation-related accidents. Many of these studies hinted at the complexity of interactions between driver cognition and visual perception versus vehicle design and roadway/crossing conditions. For example, a driver may overestimate the safe time interval for crossing the tracks because of misleading visual cues (Rahimi, 1989; Rahimi et.al, 1990; Leibowitz, 1985). Studies have shown that for vehicles of equal velocities, the larger the object, the slower the perceived velocity (a notion called “expansion of optical array”). This leads drivers to underestimate the speed of an approaching train. Also, monocular cues such as rows of trees or telephone poles lining a track create the illusion that the train is farther away than its actual position. Overall, the illusion of velocity and size (inverse relationship), the illusion of perspective, and the deceptive geometry of collisions (inverse expansion pattern) all cause drivers to overestimate the safe time interval. Studies have also shown that time-to-collision estimates are more accurate with a normal vision field, binocular vision, higher speeds, and driving experience (Viola, 1988).
Also, a number of
NTSB accident reports hint at the possibility that drivers are given
insufficient indicators for an approaching train. For instance, according to the NTSB (December
2, 2003), the probable cause of the collision earlier this year between a
commuter Metrolink train and a truck in Burbank, CA,
which also caused the train derailed, was “the design of the traffic signals’
railroad hold interval, which displayed a flashing red arrow for the eastbound
North San Fernando Boulevard left turn lane onto North Buena Vista
Street.” Moreover, according to the NTSB’s investigation, “the accident driver lost situational
awareness in an ambiguous and confusing environment that required significant
mental alertness and vigilance; consequently, he missed the cues alerting
drivers to an approaching train.”
Warning cues,
roadway or track conditions, and sun glare affecting visibility can contribute
to these accidents. Warning signals
could also be improved to provide additional visual cues to drivers, like the
two new crossbuck railroad sign designs utilizing microprismatic sheeting materials, which have already been
evaluated in Ohio (Zwahlen and Schnell, 1999). These signs have been successful in
increasing the light reflection and visibility for drivers. However, even the most successful device is
faced with a resistance for implementation on a large scale (Barry, 1999).
Recently,
efforts have been made to utilize advanced technologies and intelligent
systems. Computers, digital data
communications, and other advanced technologies can help manage and control the
railroad, locomotives, maintenance, inspection efforts, and elements of the
railroad infrastructure. For example, electronic
train monitoring and control systems could display operating instructions
governing safe train movement from train dispatchers to train crews in the
locomotives. In addition, automated
enforcement at lowered crossing gates could use sensors to detect violators and
photograph the violator’s license plate (see LAC/MTA, 1997, for a detailed
description of photo enforcement at the Long Beach Blue Line Grade
Crossing). There have also been efforts
to further develop In-Vehicle Alert System (Wanat, 1/21/00), Vehicle to Roadside Communication (Kady, Mark, Shloss, and Peter,
1995), Intelligent Vehicle Initiatives for railroads (IVI), and the Vehicle
Proximity Alert System (VPAS). These systems are being designed and evaluated
to detect the presence of on-coming trains through sensors and broadcast that
information inside specially-equipped vehicles using visual displays and voice
capacities (1997 Projects Book, 2/1/00).
In
this project, we propose that the decision styles of the drivers have a
significant impact on the way in which the HRC actions are motivated. Decision styles are the way by which
individuals receive, store, process, and transmit information for action. By decision style we mean: (1) the manner in
which the driver reacts to a given crossing situation, or (2) the manner of
interaction with other elements of the environment. This approach suggests that environmental
variables (e.g., time pressures and mental load) affect the complexity of
information processing behavior of the driver.
In this respect, the decision-making differs among drivers in two key
dimensions: amount of information used and number of decisions made. The Driver Decision Style is evaluated by a
“paper and pencil” instrument, which has been used by 400,000 individuals.
We have accomplished the above goal(s) by designing experiments based on actual crossings from down town Los Angeles (please refer to Appendix A for copies of experimental procedures, trails, tasks, and Appendix E for an electronic version of video clips). We analyzed the effect of drivers’ individual information processing behavior and decision styles on handling different grade crossings plus their performance on recalling the crossing conditions while they made the decision to cross.
Based on this research, it is concluded that: Drivers’ individual decision styles and the combination of their styles play an important role in making decision concerning to cross or not to cross under different environmental conditions and time pressure situations.
Cognitive
styles are defined as learned thinking habits, which act as indices to the
individual's total personality system, its functioning, and development. Thus, they are collectively representative of
the individual’s conceptual structure (i.e., the way he/she receives, stores,
processes, and transmits information).
By analyzing these cognitive styles, behavioral variations and
individual differences of decision makers can be explicitly identified and
analyzed (Driver, 1983). By employing
the above concept, Schroder, Driver, and Streufert (1967), and Driver and Streufert
(1969) developed a human information processing model.
This
model suggests that environmental pressures (or load) systematically affect the
complexity of information processing in persons in an inverted-U-shaped
function. Each individual can be
considered to have a unique and consistent “curvilinear information-use
pattern,” referred to as their decision style. Every individual has acquired at least one
basic or “dominant” decision style that is normally exhibited under moderate
environmental load. For most people, a
second or "backup" style emerges in extreme environmental load
conditions, such as uncertainty and time pressure. Environmental load is defined as the sum of
the effects of four basic factors: (a) information complexity (e.g.,
information load, time pressure); (b) “Noxity” or
negative input (e.g., threat); (c) “Eucity” or
positive input (e.g., support from others); and (d) uncertainty (Driver, 1979).
