219

NATIONAL COOPERATIVE
HIGHWAY RESEARCH PROGRAM REPORT
219,
APPLICATION OF TRAFFIC CONFLICT
ANALYSIS AT INTERSECTIONS

TRANSPORTATION RESEARCH BOARD

NATIONAL RESEARCH COUNCIL
TRANSPORTATION RESEARCH BOARD 1979

Officers
CHARLEY V. WOOTAN, Chairman W. N. CAREY, JR., Executive Director

Executive Committee
HENRII( E. STAFSETH, Asst. to the President, American Assn. of State Highway and Transportation Officials (ex officio)
LANGHORNE M. BOND, Federal Aviation Administrator, U.S. Department of Transportation (ex officio)
KARL S BOWERS Federal Highway Administrator, U.S.Department of Transportation (ex officio)
THEODORE C. LUTZ, Urban Mass Transportation Administratdr, U.S. Dept. of Transportation (ex officio)
JOHN M. - SULLIVAN,* Federal Railroad Administrator, U.S. Department of Transportation (ex officio)
ELLIOTT W. MONTROLL, Chairman, Commission on Sociotechnical Systems, National Research Council (ex officio)
A. SCHEFFER LANG, Consultant, Washington, D.C. (ex officio, Past Chairman 1978)
PETER G. KOLTNOW, President, Highway Users Federation for, Safety and Mobility (ex officio, Past Chairman 1979)
HOWARD L. GAUTHIER, Professor of Geography, Ohio State University (ex officio, MTRB liaison)
WILLIAM J. HARRIS, JR., Vice President (Res. and Test Dept.), Association of American Railroads (ex officio)
GEORGE J. BEAN, Director of Aviation, Hillsborough County Aviation Authority, Florida
RICHARD P. BRAUN, Commissioner, Minnesota Depdrtment of Transportation
LAWRENCE D. DAHMS, Executive Director, Metropolitan Transportation Commission, San Francisco Bay Area
ARTHUR C. FORD, Assistant Vice President (Long-Range Planning), Delta Air Lines
ADRIANA GIANTURCO, Director, California Department of Transportation
WILLIAM C. HENNESSY, Commissioner, New York State Department of Transportation
ARTHUR J. HOLLAND, Mayor, City of Trenton, N.J.
JACK KINSTLINGER, Executive Director, Colorado Department of Highways
THOMAS D. LARSON, Secretary, Pennsylvania .Department of Transportation
MARVIN L. MANHEIM, Professor of civil Engineering, Massachusetts institute of Technology
DARRELL V MANNING, Director, Idaho Transportation Department
THOMAS D. MORELAND, Commissioner and State Highway Engineer, Georgia Department of Transportation
DANIEL MURPHY, County Executive, Oakland County, Michigan
RICHARD S. PAGE, General Manager, Washington (D.C.) Metropolitan Area Transit Authority
PHILIP J. RINGO, President, ATE Management & Services Co.
MARK D. ROBESON, Chairman, Finance Committee, Yellow Freight Systems
GUERDON S. SINES, Vice President (In formation and Control Systems) Missouri Pacific Railroad
WILLIAM K. SMITH, Vice President (Transportation), General Mills
JOHN R. TABB, Director, Mississippi State Highway Department
CHARLEY V. WOOTAN, Director, Texas Transportation Institute, Texas A&M University

NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM

Transportation Research Board Executive Committee Subcom,nittee for the NCHRP
CHARLEY V. WOOTAN, Texas A&M University (Chairman) ELLIOTT W. MONTROLL, National Research Council
HENRIK E. STAFSETH, Amer. Assn. of State Hwy. and Transp. Officials A. SCHEFFER LANG, Consultant, Washington, D.C.
KARL S. BOWERS, U.S. Department of Transportation W. N. CAREY, JR., Transportation Research Board

Field of Traffic
Area of Safety
Project Panel, G 17-3
ARCHIE C. BURNHAM, JR., Georgia Dept. of Trans. (Chairman) MARTIN R. PARKER, JR., Va. Hwy. and Trans. Res. Council
WILLIAM T. BAKER, Federal Highway Administration DAVID E. PUGH, Washington State Dept. of Trans.
RICHARD D. BRUSTMAN, New York State Dept. of Trans. JAMES I. TAYLOR, University of Notre Dame
ALBERT BURG, Safety Consultant DAVEY L. WARREN, Federal Highway AdministratiOn
SIDNEY Q. KIDD, Mississippi State Highway Dept. DAVID K. WITHEFORD, Transportation Research Board
GEORGE E. MOBERLY, Illinois Dept. of Transportation JAMES K. WILLIAMS, Transportation Research Board
RICHARD D. PADDOCK, Ohio Dept. of Transportation

Program Stafi
KRIEGER W. HENDERSON, JR., Program Director ROBERT J. REILLY, Projects Engineer
LOUIS M. MAcGREGOR, Administrative Engineer . HARRY A. SMITH, Projects Engineer
CRAWFORD F. JENCKS, Projects Engineer ROBERT E. SPICHER, Projects Engineer
R. IAN KINGHAM, Projects Engineer HELEN MACK, Editor
NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM
REPORT 9
APPLICATION OF TRAFFIC CONFLICT
ANALYSIS AT INTERSECTIONS

WILLIAM D. GLAUZ AND DONALD JAMES MIGLETZ

Midwest Research Institute
Kansas City, Missouri

RESEARCH SPONSORED BY THE AMERICAN

ASSOCIATION OF STATE HIGHWAY AND
TRANSPORTATION OFFICIALS IN COOPERATION
WITH THE FEDERAL HIGHWAY ADMINISTRATION

AREAS OF INTEREST:
TRANSPORTATION SAFETY
OPERATIONS AND TRAFFIC CONTROL
TRAFFIC FLOW, CAPACITY, AND MEASUREMENTS
(HIGHWAY TRANSPORTATION)

TRANSPORTATION RESEARCH BOARD

NATIONAL RESEARCH COUNCIL
WASHINGTON, D.C. FEBRUARY 1980
NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM NCHRP Report 219

Systematic, well-designed research provides the most ef- Project 17-3 FY '78
ISSN 0077-5614
fective approach to the solution of many problems facing
ISBN 0-309-03016-1
highway administrators and engineers. Often, highway
problems are of local interest and can best be studied by
highway departments individually or in cooperation with L. C. Catalog Card No. 80-65311