The decision style model is based on two primary dimensions: information use and focus. Information use refers to the amount and complexity of information actually used in thinking and decision-making. Focus is defined as the number of alternatives, which are contained in the final solution, reached. Focus is a continuous dimension ranging from unifocus, in which a single alternative forms the outcome, to multifocus, in which many different options are included in the final solution. The unifocus style takes a given amount of data and connects it around a single solution or decision alternative, whereas the multifocus style takes the same amount of data and integrates it into several outcomes simultaneously or within a very short time.
The information use dimension can be split at some point between two extremes; at one extreme are those individuals who habitually use as much non-redundant information as is available, termed “maximizers.” At the other extreme are those individuals who use just enough information to generate one or two useful alternatives, termed “satisficers.” The maximizer/satisficer dimension suggests a high vs. low degree of integration, or the type and amount of connections between information units during analysis. By combining the dimensions of focus and information use, five distinct decision styles can be recognized: “Decisive” (unifocus, satisficer), “Hierarchic” (unifocus, maximizer), “Flexible” (multifocus, satisficer), “Integrative” (multifocus, maximizer), and “Systemic” (combination of Integrative and Hierarchic) (Driver, Brousseau, and Hunsaker, 1993).
The
decision style model has several implications for decision-making in a
naturalistic milieu and in time-pressured tasks. For instance, in mental task performance,
different styles consistently demonstrate distinctively different
reactions (e.g., perceived difficulty) to the same task load level and
environmental demands, as reported by Meshkati and Driver (1984), and Meshkati
and Loewenthal (1988).
Just
as there are marked differences in the decision style of individuals or teams,
many differences can be seen in the degree of complexity that individuals are
motivated to utilize in their decision making (Schroder,
Driver & Streufert, 1967). In general, complexity motivation is the
degree to which individuals are driven to use a variety of information, skills,
and methods in the process of completing a task. Information refers to either sheer volume or
the rate of information per time unit.
The
other component is the differentiation or the number of types of information
that are used in an information network.
At one extreme are individuals who prefer to keep the process simple by
working with a limited amount of information and number of methods. At the other extreme, some individuals prefer
a high degree of complexity, actively seeking a myriad of kinds and sources of
information as well as a variety of techniques to complete their jobs.
A
second complexity motive is the need for order as compared to uncertainty,
defined as the uncertainty motive. Some
individuals require a high degree of structure, order, and feedback while
others thrive on novelty, ambiguity, risk or conflict.
Individuals
seek to balance their desired level of complexity/uncertainty and the actual
level of complexity/uncertainty that their work requires. In areas where the desired level of
complexity is lower than the actual complexity required, individuals are
considered to be experiencing “strain.”
In areas where the desired level is higher than the actual complexity
required individuals are said to operate under a “growth” situation. In areas where the two match, individuals are
said to be in balance. Similarly,
individuals can experience levels of strain, growth, or balance with regard to
their uncertainty motive.
The
primary objective of this research is to explore the relationship between the
decision styles of drivers versus their highway-rail crossing behavior. In order to accomplish this objective, a
number of statistical analyses were performed, as described below.
Analysis
of variance (ANOVA) is a statistical tool that defines the impact of several
independent variables on a dependent variable, in an experimental design
format. For this analysis, we defined a
number of variables associated with the design of HRCs
and sought to measure their impacts on driver decision styles.
After
studying the literature and interviewing a number of experts from MTA, we have
defined two variables as the most influential in this respect. These variables
were labeled as “time pressure” and “environmental complexity.” Other variables studied, but not considered
for this experiment, were familiarity of the drivers with the intersection,
speed of vehicle, and acceleration and deceleration of the train. The decision
style variables have been discussed in the previous sections.
In
our study, time pressure refers to situations in which drivers at an HRC are
under strict time constraints and are thereby forced to form their decisions
quickly and efficiently. Environmental
complexity encompasses a variety of factors, including intersection complexity,
traffic congestion, and distractions in the environment viewed by the driver. As explained below, time to completion was
considered as a measure of time pressure and signal to noise ratio was used as
a measure of environmental complexity.
We designed
our ANOVA to have four levels for the decision style (DS) variable and two
levels for environmental complexity (EC) variables. Within two levels of EC, we have nested two
levels of “low” versus “high” to differentiate the degree by which these EC
variables are measured. This gives us a
2x2x4 mixed (nested) ANOVA table below.
Table 1. ANOVA Table for the Experimental Trials
|
Decision Style |
Time Pressure |
Environmental Complexity |
||
|
Low |
High |
Low |
High |
|
|
Decisive |
|
|
|
|
|
Flexible |
|
|
|
|
|
Integrative |
|
|
|
|
|
Hierarchical |
|
|
|
|
|
Systemic |
|
|
|
|
The
mixed design indicates that we have within-subjects variability for EC levels
and between-subject variability for Low versus High levels. The subjects were shown approximately 20
seconds of video footage of trains passing by on the left-hand side of the
screen, to accustom them to the notion that trains are present in these
intersections. Then, they were shown
video footage of the different scenarios and their responses were monitored.
Two
scenarios were created for time pressure, low and high. In the low time pressure footage the subjects
watched a traffic light in the intersection that stayed green while a vehicle
in front of the driver pulled into the left lane, in preparation for a
left-turn on the signal. The front vehicle turns left and the experimental
vehicle follows, making a simple left-turn across the HRC. We note that “left-turn” across HRCs are the most accident prone
maneuvers by automobile drivers. In this
setting, the subject is under very little time pressure when making his/her
decision to turn left. In the high time
pressure footage the subjects viewed a vehicle in front of the experimental
vehicle at time 0 seconds while the left turn arrow light was green. At time 1 second, the light turned yellow,
while the vehicle in front began moving into the intersection for a left-turn. At time 2 seconds, the front vehicle
completed a left-turn while the light stayed yellow. Meanwhile, the experimental vehicle moves up
to the stop line. At time 3 seconds, the
left turn arrow turned red and the vehicle in front had completed its
left-turn. Some drivers in this
situation, have reported difficulty making a decision whether to cross or not –
potentially a source of HRC accident in some left-turn situations. For these two scenarios, we measured the time
it takes for the subject to make the decision to cross.
After
the video footage for the time pressure scenarios were shown to the subjects
and their measures were taken, the subjects were asked to draw the intersection
they just observed in the video on a piece of paper. They were told to draw all objects they remembered
from the scenes. We were specifically
interested to see whether they remembered the sign panel which contains the
flashing train picture, the left-turn arrows, the left turn sign, and traffic
lights for through traffic, and bar indicators showing the direction of traffic
movement. They were then given a
questionnaire including questions such as:
·
What signs did you pay attention to?
·
Do you know the meaning of each sign and its purpose?
·
Which signs give indications that are important to you?
From
literature we knew that drawing the scene causes the subject to use their
recall memory, while identifying the purposes of the signs calls upon
recognition memory. The recall scene is a simple drawing of the intersection
with all objects in the scene in place.
The recognition scene is a drawing of the scene with some of objects in
place. The subjects were asked to add items of importance in this scene. The questionnaires are included in Appendix
A.
As
explained above, the tests have been designed for four different scenarios
based on four experimental situations. Every situation was considered to
measure the impact of time pressure (TP) and environmental complexity (EC).
Each time and environmental complexity variable was divided to two statuses as
low and high to have a spectrum with high and low limits and were defined as
HTP (High Time Pressure), LTP (Low Time Pressure) and HEC and LEC for high and
low environmental complexity.
The
following table shows high and low criteria for environmental complexity. For
example a HEC intersection contained all marked objects under that column
(e.g., light pedestrians, trees, sidewalks, etc.).
Table 2. Objects in Intersections (Low vs. High Env. Complexity)
|
Objects |
LEC |
HEC |
|
Pedestrians |
|
P |
|
Vehicles |
|
P |
|
General signs |
P |
P |
|
Traffic direction |
P |
P |
|
Vendors |
|
P |
|
Buildings |
P |
P |
|
Number of lanes of through traffic |
P |
P |
|
Number of lanes of perpendicular traffic |
P |
P |
|
Lighting poles |
P |
P |
|
Trees |
P |
P |
|
Signal poles |
P |
P |
|
Cross walk |
P |
P |
|
Sidewalk |
P |
P |
|
Train-related signs |
P |
P |
|
Train station |
|
P |
|
Passing train |
|
P |
|
Train tracks |
P |
P |
In order
to choose which intersections represented these conditions the best, we
videotaped a large number of HRC intersections in downtown Los Angeles. More than twenty hours of videotapes were
analyzed for these criteria. In addition
to the above criteria, other parameters such as number of accidents in those
intersections were taken into account.
We used MTA’s Metro line data to match
accident statistics with the assigned intersections. For instance, the intersections we taped were
among the top three ranked intersections in terms of their incident rates from
July 1990 to March 2000 (Summary of Metro
Blue Line Train/Vehicle and Train/Pedestrian Accident May 19, 2000 Prepared
by Risk Management Dept.). The final
selection for LEC was the intersection of Los Angeles Street and Washington,
and the HEC was the intersection of Grand and Washington. The direction of light train moving was
East/West along the Washington Street.
We then simulated our scenarios based on these intersections and
videotaped them for our experimental setup.
We
videotaped the actual live scenes from inside an automobile, while crossing a
high-incident intersection in downtown Los Angeles. A camera was mounted at the driver’s eye
level and taped the subject’s actual visual scene while driving a vehicle. For the experimental setup, each session
playback represented a cell within our nested ANOVA design. Post-experiment questionnaires are shown in
the Appendix A for the four conditions of LEC, HEC, LTP, and HTP.
We used
Minitab statistical software to compute the F-ratios for each ANOVA. The following tables contain the results of
the analysis in the form of Minitab output tables.
Table 3. The effect of Decision Style, LEC, and Recall
Table 4. The effect of Decision Style, LEC, and Recognition
Table 5. The effect of Decision Style, HEC, and Recall
Table 6. The effect of Decision Style, HEC, and Recognition
Based
on the calculated p-values, only the first effect of decision style is
statistically significant, at an a = 0.05. This indicates that there are differences in
decision styles of our subjects considering the low environmental complexity
and when it involves recalling information about the HRC intersection. One way of interpreting this result is that
those who are, for example, decisive decision-makers, seem to recall the HRC
environmental information differently.
This also means that this difference becomes larger when the decisive
decision-makers are faced with fewer signals and distracters while making a
left-turn in a non-busy intersection.