their state universities and others. However, the accelerat- Price: $7.60
ing growth of highway transportation develops increasingly
complex problems of wide interest to highway authorities.
These problems are best studied through a coordinated
program of cooperative research.
In recognition of these needs, the highway administrators
of the American Association of State Highway and Trans- Notice
portation Officials initiated in 1962 an objective national The project that is the subject of this report was a part of the
highway research program employing modern scientific National Cooperative Highway Research Program conducted by the
Transportation Research Board with the approval of the Governing
techniques. This program is supported on a continuing Board of the National Research Council, acting in behalf of the
basis by funds from participating member states of the National Academy of Sciences. Such approval reflects the Governing
Association and it receives the full cooperation and support Board's judgment that the program concerned is of national impor-
tance and appropriate with respect to both the purposes and re-
of the Federal Highway Administration, United States sources of the National Research Council.
Department of Transportation. The members of the technical committee selected to monitor this
project and to review this report were chosen for recognized
The Transportation Research Board of the National Re- scholarly competence and with due consideration for the balance
search Council was requested by the Association to admin- of disciplines appropriate to the project. The opinions and con-
ister the research program because of the Board's recog- clusions expressed or implied are those of the research agency that
performed the research, and, while they have been accepted as
nized objectivity and understanding of modern research appropriate by the technical committee, they are not necessarily those
practices. The Board is uniquely suited for this purpose of the Transportation Research Board, the National Research Coun-
cil, the National Academy of Sciences, or the program sponsors.
as: it maintains an extensive committee structure from Each report is reviewed and processed according to procedures
which authorities on any highway transportation subject established and monitored by the Report Review Committee of the
may be drawn; it possesses avenues of communications and National Academy of Sciences. Distribution of the report is ap-
proved by the President of the Academy upon satisfactory comple-
cooperation with federal, state, and local governmental tion of the review process.
agencies, universities, and industry; its relationship to its The National Research Council is the principal operating agency of
parent organization, the National Academy of Sciences, a the National Academy of Sciences and the National Academy of
Engineering, serving government and other organizations. The
private, nonprofit institution, is an insurance of objectivity; Transportation Research Board evolved from the 54-year-old High-
it maintains a full-time research correlation staff of special- way Research Board. The TRB incorporates all former HRI3
activities but also performs additional functions under a broader
ists in highway transportation matters to bring the findings scope involving all modes of transportation and the interactions of
of research directly to those who are in a position to use transportation with society.
them.
The program is developed on the basis of research needs
identified by chief administrators of the highway and trans-
portation departments and by committees of AASHTO.
Each year, specific areas of research needs to be included
in the program are proposed to the Academy and the Board
by the American Association of State Highway and Trans-
portation Officials. Research projects to fulfill these needs
are defined by the Board, and qualified research agencies
are selected from those that have submitted proposals. Ad-
ministration and surveillance of research contracts are Published reports of the
responsibilities of the Academy and its Transportation
Research Board. NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM

The needs for highway research are many, and the National are available from:
Cooperative Highway Research Program can make signifi-
cant contributions to the solution of highway transportation Transportation Research Board
National Academy of Sciences
problems of mutual concern to many responsible groups.
2101 Constitution Avenue, N.W.
The program, however, is intended to complement rather
Washington, D.C. 20418
than to substitute for or duplicate other highway research
programs. Printed in the United States of America.
FORE\WORD Traffic engineers, safety specialists, highway designers, and researchers con-
cerned with traffic safety and operations at intersections will be interested in the
By Staff research findings presented in this report. Traffic conflict analysis, although still
Transportation in the developmental stage, is considered to have the potential of providing a
Research Board reliable and inexpensive tool to be used in lieu of, or in addition to, accident data
to diagnose safety and operational deficiencies and permit evaluation of improve-
ments within a short period of time. The traffic conflict analysis technique used
in this study was field tested to determine the acceptability of the conflict defini
tions, the practicality of the approach, and the accuracy of the results. A pro-
cedures manual and instructor's guide are provided in the appendixes to facilitate
future data collection efforts.

Previous research studies conducted in the United States, Canada, England,

and Sweden have developed a promising technique to evaluate intersection per-
formance, especially in regard to safety, through observation of conflicts between
vehicles as they pass through an intersection. Although not yet proven, traffic
conflicts may provide a surrogate for accidents so that an intersection's safety
characteristics can be determined without relying on historical accident reãords.
This approach would offer designers and traffic engineers the advantage of being
able to analyze an intersection immediately following its construction or improve-
ment rather than having to wait years for an accident history to develop.
In addition to the previous research that has been conducted on traffic con-
flict analysis, several highway agencies are using the technique on a routine basis.
However, conflicts definitions and sampling procedures vary significantly among
individual studies and applications. Therefore, the objective of this project was to
develop standard definitions in regard to what constitutes a traffic conflict and to
design a data collection procedure that will minimize individual differences in the
observation and recording of conflicts. Midwest Research Institute has fulfilled
that objective through a state-of-the-art review; development of candidate defini-
tions; comprehensive field studies (including the recruitment and training of con-
flicts observers); and, following an assessment of the collected data and the field
procedures, documentation of the findings in a form for direct use both in other
research projects and in practical applications. This report provides sufficient
information for individuals having no prior experience in the use of the technique
to initiate new applications. The procedures manual, instructor's guide, and sample
data collection forms provided in the appendixes should be particularly useful.
However, there remains a major deterrent to widespread application of this
technique. That deterrent is the lack of a proven, direct relationship between
accidents and conflicts. Without that relationship, an analysis of conflicting
maneuvers can only provide insights into the performance of an intersection.
By using the traffic conflicts technique in its present form, it is not possible to
reliably estimate the expected number of actual accidents or to quantify accident
costs for use in benefit-cost analysis. Future research to develop the needed
relationships is recommended.
 CONTENTS

1 SUMMARY

PART I
2 CHAPTER ONE Introduction and Research Approach
Problem Statement
Research Approach

4 CHAPTER TWO Findings

State of the Art
Development of Candidate Traffic
Conflict Definitions
Field Studies'
Field Study Results

14 CHAPTER THREE Interpretation, Appraisal, Application

Uses of Traffic Conflicts
Conflict Categories
Conflict Observations
Application of Conflicts Results
Training and Implementation

18 CHAPTER FOUR Conclusions and Suggested Research

Conclusions
Research Needs

19 REFERENCES

PART II

20 APPENDIX A Literature Review and Bibliography

26 APPENDIX B The State of Practice of Traffic Conflict Analysis

29 APPENDIX C Philosophical Considerations in Traffic Conflict

Definitions

34 APPENDIX 0 Theoretical Framework for the Accident—Con-

flict Relationships

36 APPENDIX E Conflict Definitions Used in Field Studies

45 APPENDIX F Field Studies

67 APPENDIX G Data Analysis Process and Results

84 APPENDIX H Procedures Manual for Traffic Conflicts Ob-

servers

101 APPENDIX I Instructor's and Engineer's Guide

ACKNOWLEDGMENTS

The research reported herein was performed under NCHRP Many members of MRI's staff helped in the conduct of the
Project 17-3 by Midwest Research Institute (MRI). William study. They are too numerous to give individual acknowledg-
D. Glauz, Associate Director, Safety and Engineering Analysis, ments. However, the authors would draw special attention to
was the principal investigator. He was assisted by Donald Michael C. Sharp for his experimental design and statistical
James Migletz, Assistant Traffic Engineer. analysis expertise; and Rosemary Moran for data management,
Grateful acknowledgment is made to the many individuals computer programming, and analysis. John C. Glennon, Over-
who have, in various ways, given ideas and otherwise played an land Park, Kansas, was a prime consultant to the project, and
important role in the development of this research. Included he contributed in a variety of ways, especially in the formative
are Christen Hydén of the Lund Institute of Technology, and conceptual aspects. Dr. James E. Aaron, Southern Illinois
Lund, Sweden; John Older and John Shippey of the Trans- University, also a consulting member of the project •team,
portation and Road Research Laboratory, England; Göran assisted in the training and testing portions of the project.
The efforts of the 17 conflicts observers, who put in long
Nilson of the National Swedish Road and Traffic Research
hours, sometimes under adverse conditions, were extremely
Institute, Sweden; and Professor Ezra Hauer of the University helpful.
of Toronto, Canada. Finally, special thanks also go to William T. Baker, Fed-
Special thanks are extended to the traffic engineers from eral Highway Administration, who has been involved in the
several states who have been active in the TCT field and who TCT almost from its inception, and is one of the key persons
freely shared their experiences and ideas. Special thanks are pushing for a better understanding of its applicabilities and
due to David Pugh (Wasington), Richard Paddock (Ohio), limitations. His insight, assistance, and enthusiasm were most
Charles Zegeer (Kentucky), and Martin Parker, Jr. (Virginia). appreciated.
 APPLICATION OF TRAFFIC CONFLICT
ANALYSIS AT INTERSECTIONS•