None
of the other conditions in our ANOVA resulted in significant differences among
the DS categories. We believe that this
was primarily due to the parametric nature of the test and also the low power of
this test. In other words, with ten subjects in each cell the power of the
statistical test was not sufficient to detect any significance. A much larger number of subjects (around 300)
would have given us a better indication of these differences, when using a
parametric regression test such as ANOVA.
However, conducting an experiment with this many subjects would have
required substantially larger resources, not available to the researchers here.
So, we began developing a non-parametric test for our data.
After
the results of our ANOVA, we reorganized our data to conduct a non-parametric
test. One such test is a chi-square
test. This time, due to the low number
of subjects, we grouped the decision style into two levels. Subjects were grouped into two (18-member)
groups based on their decision style: single alternative (SI) or multiple alternative (MU). SI is a subject whose decision style test
is Decisive, Hierarchical or Systemic and MU is a subject whose decision style is
either Flexible or Integrative. Decision style research indicates that such
dichotomy makes sense when the subjects are dealing with gathering and
processing a large amount of information from objects in an environment.
The
subjects decision style scores are given in Appendix
B. The following table shows how the
subjects are grouped according to their decision style versus two levels of TP
and two levels of EC.
Table 7. Chi-Square Test Table (Time Pressure Vs. Env. Complexity)
|
|
Decision style as
Grouping Variable ß |
|
|
|
Nested System
Variables Þ |
|
TP (H,L) |
EC (H,L) |
|
|
Decisive + Flexible |
|
|
|
|
Integrative + Hierarchical + Systemic |
|
|
Each
system variable (independent variable) is nested with two levels of “high” and
“low.” The variables are Time Pressure
(TP) and Environmental Complexity (EC).
The subjects are licensed drivers assigned to cells i
(between-subject design) and j (within-subject design). The subjects will be randomly assigned to
each cell to reduce learning effects.
For each subject, we measured the aggregate number of signals and noise
items that they identified in their respective scenes, as explained below.
A
signal is an environmental object or item, which draws the attention to the
possibility of a crossing train presence.
On the other hand, a noise item is an environmental object, which takes
away the attention of the subject from the possibility of a crossing train
presence. The following table presents object’s type (noise/signal) in an
intersection with environmental complexity.
We have added a weighting factor to scale the importance of each item
within the scene. Appendix C begins with
an object identification key for each identifiable item in the intersection,
then it shows the raw number of objects identified for each driver. The number of objects identified for each
driver and its cross-tabulation for LEC and HEC, object recognition versus
object recall, signal versus noise are shown in Appendix D. Appendix E contains the verbal comments by
each subject during the experimental trials.
In this appendix, the video clips used for experimental trials are also
included.
Table 8. Environmental Complexity Objects and Their Assigned Weights
|
Objects |
Noise/Signal |
Weight |
|
Pedestrians |
N |
1 |
|
Vehicles |
N |
1 |
|
General signs |
N |
1 |
|
Traffic direction |
N |
1 |
|
Vendors |
N |
1 |
|
Buildings |
N |
1 |
|
Number of lanes of through traffic |
N |
0.5/lane |
|
Number of lanes of perpendicular traffic |
N |
0.5/lane |
|
Signal poles |
N |
0.5 |
|
Trees |
N |
0.5/tree |
|
Cross walk |
N |
1 |
|
Side walk |
N |
1 |
|
Train-related signs |
S |
2 |
|
Train station |
S |
2 |
|
Passing train |
S |
2 |
|
Train tracks |
S |
2 |
We,
therefore, designed a 4x4 chi-square table in order to study the effects of
each decision style on the HRC safety factors. In the chi-square table above, independent
variables are grouped into decision styles and system variables.
The
following tables show the chi-square test result for HEC and LEC environments.
The data for the chi-squared test is from the raw count of signal and noise in the
HR crossings (Appendix C). For each
complexity level, the tables show both the actual signal and noise items versus
the expected ones for LEC and HEC situations.
Table 9. The Results of Chi-Square Tests for LEC
|
Actual |
Recall |
Recognition |
||||
|
|
Signal |
Noise |
Total |
Signal |
Noise |
Total |
|
D+H+S |
27 |
90 |
117 |
17 |
51 |
68 |
|
F+I |
38 |
117 |
155 |
30 |
72 |
102 |
|
Total |
65 |
207 |
272 |
47 |
123 |
170 |
|
Expected |
Recall |
Recognition |
||
|
|
Signal |
Noise |
Signal |
Noise |
|
D+H+S |
27.96 |
89.04 |
18.8 |
49.2 |
|
F+I |
37.04 |
117.96 |
28.2 |
73.8 |
|
Chi square |
0.782785022 |
0.528648 |
Table 10. The Results of Chi-Square Tests for HEC
|
Actual |
Recall |
Recognition |
||||
|
|
Signal |
Noise |
Total |
Signal |
Noise |
Total |
|
D+H+S |
40 |
97 |
137 |
19 |
62 |
81 |
|
F+I |
42 |
111 |
153 |
25 |
63 |
88 |
|
Total |
82 |
208 |
290 |
44 |
125 |
169 |
|
Expected |
Recall |
Recognition |
||
|
|
Signal |
Noise |
Signal |
Noise |
|
D+H+S |
38.74 |
98.26 |
21.09 |
59.91 |
|
F+I |
43.26 |
109.74 |
22.91 |
65.09 |
|
Chi square |
0.742083 |
0.463346 |
The
Chi-square results show no significance, based on a type I error of 0.05. In other words, there was no significant difference
between the two groups of decision styles for each level of recall and
recognition. The same applied to the
signal items (i.e., train-related) versus noise items (i.e., environmental
distracters). This lack of significance
was seen in both low and high environmental complexity situations.