SUMMARY Traffic accident data are often not suitable for diagnosing safety problems at
intersections or for evaluating the effectiveness of improvements. At times, a more
rapid approach or sensitive measure is desirable. The traffic conflicts technique
(TCT) has .been developed and refined with this application in mind. Viewed
simply, a traffic conflict is a traffic event involving the interaction of two vehicles
where one or both drivers may have to take evasive action to avoid a collision.
The objective of NCHRP Project 1 7-3 was to develop a standardized set 6f opera-
tional definitions and procedures that would provide a cost-effective method for
measuring traffic conflicts. This report provides the definitions and procedures
developed to meet this objective.
The research included the proposal of various candidate TCT definitions and
procedures, and the conduct of extensive comparative field tests. Over 9 weeks of
field data were collected using 17 traffic conflict observers trained for this specific
purpose. They obtained data at more than 24 intersections having a variety of
geometric and traffic control configurations.
Analysis of the data collected led to a recommended set of traffic events that
should be observed and recorded, together with procedures for analyzing these data.
The traffic conflicts in the recommended set all have very high observer reliability.
In other words, after undergoing a modest amount of training, most persons at the
traffic technician level should be able to observe and record these events in nearly
the same way.
Traffic conflicts, as stochastic traffic events, vary quite markedly in number
and rate from day to day even under nominally identical conditions, just as do
other traffic events such as accidents and turning volumes. Therefore, even though
an individual observer could become extremely profficient and produce very repeat-
able results from day to day, it is important operationally to recognize that the
events themselves are not totally repeatable. This implies that—depending on site
characteristics, traffic volumes, and the conflict types of interest—from several
hours to several days of data collection may be required before reasonable con-
fidence should be placed on the findings.
Traffic conflicts are viewed as supplements to, rather than replacements for,
accident data. Although the recommended traffic conflict definitions and proce-
dures have high face validity as indicators of safety hazards, the necessary research
has not yet been conducted that would firmly tie such events in a statistical fashion
to traffic accidents. Nevertheless, preliminary data obtained in this study (and in
others) suggest that such relationships may indeed exist. In particular, accidents
involving cross traffic and opposing left turns may be correlated to traffic conflicts
of an analogous nature. Rear-end conflicts, however, are not clearly relatable to
their analogous accidents, but they may be useful in diagnosing the causes for
rear-end accidents.
The study developed preliminary estimates of the traffic conflict rates (which
are better measures than traffic conflict numbers) for sites that have certain geo-
metric configurations and traffic control devices. Further research is recommended
to refine these estimates and extend them to a more comprehensive set of inter-
section parameters.
 A procedures manual was developed for the use of agencies and traffic conflict
observers planning to use the technique. This manual contains complete opera-
tional definitions and descriptions together with recording forms and detailed
stepby-step procedures for the conduct of a traffic conflicts count. In addition, an
instructor's guide was prepared for use in training persons in the traffic conflicts
technique and in applying the technique.

CHAPTER ONE

INTRODUCTION AND RESEARCH APPROACH

PROBLEM STATEMENT search review continued throughout the duration of the

project, as additional work was accomplished throughout
Traffic accidents are the most direct measure of safety
the world.
for a highway location. But, attempts to estimate the rela-
The gathering of literature and the meeting with the
tive safety of a highway location are usually fraught with
international group was supplemented by on-site observa-
the problems of unreliable accident records and the time
tions with users of various forms of TCT. In Europe, first-
required to wait for adequate sample sizes. For these rëa-
hand experience was obtained in Sweden and in England
sons, the traffic conflicts technique (TCT) was developed
with researchers in those countries. Extended visits and
in an attempt to objectively measure the accident potential
field trips were made to the most active users in the United
of a highway location without having to wait for a suitable
States, which were the States of Washington, Ohio, and
accident history to evolve.
Kentucky.
Over the past 12 years, several agencies have used the
TCT as either an operational or experimental tool. This Task 2—Selection of Candidate Definitions—On the
activity, which has been international in scope, has been basis of the information gained in Task 1, a determination
dominated by research efforts as opposed to operational was made of candidate TCT definitions, techniques, and
applications. Numerous alternatives, together with esti- procedures to be examined in the remainder of the con-
mates of their applicability and practicality, have evolved tract. Definitions and philosophies abound regarding TCT,
from this research. its operational aspects, and its applicability to the serving
The objective of this research was to develop a stan- of many purposes. From this universe of viewpoints a
dardized set of definitions and procedures that would pro- rather large set of definitions, techniques, and procedures
vide a cost-effective method for measuring traffic conflicts. was selected that could be applied, in principle, by trained
A TCT method was required that would be both reli- observers at the paraprofessional or technician level—each
able and repeatable for diagnosing safety and operational was carefully defined, both in concept and from an opera-
deficiencies and for evaluating the effectiveness of im- tional viewpoint.
provements at intersections. The method also should be Task 3—Performance of Comprehensive Field Stud-
implementable through application of a readily usable ies—This task involved the conduct of a comprehensive set
procedures manual that clearly and concisely describes data of field studies testing the candidate definitions and pro-
collection procedures, analysis techniques, and evaluation cedures to obtain data pertaining to their reliability (ob-
methods. server variance) and repeatability (site variance) for a
variety of intersection conditions and characteristics. This
RESEARCH APPROACH was the major undertaking of the project, involving five
substantial subtasks and about 40 percent of the total
The objective was achieved through the accomplishment research effort and expenditure.
of the following five tasks: The first subtask was the development of a detailed plan
1. Task 1—State-of-the-Art Review—This task entailed for the conduct of the remaining subtasks.
a thorough review and examination of all research and op- The second subtask was to develop the experimental de-
erational activities currently in existence concerning traffic sign and select experimental sites. These activities had to
conflicts. It was begun with a formal literature search, proceed simultaneously because, although onecan suggest
augmented with tracking down foreign and U.S. reports experimental parameters of interest and assemble them in
not widely distributed, but known to the research team and formal ways that facilitate later statistical analyses, in ac-
members of the project panel. tuality certain intersection characteristics are more preva-
By coincidence, an international workshop on traffic con- lent than others. Thus, some situations (such as 2-lane,
flicts (1) was held at the onset of the contract, which 3-way, signalized intersections) exist rarely in the real
brought to light a great deal of additional research of an world, so deserve lesser attention. The final approach was
international nature, and, as well allowed the ad hoc diffu- to tailor the design somewhat to the availability of actual
sion of ideas and concepts between investigators. The re- intersections in the greater Kansas City metropolitan area.
 Much attention was paid to the process of locating, mine reliability and repeatability of candidate,
screening, recruiting, training, and testing of TCT observ- basic traffic conflict categories.
ers for the field studies (the third subtask). Advertising of Further statistical analyses of variance to deter-
a nonspecific nature was used (through the Missouri Em- mine reliability and repeatability of derived or col-
ployment Office), along with notices in high schools, col- lapsed categories of observed traffic conflicts.
leges, and retirement homes and complexes. A multiple- Examination of conflictcounts and categories as
step screening process was employed to find persons who they relate to intersection parameters as well as to
were available for 3-month (summer) employment and individual site characteristics.
were 18 years of age or older, had an automobile and an Characterization of most promising conflict cate-
operator's license, and had some work experience. Per- gories with respect to intersection parameters,
sonal interviews with over 80 applicants were conducted, available improvement alternatives, traffic vol-
24 persons were offered jobs, and 19 persons accepted. umes, and accident experience.
A 2-week training program followed, which heavily em- Selecting those conflict categories that are highly
phasized in-the-field experience and careful monitoring and rated from both reliability and repeatability stand-
testing of progress. The program also utilized specially points and that are also practical and economically
developed 16-mm training movies, 35-mm slides and video- efficient.
tapes, and a videotape obtained from Sweden illustrating 5. Task 5—Preparation of Project Documentation—This
special concepts. One person dropped out and another was task included the preparation of a final research report, a
released because of inability to comprehend the concepts. TCT procedures manual, and an instructor's guide for
The remainder were utilized in the field experiments (al- training TCT observers. This report incorporates all of this
though three received an additional 1 to 2 days of super- documentation.
vised field practice beyond the formal training program). Chapter Two summarizes the findings of the research
The fourth subtask was to develop the field study pro- activities. Special attention is paid to the following: the
cedures. This activity relied heavily on practices presented existing state of the art as appraised through current usage
in the literature and/or observed in current operation, with and knowledge; statistical and other findings concerning the
adaptations to fit the experimental nature of this project. tested operational definitions of the TCT; and site-specific
Thus, special recording forms were developed that: (a) fit characteristics of traffic conflicts.
certain types of intersections, (b) allowed direct recording Chapter Three appraises the research findings from an
of conflicts meeting certain requirements, (c) enabled the operational viewpoint, recommends procedures for making
simple determination of other conflict categories of inter- and analyzing traffic conflict counts, and indicates manage-
est, and (d) included subsidiary information that would ment steps required by an agency to implement a TCT pro-
facilitate the later analyses. Daily operating practices were gram. The last chapter presents the project conclusions and
outlined, time tables were drawn up, and workday activity suggests areas where future research might further the use
lists and equipment needs were detailed. Finally, individ- of the TCT as a diagnostic/evaluative tool in traffic safety
ual, daily assignments were made for each observer in and traffic operations.
accordance with the general plan and the experimental The appendixes provide substantiating and augmenting
design. material to this report. The first seven appendixes give
The final subtask was the actual conduct of field experi- backup details; the last two include material for use in im-
ments. The experimental plan anticipated three phases of plementing the project recommendations.
3 weeks each, for a total of 9 weeks during the summer of Appendix A is a comprehensive and critical review of the
1978. Data collection began on Monday, June 12, and was TCT literature, which includes a bibliography and a topi-
scheduled to end Thursday, August 10. Because of weather cal cross reference matrix. It has a companion Appen-
and other problems, make-up work was scheduled from dix B, which concentrates on the TCT as an operational
August 15-17. device, as distinct from a research tool.
4. Task 4—Analysis of Field Data and Generation of Appendix C examines traffic conflicts in a philosophical
Recommendations—The analysis process that was devised context. Questions considered include, "What is a traffic
included the following major steps, some of which were conflict?" "What attributes should it have?" "What re-
subdivided in various ways: search problems have been faced before?" "How are they
Field screening and peer-review of data, on an related to accidents?" "How can they be measured?" and
essentially as-collected time frame. "What are the sources of error?" Appendix D gives mathe-
Supervisory review of field data, typically within matical insight into the accident—conflict relationship, and
24 to 48 hr. concludes with estimates of the tradeoffs between using
Data encoding and keypunching. conflicts and historical accidents to predict future accidents.
Data listing, by site, and manual scanning for in- Appendix E provides detailed descriptions of the TCT
consistencies. definitions used in the field studies for this project. Ap-
Elimination of conflict categories with obvious pendix F presents numerous details of the planning for, and
flaws for operational utility, such as rarity of oc- conduct of, the field studies.
currence or high observer unreliability, as deter- Appendix G is rather extensive because it includes the
mined by visual inspection of data summaries. analysis concepts, procedures, and detailed results.
Formal, statistical analyses of variance to deter- The last two appendixes provide implementation mate-
rial. Appendix H is a set of instructions and related ma- mented by Appendix I, an instructor's manual that in-
terials that may be adopted by persons trained to make corporates suggested training concepts, materials, and
traffic conflicts counts using the recommendations of this schedules.
project. It is a user's manual. This material is supple-

CHAPTER TWO

FINDINGS

STATE OF THE ART What are the basic applications of traffic conflicts?
(Accident prediction, hazardous location identification,
More than 60 documents related to traffic conflicts were
hazardous location diagnosis, traffic improvement evalua-
obtained, reviewed, and summarized. In addition, several
tion.)
states currently using traffic conflicts operationally were
To what specific types of applications do traffic con-
visited and operational experience was acquired. The find-
flicts lend themselves? (Intersections, construction zones,
ings from these two activities are detailed in Appendixes
acceleration lanes, lane drops, pedestrian and pedalcyclists,
A and B, respectively, and are summarized as follows.
warrants for traffic control devices, the movement of over-
size loads, etc.)
Historical Developments