In
our previous research (e.g., Meshkati, et. al. 1999), we indicated that
different decision styles have a different but predictable tolerance for, and response
to, different task dimensions such as complexity, uncertainty and information
load. It is expected that those decision
styles which are categorized as multifocus (i.e.,
Integrative and Flexible) will have more options and perform better when responding
to events containing high levels of uncertainty. In addition, those with information
maximizing styles (i.e., Integrative, Hierarchic and Systemic) experience less
strain (respond more smoothly) to high levels of complexity than those with
information satisfying styles (i.e., Decisive and Flexible). This is particularly critical during events
requiring unfamiliar reasoning and diagnosis where uncertainty and complexity
are typically high. To illustrate this
point, consider an individual who has a unifocus,
information satisfying style (Decisive) that is required to contend with an
unfamiliar grade crossing. This
individual will experience a high degree of stress when attempting to analyze
the situation. On the other hand, this
kind of unfamiliar scenario is more comfortable for multifocus,
information maximizers (i.e., Integrative), provided
the time pressure was not excessive.
That is, a routine situation requiring a quick decision would be best
handled by Decisives (unifocus,
information satisficers) as compared to those who are
multifocus, information maximizers
(i.e. Integrative).
Also,
based on previous and this research, it is concluded that the drivers’
individual decision styles and the combination of their styles play an
important role in making decision concerning to cross or not to cross under
different environmental conditions and time pressure situations. (For a detailed account of the subject’s
verbal responses during the experimental trials, see Appendix D). Finally, drivers’ decision styles affect both
their recall and recognition performance of grade crossing signage warming
systems. The important implication of
this research is that the grade crossing and warning light and signage design must
take into consideration the different informational needs of drivers performing
different types of driving tasks.
In the scientifically oriented literature on HRC accidents, there are scant references to the root causes of these accidents, such as a disregard for stop signs, failure to look for trains, distraction, judgment error, inattention, fatigue, or drug impairments. In popular press, on the other hand, we often see these causes reduced to “human error.” Should we be satisfied with this reductionism? Some argue that this is sufficient due to the success record of bringing this argument in the legal challenges against MTA. We argue that the precautions required by the legal and regulatory statutes are not sufficient to ensure safety in the Los Angeles HRC context. While this argument may be legally defensible, it is scientifically and morally questionable. The proportionally high rate of both fatal and nonfatal accidents should be a sufficient indication that the current procedures for designing and operating warning systems are not adequate.
The critical failure of current rail safety technical analysis is that it does not investigate how people with different information processing behavior and risk perception actually respond to light rail warning systems and protective safety features. An “average” person under normal conditions with maximum attentional resources may have no problem recognizing what warning signs are telling them. But someone who is under time pressure, distracted, has slower reflexes, or is unfamiliar with the location, will be less likely to notice and realize that s/he is in danger at the critical moments. The responsibility in these situations should not be placed entirely on the driver and the pedestrian because it is unrealistic to expect that individuals will always be in ideal cognitive circumstances when they come upon the warning signs. Warning features should be designed for the most at risk individuals, not just for the “average” individual.
The drivers and pedestrians in Southern California are especially vulnerable to this problem because light rail is new here. The population in general is not familiar with trains running across their streets and backyards, and they have not internalized the inherent risks associated with HRCs in their cognitive and behavioral characteristics. For example, whenever we are in the “left-turn only” lane of an intersection, we are used to be watchful of the oncoming traffic. Our risk perception domain is not sensitized to a situation that requires us to look into the side mirror searching for a potential vehicle passing us on the left side. Therefore, a driver at a rail crossing, attempting to turn left would have a higher chance of collision with a train approaching from behind, because the driver does not expect a vehicle coming from behind. Not surprising, statistics indicate a significantly higher collision rates for MTA trains by drivers attempting to make a left turn in violation of standard traffic signals correctly displayed at HRCs. In these cases, the drivers may have thought that they are risking a traffic ticket, but they are probably risking their lives without the intersection and its environment conveying the risk to the driver.
Despite their genuine interest in protecting public safety and dedicated safety staff, neither the MTA nor Metrolink have been successful at addressing the complex issue of grade crossing safety. Therefore, we recommend that the MTA and Metrolink begin a systematic investigation on how people react to warning signs, signals, symbols, and pedestrian protective systems, and how these designs are perceived and responded to by diverse groups of people under a variety of conditions and situations. The MTA’s $400,000 portable multimedia theater for public education and Metrolink’s 177 public presentations in 2002 may constitute necessary steps in the right direction, but they are far from being sufficient at preventing HRC accidents at this time. Presently neither the Federal Transit Administration nor the California Public Utilities Commission (CPUC), as the designated state safety oversight agency, have a readily available standard or a sensible guideline for taking into account the local population’s characteristics, limitations and capabilities in the design of warning signs and protective systems at grade crossings. The CPUC’s voluminous safety audit of the MTA in 2002 concentrated only on how the operation and maintenance of a few randomly chosen “crossing warning” systems and devices conformed with “(LAC)MTA signal maintenance standards.” The safety audit addressed neither the adequacy of design features of such systems and devices nor their critical human factors considerations that affect intended users behavior.
This is not the time for recriminations. We do need to spend resources on rail transportation infrastructure in the region; yet, its ultimate success depends on improved rail safety performance. To do this we need a paradigm shift in how we implement our light rail infrastructure. To paraphrase the American philosopher William James who once said, “Great emergencies and crises show us how much greater our vital resources are than we had supposed”. In order to stop and reverse the dangerous direction we are headed in the MTA, Metrolink, the California Transportation Commission, the CPUC, the Federal Transit Administration, the Federal Railroads Administration, and the research community must rise to the challenge. We need to work together to systematically analyze all the pervious accidents, to study all the existing crossings, and to come up with redesign of crossings and warning signs and signals based on sound technical and human factors considerations.
We believe that we have a serious and urgent safety problem at our nation’s grade crossings, and further research is needed to study and address this problem.
Anandarao, S. and C. Martland, (1998). Level-Crossing Safety on East
Japan Railway Company: Application of
Probabilistic Risk Assessment Techniques.
Transportation, 25(3):
265-286.
Barry, P., (1999). U.S. Transportation
Secretary Slater Conducts Roundtable Meeting on Highway-Rail Crossing Safety.
http://www.itsa.org/85256201003EFA03/0/ D4878AF071 A1778C85256752006B35E6?Open
Carlson, P. and K.
Fitzpatrick, (1999). Violations at Gated
Highway-Railroad Grade Crossings.
Journal of the Transportation
Research Board: Transportation Research Record, 1692: 66-73, Paper No.
00-0259.
Department of Transportation, Highway-Rail
Crossing and Trespasser Facts: Always
Expect a Train, 1/11/00. www.fra.dot.gov/o/public/factoid.htm.
Driver, M.J. (1979). Individual decision making and creativity. In
Driver, M.J. (1983).
Decision styles and organizational behavior: Implications for
academia. The Review of Higher Education, 6(4), 387-406.
Driver, M.J., Brousseau, K. R., & Hunsaker,
P.L. (1993). The
Dynamic Decision Maker: Five Decision Styles for Executive and Business
Success.
Driver, M.J., and Streufert,
S. (1969). Integrative complexity: An approach to individuals and groups as
information processing systems. Administrative Science Quarterly, 14, 272-285.
Kady, M. and P. Shloss, (1995). 4th Annual Meeting: Electronic
Messaging Using VRC.
http://www.itsa.org/8525626E0028B517/0/9EB4E254E746707E8525626E0029F0F6?Open&Highlight=2,Kady.
Leibowitz, H. W., (1985). Grade Crossing Accidents
and Human Factors Engineering. American Scientist, 73(6): 558-562.
Meeker, F., (1997). A
Comparison of Driver Behavior at Railroad Grade Crossings With
Two Different Protection Systems. Accident Analysis and Prevention,
29(1): 11-16. 1997 Projects Book:
Railroad Crossing Vehicle Proximity Alert System, 2/1/00.
http://www.itsa.org/852564980079857D/0/56A8582E677DCA718525649800677A7D?Open.
Meshkati, N., Buller, B.J.,
Tan, W.M, and Lesjak, B. (1999, June). Research Validating a Model of the Effect of
Operator Information Processing Behavior on Performance while Using a
Traditional and Ecological User Interface for a Nuclear Power Plant. Prepared for the U.S. Nuclear Regulatory
Commission (Contract No. NRC 04-91-102). Los Angeles, California: University of
Southern California.
Meshkati and Driver (1984) Individual information processing behavior in perceived job
difficulties: A decision style and job design approach to coping with human mental
workload. In H.W. Hendrick & O. Brown, Jr. (Eds.), Human Factors in
Organizational Design and Management. Amsterdam: North-Holland, 533-538.
Meshkati and Loewenthal (1988). The effects of individual
differences in information processing behavior on experiencing mental workload
and perceived task difficulty: An experimental approach. In P.A. Hancock and N. Meshkati (Eds.).
Human Mental Workload. Amsterdam: North-Holland,
269-288
National Transportation
Safety Board (NTSB) Safety at Passive Grade Crossings: Volume 2 – Case
Summaries, 1/22/00.
Rahimi, M. (1989). A task, behavior, and
environmental analysis for automobile left-turn maneuvers. Proceedings, Human Factors Society 33rd
Annual Meeting, Santa Monica, CA: Human Factors Society.
Rahimi, M., Briggs, R.P.,
and Thom, D.R. (1990). A field evaluation of driver eye and head movement strategies
toward environmental targets and distractors. Applied Ergonomics,
21(4), 267-275.
Rahimi, M. and Meshkati, M. (2001). Human factors in highway-rail crossing
accidents: the influence of driver decision style. Driving Assessment 2001,
Schroder, Driver & Streufert, (1967). Human information processing. New
York: Holt, Rinehart and Winston.
Grade Crossing Handbook (2nd
Edition), Report No. FHWA
TS-86-215, Federal
Highway Administration, Washington, D.C.: 273.
Voiola, C., (1988).
Visual Information and Skill Level in
Time-to-Collision Estimation. Perception, 17(5): 623-632.