In 1967, Perkins and Harris (2) of the General Motors Operational Applications
Corporation wrote the first paper on traffic conflicts. Al-
Although some European countries have applied traffic
though the traffic conflicts technique (TCT) was originally
conflicts to the solution of operational problems, appar-
developed to investigate whether General Motors vehicles
ently the only routine operational applications are in the
were driven differently from others, the method was soon
United States. The States of Washington and Ohio have
considered more appropriate to evaluate accident potential
had the most experience in this regard, although several
and operational deficiencies of intersections without de-
other states such as Virginia and Kentucky have done ex-
pending on accident data for these analyses. Then, in 1969,
tensive work, also. Both Washington and Ohio maintain
the U.S. Federal Highway Administration awarded con-
teams of technicians who are trained in traffic conflicts tech-
tracts to the States of Washington, Ohio, and Virginia to
nique and who more or less 'routinely are sent out to various
aid in the evaluation of the technique (3). Since that time,
intersections of interest to obtain conflicts data (5, 6). Both
the technique has generated a great deal of national and
states use slightly modified versions of the original GM
international interest, with many countries funding exten-
technique. Most of their applications are for diagnosing
sive research into various aspects of the TCT.
safety problems at intersections, after the intersection has
Many technical meetings have been held on the subject.
been singled out because of accident data or public com-
A traffic conflicts seminar was held in Washington, D.C., in
plaint, or for evaluating intersection improvements. Each
March 1974. Subsequently, the Transportation Research
state has developed its own detailed data collection and
Board formed subcommittee A3Al2(l), as part of its
analysis procedures, and has accumulated an extensive
technical committee structure, just to deal with this subject.
amount of experience and data over the last several years.
An international workshop was held in Oslo, Norway, in
September 1977, and proceedings were published (1). Re-
DEVELOPMENT OF CANDIDATE TRAFFIC
cently (May 1979), a second international workshop was
CONFLICT DEFINITIONS
conducted in Paris, France.
In the broadest sense, a traffic conflict is a traffic event
Recent and Current Research involving the interaction' of two vehicles, where one or both
drivers may have to take an evasive action to avoid a col-
The national and international research involving or us-
lision. Traffic conflicts do not cause accidents, but they are
ing traffic conflicts techniques is extensive. It includes the
following major directions: probably symptomatic of the same things that cause or con-
tribute to accidents. In a sense, an accident is simply a
What are the relationships between traffic conflicts conflict where the evasive action was too little or too late.
and accidents? This section of the report presents some of the philo-
What are the "best" definitions of traffic conflicts? sophical and statistical considerations in traffic conflict' defi-
How should traffic conflicts be measured? (For ex- nitions, which are given in more detail in Appendixes C
ample, using human observers, film or videotape, auto- and D. Then, the final set of operational traffic conflict
mated sensors, etc.) definitions and categories that were used in the field tests
is presented. Detailed descriptions of these are included in The last three requirements are the major focus on this
Appendix E. research effort. The second requirement was also consid-
ered, to the extent that the limited amounts of data per-
Fundamental Attributes of TCT Definitions
mitted. However, the relationship between traffic conflicts
There is great divergence of opinion, philosophically, and accidents was not central to this research; rather, it
about traffic conflict definitions. One school of thought (7) was implicitly assumed that there is some correspondence
is that a proper definition of a traffic conflict must ensure between the observation of evasive action and the presence
that every accident be preceded by a conflict. Although this of safety or operational deficiencies at an intersection.
is an appealing concept, it can lead to unrealistic data Having agreed on the basic attribtues to be satisfied by
collection requirements. traffic conflicts, it next is important to state a generalized
Another viewpoint is to examine the purpose of using the definition of a traffic conflict. This generalized definition
traffic conflict technique. Some believe that traffic conflicts will form the basis for specific operational definitions. One
are a surrogate measure of accidents—that is, they are excellent generalized definition was generated at the inter-
events whose counts are directly comparable to accident national conference in Oslo (1):
counts. However, attempts to find strong correlations be-
A traffic conflict is an observable situation in which two
tween conflicts and accidents have, for the most part, been or more road users approach each other in space and
either unfruitful or misleading for a number of reasons (8). time to such an extent that there is a risk of collision if
Such attempts, nevertheless, should be continued because their movements remain unchanged.
of the potential benefits achievable even if relationships The foregoing definition was modified slightly, to add a
found are relatively weak. The science of predicting acci- little more specificity for purposes of this research:
dent rates or numbers is far from perfect. Because of in-
herent statistical (random) fluctuations in such events, even A traffic conflict is a traffic event involving two or more
historical accident data themselves are usually inadequate road users, in which one user performs some atypical or
unusual action, such as a change in direction or speed,
for making accurate estimates of future accidents. This is that places another user in jeopardy of a collision unless
true unless the numbers involved are very large, which is an evasive maneuver is undertaken.
seldom the case at individual intersections or even for small
groups of intersections having some common attributes. The latter definition has several important implications.
Best available data suggest that even with imperfect rela- First, the unusual action instigating the conflict is one that
tionships between traffic conflicts and accidents, and after the typical road user would not perform under the same
accounting for inherent conflict count variability, conflict circumstances, such as precautionary braking by a motorist
counts obtained by well-trained observers are likely to pro- driving through an intersection even though there is no
duce more accurate predictions of accidents than historical cross traffic. This restriction does, however, rule out cer-
accident data (see App. D). tain types of movements that most users initiate under the
Despite the promise of using traffic conflicts as accident same conditions, such as stopping for a stop sign. It is not
surrogates, it is probably more acceptable, at least in the necessary according to this definition that there actually be
United States, to view traffic conflicts as logical indicators an evasive maneuver or an impending collision. It suffices
of safety or operational problems, even if the relationship that the instigating action or maneuver threatens another
cannot yet be placed on sound statistical grounds. This is user with the possibility of a collision, and thereby places
because operational definitions of traffic conflicts not only the user in the position of probably taking some evasive
maneuver.
must imply a safety-related attribute, they also must satisfy
several other desirable attributes. Altogether there are at The definition excludes "evasive maneuvers" that are
least five: strictly precautionary in nature. For example, it does not
include braking or swerving of a through-vehicle in re-
Safety-relatedness—At least in a conceptual sense, sponse to an anticipated opposing left turn, instigated by
conflicts should be related statistically to accidents. the opposing driver turning his wheels, but not encroaching
Site-relatedness—They should be useful in diagnosing on the lane of the through-vehicle. Also, it does not in-
problem locations or measuring the effectiveness of a site clude braking or swerving occasioned by the presence of a
improvement. stopped vehicle on a cross street, which may "threaten" to
Reliability—The definition should provide minimum encroach but does not actually do so. Another general class
variation between different observers when recording the of exclusions are violations such as "run red light" and "run
same event. Reliability is dependent on the explicitness of stop sign" unless such violations occur in the presence of
conflict definitions, technique tractability, and observer a through-vehicle that is placed in jeopardy of a collision.
knowledge.
Repeatability—The definition should result in an ac-
Candidate Conflict Categories
ceptable level of variation in repeated observations by the
same observer at the same site under nominally identical If they are to be implemented widely in the United
conditions. This attribute has an important impact on States, operational definitions must avoid or minimize the
determining meaningful sample sizes. use of sophisticated equipment or painstaking measure-
Practicality—Reliable, repeatable, safety-related, and ments, whether in the field or later in the office. Thus,
site-related data should be obtainable in a reasonable time operational definitions must be suitable for direct applica-
with reasonable resource expense. tion by human observers. Moreover, such definitions
should not require highly educated and experienced traffic
engineers as observers; they should be amenable to use by
persons such as traffic technicians, with suitable training.
Therefore, operational definitions must encompass readily
observable events.
To be observable, the traffic event must elicit an evasive
maneuver (braking or swerving) by the offended driver.
In this respect, the operational definition is more restric-
tive than the generalized definition, and is like the GM
work (2, 3).
An intersection traffic conflict can then be described,
operationally, as a traffic event involving several distinct
stages:

One vehicle makes some sort of unusual, atypical, or

unexpected maneuver.
A second vehicle is placed in jeopardy of a collision.
The second vehicle reacts by braking or swerving.
The second vehicle then continues to proceed through
the intersection.