Wanat, J., 1/21/00. Highway-Rail Intersection.
www.path.berkeley.edu/~leap/TTM/
High_Rail_Intersect/index.html
Wilner, F.
(1998). Asleep
at the Throttle? Journal of Commerce
Inc., May: 14.
Zwahlen, H. and T. Schnell, (1999). Efficacy of Safety-Related
Investments to
Reduce Fatalities on the East Japan Railway. Journal
of the Transportation
Research Board: Transportation Research
Record, 1692: 82-93.
Instructions for Participants
The investigators would like to thank you for participating in this research.
In this session, you will be watching four short video clips of traffic in downtown Los Angeles. After each clip, you will be asked to fill out a short questionnaire.
When you receive a signal from the investigator, the video clip begins running. While watching this video, please try to imagine yourself as a driver sitting in a driver seat of a car. Make sure that you watch the scene very carefully. We want you to concentrate on the video monitor while the clip is running in its entirety.
The investigator will hand you the questionnaire immediately after each clip. Your task is complete after responding to the questionnaire for the fourth video clip.
Please do not share any information
about this experiment and what you saw in the video clips with any other
person. Others may be participants in
this experiment, and we want them to be unbiased and “fresh” concerning any
aspect of this study.
Thank you!
Please write last 4 digit of your SS#___________________
Post Video Survey HEC
Do you want to cross the intersection or make a left turn?
You are asked to answer the following questions:
Visualize the scene at the intersection. Draw and label everything that you
remember in this clip. A simple sketch
is provided below to help you identify the location of items you want to
identify.
There is no time pressure, but please finish
this drawing in less than five minutes.
Visualize the scene at the intersection.
A simple sketch is provided below to help you identify the location of
items. Draw and label everything that you remember in this clip. You may add, draw and label anything that
is missed from the sketch.
There is no time pressure, but please finish it
in about 5 minutes.
Which piece or pieces of information did you use when making your decision to cross the intersection or turn left or stay for the next light?
Was there any railroad track visible?
Was there any sign or signal to draw your attention to the presence of the railroad track?
Were there any signs or signals at the scene, other than railroad related?
If so, please name the sign(s) or signal(s):
What does each of the above signs or signals mean to you?
While waiting, what were you looking at? Please mention everything that you were looking at.
Which signs are meant for you when you are waiting to cross the intersection or make a left turn?
Were you familiar with this intersection prior to today’s viewing?
As you were making a left turn, was there anything you would like to see?
Please write down last 4 digit of your SS#___________________
Post Video Survey HTP
Would you have crossed the intersection or made the left turn here?
Empty intersection for memory recall
Visualize and draw
the signs and traffic lights you saw in the scene.
Visualize the scene at the intersection. A simple sketch is provided below to help you
identify the location of items. Draw and label everything that you
remember in this clip. You may add, draw
and label anything that is missed from the sketch.
There is no time pressure, but please finish it
in about 5 minutes.
Which piece or pieces of information did you use when making your decision to cross the intersection or turn left or stay for the next light?
Was there any railroad track visible?
Was there any sign or signal to draw your attention to the presence of the railroad track?
Were there any signs or signals at the scene, other than railroad related?
If so, please name the sign(s) or signal(s):
What does each of the above signs or signals mean to you?
While waiting, what were you looking at? Please mention everything that you were looking at.
Which signs are meant for you when you are waiting to cross the intersection or make a left turn?
Were you familiar with this intersection prior to today’s viewing?
Please write down last 4 digit of your SS#___________________
Post Video Survey LEC
Do you want to cross the intersection or make a left turn?
You are asked to answer the following questions:
Visualize the scene
at the intersection. Draw and label everything
that you remember in this clip. A simple
sketch is provided below to help you identify the location of items you want to
identify.
There is no time
pressure, but please finish this drawing in less than five minutes.
Visualize the scene at the intersection. A simple sketch is provided below to help you
identify the location of items. Draw and label everything that you
remember in this clip. You may add, draw
and label anything that is missed from the sketch.
There is no time pressure, but please finish it
in about 5 minutes.
Which piece or pieces of information did you use when making your decision to cross the intersection or turn left or stay for the next light?
Was there any railroad track visible?
Was there any sign or signal to draw your attention to the presence of the railroad track?
Were there any signs or signals at the scene, other than railroad related?
If so, please name the sign(s) or signal(s):
What does each of the above signs or signals mean to you?
While waiting, what were you looking at? Please mention everything that you were looking at.
Which signs are meant for you when you are waiting to cross the intersection or make a left turn?
Were you familiar with this intersection prior to today’s viewing?
As you were making a left turn, was there anything you would like to see?
Please write down last 4 digit of your SS#___________________
Post Video Survey LTP
Would you have crossed the intersection or made the left turn here?
Visualize and draw the
signs and traffic lights you saw in the scene.
Visualize the scene at the intersection. A simple sketch is provided below to help you
identify the location of items. Draw and label everything that you
remember in this clip. You may add, draw
and label anything that is missed from the sketch.
There is no time pressure, but please finish it
in about 5 minutes.
Which piece or pieces of information did you use when making your decision to cross the intersection or turn left or stay for the next light?
Was there any railroad track visible?
Was there any sign or signal to draw your attention to the presence of the railroad track?
Were there any signs or signals at the scene, other than railroad related?
If so, please name the sign(s) or signal(s):
What does each of the above signs or signals mean to you?