The last stage is necessary to convince the observer that the

second vehicle was, indeed, responding to the offending Figure 1. Left-turn, same-direction conflict.
maneuver and not, for example, to a traffic device.
Within this framework, a basic set of operational defini-
tions can be stated, corresponding to different types of in-
TABLE I
stigating maneuvers. One example, called a left-turn, same-
direction conflict occurs when an instigating vehicle slows BASIC INTERSECTION CONFLICTS
to make a left turn, thus placing a following, conflicted
Left turn, same direction
vehicle in jeopardy of a rear-end collision. The conflicted
vehicle brakes or swerves, then continues through the inter- Right turn, same direction
section (see Fig. 1). A total of 13 basic conflicts was de-
fined as candidates to be field tested, as described in Table 1 Slow vehicle, same direction
and in Appendix E.
For each basic type of conflict there can also be defined Lane change
a traffic event called a secondary conflict (it is comparable
Opposing left turn
to the GM-defined previous conflict). The secondary con-
flict involves an additional vehicle that is conflicted with by
Right turn cross traffic, from right
an instigating vehicle that slowed or swerved in response
to some other conflict situation. Each secondary conflict is Left turn cross traffic, from right
identified by the type of initial conflict with which it is
associated. Thru cross traffic, from right
Alternative operational definitions were also used in the
field tests to determine their value relative to these basic Right turn cross traffic, from left
definitions. For each of the 13 basic types of conflicts,
Left turn cross traffic, from left
other more restrictive or less restrictive definitions were
examined.
Thru cross traffic, from left
For the same-direction conflicts in Table 1, the original
GM work specified that the vehicles must be traveling as Opposing right turn on red (during protected
a pair (e.g., a "car-following" situation must exist). In
actual practice, however, some users prefer to include all left turn phase)
situations where a second vehicle brakes or swerves even
if it came upon the leading vehicle several seconds later. Pedestrian
The all-inclusive definitions are less restrictive, and include
the GM definition that requires paired-vehicle situations.
For the other types of conflicts in Table 1, the GM study scribed; they become such only if an opposing vehicle is
suggested counting vehicles. Many users do this, believing relatively close and it reacts by braking or swerving.
that such data may be useful. An alternative terminology is The foregoing descriptions cover 39 different operational
suggested, "opportunities." Thus, for example, all opposing definitions of traffic conflicts. All were used, but pedestrian-
left turns (except during a protected-left-turn phase) repre- related conflicts were so rare that they were not analyzed,
sent opportunities for traffic conflicts as previously de- leaving 36 definitions for analysis.
 Another set of 13 traffic conflicts witha more restrictive descriptive parameters shown in Figure 2. This figure also
definition is the subset that exceeds some threshold level of includes the four additional sites used in a subsidiary ex-
"severity." Of the numerous approaéhes toward develop- periment. Each intersection was observed from two op-
ment of descriptive definitions of severe conflicts in the posing legs, and the number of traffic lanes of each of these
United States and elsewhere, it was determined that most legs was one of the parameters. The presence or absence
promising was that developed by Hydén (9) in Sweden— of signalization (vis-a-vis stop-sign control for the cross
a time-to-collision measurement, defined as the time inter- traffic) was a second parameter, and the geometric con-
val from when a conflicted vehicle reacts (brakes or figuration (3-way vs. 4-way) was a third. The speed limit
swerves) until a collision (or a near miss) would have oc- on the observed legs was another parameter, with 40 mph
curred had there been no reaction. Specifically, a conflict being the dividing point between high and low speed. Thus,
was defined to be severe if the time-to-collision value was there were 12 combinations of factors in the basic experi-
less than the threshold of 1.5 sec, as determined subjec- ment—three groupings of number of lanes and traffic con-
tively by trained observers. trol, two speeds, and two geometric configurations. Two
In addition to the foregoing conflict categories, all of sites (replicates) were used for each combination, giving a
which were observed and recorded in the field, a number total of 24 sites. Traffic volumes were also monitored, but
of others were created by grouping or collapsing categories used an implicit rather than an explicit variable. Most of
in the analysis process. For example, for each of the rear- the sites were in rural and suburban areas. There Were
end categories, "paired-vehicle" and "not-paired-vehicle" some in areas zoned for business or industry, but none in
conflicts were actually recorded. These two, plus their the CBD.
combination, were analyzed, however. Then, all of the The basic experiment used 12 trained observers who
rear-end conflicts were combined in various ways and worked in pairs, alternately (every half-hour) viewing from
analyzed. Similarly, the cross-traffic conflicts were grouped opposing legs of the intersections. Each observer collected
in several ways. Further, secondary conflicts were com- data at a specified site for half a day with a designated part-
bined with their causative conflict categories. Finally, sev- ner, then moved to another site for half a day to work with
eral kinds of "grand totals" were also created. All told, a different partner. With 12 observers, each of the 24 sites
62 conflict categories were subjected to formal analysis (see could be visited once every 2 days (4 half-days), and each
Table G-1 for a complete listing), not including the severe observer could be paired with each of the others once every
conflicts that were analyzed separately manually. 11 half-days.
These factors were combined with a 4-day 40-hr weekly
FIELD STUDIES
schedule to create a basic experimental phase of 3 weeks,
in which each site was observed for three mornings and
Extensive field tests were conducted during the summer three afternoons, and each observer worked with every
of 1978 to obtain data on the candidate operational defini- other observer at least twice. Three such phases were con-
tions of traffic conflicts. These field tests, conducted in the ducted, the results of which could be analyzed separately
greater Kansas City metropolitan area, used observers, and/or compared or combined.
without special abilities or experience, who received a This plan enabled the determination of the contributions
2-week training program designed and administered as part of each of many factors to the various conflict count cate-
of this project. gories, to the variability in the counts, and the statistical
significance of these quantities. Mathematically, the vari-
Experimental Plan ance, o.,2, obtained as a result of repeated, short observa-
tions of the same type of conflict, Y, over a period of weeks
The basic experiment involved 24 intersections with the at numerous sites, by different persons, and at different

- High Speed Low Speed

Experiment Lanes 4-Way

Basic 4 No X X X

4 Yes X X X

2 No X X X

Subsidiary 2 Yes X - X

a! Each X represents two physical sites, each with two legs or approaches

being observed.
Figure 2. Experimental design frame work.
times of day of different days, can be assigned to the iden- which had been analyzed by the Institute to determine the
tifiable factors according to their numerical combinations time-to-collision value.
to 0-,,2. That is: There were procedures of several types. Early in the
training program several "quizes" were given. Videotapes
cr2 = cr2 + crt 2 + D2 + (TN 2 + . . . + a (1)
were displayed and the trainees were asked to observe and
where each term on the right side of Eq. 1 has a particular record conflicts. But the major test procedure involved the
meaning: field work itself. The field experiences emphasized small
group observations with individual recording of results.
= observer variance (reliability)—the variation due The composition of .the groups was continuously changed.
to systematic biases between observers;
All field work was turned in and scored on a comparative
= the variance between the short observation inter- basis, using statistical techniques. Conflict categories re-
vals at a site;
ceiving high interobserver variances were singled out for
= the variance between days of week at a site; more intensive review and discussion. Individuals, on the
N2 = the variance between 3-leg and 4-leg sites;
0-92 = the variance between low-speed and high-speed
other hand, who appeared to have discrepant counts in one
or more categories were given extra attention to resolve the
sites;
difficulties.
= the variance between 2-lane and 4-lane, unsignal- The field tests, in particular, were relied on very heavily
ized intersections;
in determining the 12 persons who would form the basic
0.
e2= the variance between legs at a site; team of observers, and whether or not the remainder would
0ü2 = the variance between 4-lane intersections with sig-
be suitable substitutes and "extra" observers.
nalized and unsignalized traffic;
0-2 = the variance between "replicate" sites of nominally Conduct of the Studies
the same type (same speed, number of lanes, and
traffic control); and Data were collected on Monday through Thursday dur-
= residual variance, or error, which is the "repeat- ing the first and third 3-week phase, and on Tuesday
ability" sought by the project. This is the variance through Friday during the second phase. The 12 regular
of repeated observations by the same observer observers worked, in pairs, at two of the 24 basic test sites
under (theoretically) identical conditions (same each day. The remaining five observers were given daily,
physical site, same time of day and day of week, special assignments. These people observed conflicts at the
etc.). extra locations, replaced regular observers in emergency
situations (sickness, car trouble), performed make-up work,
Observer Training and Testing and did project-related office work.
The 10-hr workday started at 0700 hr and ended at
A training program lasting 9 days was devised far the 1800 hr. Four hours of observation occurred at each morn-
19 persons hired for this purpose. One dropped out. An- ing and afternoon location. The remainder of the workday
other was released, and three received an additional 1 to was allocated to travel time, preparation time, and sched-
2 days of training because of results from a testing pro- uled breaks. Each morning and afternoon was divided into
gram, described subsequently. The observers were male eight half-hour count periods. In each half-hour period,
and female, and most were relatively young (early 20's). 15 min were spent observing conflicts that were then re-
The training, itself, is described in Appendix F. Multi- corded; 5 min were spent counting traffic volumes that were
modal instruction was used, but supervised field practice then recorded; and the remaining time was used to move
was emphasized. Of the 72 total hours allocated to train- to the opposite approach of the intersection. The two
ing, 42 hr (58 percent) were spent in the field. The re- people at each intersection observed conflicts and traffic
mainder of the hours were mostly in the classroom and volumes on the same street and at the same time, but from
involved lecturing; question and answer periods; discussion opposite approaches. They switched approaches every
sessions; testing and demonstrations using slides, films, and half-hour.
videotape. The four types of conflict forms used in the study (as
Because no totally satisfactory training films on traffic shown in Figs. F-8 through F-il) were specifically de-
conflicts were available, one was created using staged signed for different types of intersections. Each of the
vehicle-vehicle and vehicle-pedestrian conflicts. This film is forms illustrates the categories of conflicts that are most
an edited, 17-mm, silent 16-mm color film with subtitles. common to the appropriate type of intersections. The
The conflict categories are presented in logical groupings to forms also contain blank columns for field use, so that
facilitate comprehension and review. A parallel, 35-mm other conflict categories can be added if observed. In addi-
slide presentation was also used. tion, the observers were encouraged to write or diagram
Several videotapes were used. One was created during a any of their observations not covered by the formal defini-
field visit to Ohio, and it contains traffic conflicts observed tions, and to add notes and comments.
at several intersections. Videotapes were also made in Kan- Daily site visits were made by at least one member of the
sas City at several selected intersections. Finally, a video- project staff. During this time, questions were answered
tape created by the Lund Institute of Technology in Sweden and supplies replenished. The observers were provided
was acquired. This tape, designed to help teach the severe with the office and home telephone numbers of the project
conflict concepts, contains 41 actual conflicts, each of staff and urged to call if any problems needed immediate
attention. On days when it rained, observation stopped Analyses were also performed of their distributions
until the roads appeared to be drying. among observers, time periods, and days of the week, tak-
The individual site visits provided good contact, but only ing into account the relative frequencies of observation
on a one-to-one basis, so the observers had little oppor- (e.g., counts were taken on nine Tuesdays, Wednesdays,
tunity to maintain contact with the group after the training and Thursdays, but only on six Mondays and three Fri-
program. Therefore, a periodic newsletter was prepared days). The results detailed in Part 6 of Appendix 0
and distributed to all observers, which included any gen- showed no significant differences by day of week. Further-
eral statements concerning policy and procedures, as well more, they were rather uniformly distributed throughout
as other items of interest. (Two newsletters are shown in the morning and early afternoon hours, with no marked
Figs. F-l3 and F-14.) morning peak. However, they began to be much more
The first phase proceeded smoothly with only minor prevalent by midafternoon (2:30 to 3:00 p.m.) and peaked
operational difficulties. The second phase was interrupted greatly in the late afternoon, as shown in Figure 3.
by weather problems (rain), although most of the data Analyses showed significant differences between observ-
were collected as planned. The third phase was concerned ers, with 4 of the 17 observers recording essentially half
with numerous problems including frequent rains. One (51/104) of the severe conflicts. Thus, these measures
site was lost completely because of a nearby bridge failure suffer from a lack of reliability, as well as being infrequent
and thus a road closure. Another site was essentially lost and not site-discriminating.
because roadside construction work necessitated traffic di- Severe conflicts were also examined to determine if they
version. Although significant attempts were made to "make were distributed among "types" in the same way as regular
up" this data collection by using time extensions and conflicts. For this purpose, four groupings were used:
pressing the extra observers into heavier service, it was rear-end or same-direction conflicts; opposing left-turn con-
finally decided to use these data to fill voids in the first two flicts; cross-traffic-from-right conflicts; and cross-traffic-
phases and to enable the conduct of auxiliary analyses from-left conflicts. The analysis showed that the distribu-
pertaining to observer reliability and to certain site parame- tions were greatly different. Whereas about 83 percent of
ter effects. all conflicts were of the same-direction variety, only 55 per-
cent of the severe conflicts were of this type. Instead, the
FIELD STUDY RESULTS severe conflicts were more likely to be of the cross-traffic or
opposing left-turn variety—especially the latter, which
The findings from the analyses of the field study data are comprised 18 percent of the severe conflicts as opposed to
summarized in this section. These analyses dealt with 4,000 5 percent of the regular conflicts.
observer-hours of conflict and volume counts. Appendix G
contains more detailed tabulation and discussion of these Rarely Observed Conflict Categories
results.
There were 36 basic conflict categories recorded rou-
tinely during the field tests. Table 2 lists those that should
Severe Conflicts
be dropped as useful concepts, because they are so rare as
Most researchers and users of TCT utilize some type of to be impractical observational measures. Essentially, the
severity classification. European users tend to place more tabulated conflicts each occurred, at most, only about once
reliance on such classifications than do American users, for every 8 observer-hours of observation, equivalent to
who have not been able to relate them to other traffic about 2 workdays. This observation rate is not considered
events. The findings of this research concur with the latter. useful in a practical sense.
A grand total of 104 severe conflicts at the 28 test sites The two major exceptions to this 8-hr limitation are the
was noted, an average of about one per 18 observer-hours right-turn-from-left opportunity and the right-turn-on-red
of observation. Six of these 104 were accidents. Chi-square opportunity. Further examination of the data indicated
analyses showed that there were no significant differences in that, in addition to the relative rarity of these events, the
the counts attributable to the factors characterizing the interobserver variance was unusually high. The majority of
sites, such as the number of lanes, presence or absence of these (few) counts were obtained (probably erroneously)
signalization, etc. Analyses of the severe conflicts at the by just a few of the observers. The definitions of these
24 basic sites (an average of four per site) showed no per- events are apparently difficult, conceptually, leading to low
vasive evidence that the location is associated with severe reliability.
conflict counts. The number per site ranged from 1 to 8 It is noted that Table 2 includes all the secondary con-
with astandard deviation of 2.02 (variance of severe con- flicts categories tested, except for those of the rear-end
flicts per site of 4.09). If the severe conflict rates were the variety. This implies that, in general, secondary conflicts,
same at all sites and were distributed Poisson at each site, by themselves, are not generally useful measures. The table
the collection of 24 site results would also be distributed also includes all three conflict categories related to the spe-
Poisson. This assumption is not rejected by a goodness of cialized right-turn-on-red and right-turn-from-left concepts,
fit test (x2 (6) = 3.88). That is, based on the extensive and two of the three categories of the lane-change type.
amount of data collected during the experiment,, no particu- Certain types of conflicts and other traffic events were
lar site and no particular site characteristics could be iden- not recorded routinely because of their expected rarity.
tified as being particularly hazardous using severe conflicts With one exception, this assumption proved correct. The
as a criterion. exception was at a 3-way site where a driveway formed, in
 10