Which signs are meant for you when you are waiting to cross the intersection or make a left turn?
Were you familiar with this intersection prior to today’s viewing?
Decision Styles of the
Drivers Participating in the Experiment (N=43)
|
ID NUMBER |
H OP |
I OP |
F OP |
D OP |
S OP |
PRIM OP |
BACK OP |
H LEVEL |
I LEVEL
|
F LEVEL |
D LEVEL |
S LEVEL |
|
8,337 |
47.75 |
29.90 |
14.76 |
34.12 |
|
H |
I |
MH |
M |
L |
L |
|
|
6,044 |
42.95 |
27.30 |
44.50 |
16.58 |
|
F |
H |
MH |
ML |
MH |
VL |
|
|
3,599 |
45.10 |
44.10 |
39.50 |
13.60 |
44.60 |
S |
F |
MH |
MH |
MH |
VL |
MH |
|
9,837 |
37.30 |
16.80 |
44.50 |
21.97 |
|
F |
H |
M |
L |
MH |
L |
|
|
2,714 |
32.00 |
30.20 |
9.52 |
42.60 |
31.10 |
S |
D |
M |
M |
L |
ML |
M |
|
6,185 |
13.35 |
26.20 |
49.50 |
23.48 |
|
F |
I |
L |
ML |
H |
L |
|
|
2,899 |
18.85 |
21.85 |
9.52 |
49.77 |
|
D |
I |
L |
ML |
L |
MH |
|
|
2,835 |
20.30 |
20.00 |
9.52 |
49.90 |
20.15 |
D |
S |
ML |
L |
L |
MH |
ML |
|
4,827 |
12.90 |
24.60 |
44.50 |
27.50 |
|
F |
I |
L |
ML |
MH |
L |
|
|
5,523 |
16.20 |
18.65 |
20.00 |
45.05 |
|
D |
F |
L |
L |
L |
MH |
|
|
2,109 |
26.80 |
17.55 |
29.50 |
35.22 |
|
D |
F |
ML |
L |
M |
M |
|
|
3,465 |
29.75 |
48.60 |
64.50 |
0.55 |
|
F |
I |
M |
MH |
VH |
VL |
|
|
9,307 |
40.15 |
32.40 |
20.00 |
32.48 |
36.28 |
S |
D |
MH |
M |
L |
L |
M |
|
5,295 |
20.00 |
51.80 |
64.50 |
2.73 |
|
F |
I |
L |
H |
VH |
VL |
|
|
1,833 |
43.60 |
21.65 |
14.76 |
38.25 |
|
H |
I |
MH |
ML |
L |
L |
|
|
2,910 |
26.10 |
17.55 |
34.50 |
32.12 |
|
F |
H |
ML |
L |
M |
L |
|
|
2,285 |
21.60 |
34.00 |
49.50 |
18.13 |
|
F |
I |
ML |
M |
H |
VL |
|
|
2,129 |
20.05 |
22.45 |
34.50 |
32.50 |
21.25 |
F |
S |
ML |
ML |
M |
L |
ML |
|
4,887 |
29.50 |
39.90 |
14.76 |
36.87 |
|
I |
H |
M |
MH |
L |
L |
|
|
9,079 |
43.65 |
41.65 |
- |
38.23 |
42.65 |
S |
D |
MH |
MH |
VL |
L |
MH |
|
8,464 |
32.40 |
18.90 |
9.52 |
46.23 |
|
H |
D |
M |
L |
L |
ML |
|
|
9,392 |
21.25 |
17.80 |
34.50 |
33.65 |
|
F |
H |
ML |
L |
M |
L |
|
|
1,021 |
45.90 |
23.00 |
24.74 |
30.37 |
|
H |
F |
MH |
ML |
ML |
L |
|
|
2,602 |
19.80 |
46.00 |
20.00 |
34.73 |
|
I |
D |
L |
MH |
L |
L |
|
|
1,416 |
18.00 |
25.70 |
54.50 |
18.77 |
|
F |
I |
L |
ML |
H |
VL |
|
|
2,966 |
46.20 |
67.25 |
49.50 |
(1.15) |
|
I |
F |
MH |
VH |
H |
VL |
|
|
8,763 |
8.40 |
3.30 |
24.74 |
49.43 |
|
D |
F |
VL |
VL |
ML |
MH |
|
|
1,369 |
16.65 |
16.10 |
49.50 |
25.75 |
|
F |
D |
L |
L |
H |
L |
|
|
3,750 |
70.15 |
91.90 |
49.50 |
(17.35) |
|
I |
H |
VH |
VH |
H |
VL |
|
|
4,850 |
36.00 |
108.00 |
54.50 |
(14.67) |
|
I |
F |
M |
VH |
H |
VL |
|
|
4,924 |
34.45 |
14.80 |
14.76 |
43.58 |
|
H |
D |
M |
L |
L |
ML |
|
|
5,004 |
26.70 |
31.35 |
14.76 |
40.65 |
|
I |
H |
ML |
M |
L |
ML |
|
|
6,675 |
47.40 |
33.60 |
39.50 |
16.33 |
|
H |
F |
MH |
M |
MH |
VL |
|
|
9,253 |
25.40 |
46.35 |
59.50 |
6.08 |
|
F |
I |
ML |
MH |
VH |
VL |
|
|
2,575 |
34.00 |
59.40 |