0.20 TABLE 2
C
.2
6
1 RARE CONFLICTS
S
0.15 - Conflict Type -- Observer Hours
per Occurrence

o Right-Turn-on-Red Secondary Conflict 001/

0.10 .','.
0. Right-Turn-from-Left Secondary Conflict 250.0
U
Lane-Change Secondary Conflict 62.5
C
0.05 /\•_•
.-.
Right-Turn-from-Left Conflict 33.3
>
Cross-Traffic-from-Left Secondary Conflict 23.8
0 I I I
0700 0900 1100 1300 1500 1700 Croas-Traffic-from-Right Secondary Conflict 15.9
Time of Day
Opposing-Left-Turn Secondary Conflict 13.5
Figure 3. Severe conflicts by time of day.
Right-Turn-from-Right Secondary Conflict 11.6
effect, a fourth leg. Traffic into and out of the driveway Left-Turn-from-Right Secondary Conflict 11.2
caused many conflicts. Other conflict types and traffic
Right-Turn-on-Red Conflict
events noted occasionally included pedestrian conflicts, red 9.4
light violations, and U-turns. Left -Turn- from-Le ft Secondary Conflict 8.3

Reliability Lane-Change Conflict 6.4

Reliability is the degree to which different observers re- Right-Turn-from-Left Opportunity 4.1
cord identical results when observing the same traffic Right-Turn-on-Red Opportunity 3.0
events. It is quantified by the interobserver variance, 0-02.
The conflict counts from the field experiments, in gen-
None observed
eral, tend to be small quantities with large variances. That
is, they come from a very skewed distribution such as the
negative binominal rather than a normal distribution. As 109 percent with nearly all of them under 50 percent. The
such, the coefficients of variation (CV—standard deviation worst was right-turn-on-red opportunities, whose high CV
indicates lack of uniform understanding among the observ-
divided by the mean) are typically greater than 1. One
consequence of this property of conflict counts is that the ers. The three paired-vehicle conflict categories previously
data must undergo a mathematical transformation to en- mentioned also had high CV's, as did the slow-vehicle,
able valid statistical testing. This is described in Part 4 of paired-vehicle conflict. These findings indicate observer
Appendix G. difficulties with the paired-vehicle concept. This is clearly
illustrated by Table 3. Particularly for the left- and right-
The interobserver variances, 0-02, were calculated sepa-
turn categories, the over-all reliability is very good, but it is
rately for each of the first two 3-week phases, and com-
much poorer (high CV) when subdivided into paired-
pared (see Table G-9). For all practical purposes, they did
vehicle and not-paired categories. Table 3 shows a similar
not differ between phases, meaning that no noticeable dif-
tendency for the slow-vehicle categories but, here, even the
ferential change between observers occurred as a result of
total reliability is not good. Clearly, the observers were not
"long-term" learning or practice effects—the training pro-
as uniform as desirable in separating driver responses to
gram had effectively completed this process. The lack of
slow vehicles from, say, responses to traffic controls or,
differences also means it is sufficient to examine just one of
perhaps, secondary conflicts.
the phases, in detail.
All of the foregoing reliability findings are based on data
The details of the interobserver variances are given in
collected by the 12 "regular" observers. Analysis of data
Table G-7 and the coefficients of variation in Table G-8,
obtained by the extra observers showed generally com-
for all the observed and derived conflict categories except
parable findings, with just a few exceptions. Some of the
those dismissed in Table 2. In general, 0-02 represents only
paired-vehicle and secondary conflict concepts were ap-
a small part of the total variation in conflict counts (typi-
parently less well understood by a few of the extra observ-
cally, a few percent); other factors appear to be more im-
ers, because higher variances (lower reliabilities) were
portant. A few exceptions are notable. The following had
obtained.
poor reliability as indicated by comparatively large 0- 2
(over 10 percent of the total variance): left-turn, same-
Repeatability
direction, paired-vehicle conflict; right-turn, same direction
conflict; and all rear-end, paired-vehicle conflicts. Several The ability of an observer to achieve uniformity in the
other rear-end conflict types had reliabilities nearly as poor, number of conflicts counted repetitively at a given site
as did some cross-traffic opportunities. under "identical" conditions is called the repeatability.
The coefficients of variation ranged from 9 percent to Conceptually, it could be measured in the field by staging
 11

sequences of traffic events to occur repeatedly. A more TABLE 3

practical approach might be to video tape or film such PAIRED-VEHICLE RELIABILITIES
events and review them repetitively in the office or labora-
tory. However, this procedure lacks realism and may not Coefficient of Variation
lead to results translatable into field practice. Conflict Category (00/mean, percent)
From a practical viewpoint, the observer should be asked
Left-Turn, Same Direction
to view real traffic many times under conditions as nearly
alike as possible. This is effectively what was done. The Paired Vehicle 67.63
factors that might introduce variability into conflict counts, Not Paired 35.87
such as time of day and day of week, were identified and Total 21.19
accounted for, as described previously. What remained
(the residual variance, cr 2 ) was due to two effects: Right-Turn, Same Direction

The true or theoretical repeatability that might be ob- Paired Vehicle 42.96
tained by a hypothetical experiment as previously de- Not Paired 101.82
Total 19.50
scribed.
The inherent variability in real conflicts as traffic
Slow-Vehicle
events, totally analogous to the well-known variance ob- Paired Vehicle 54.25
served in repeated traffic counts. Not Paired 34.16
Total 41.20
The combination of these two effects is the practical re-
peatability—the result that can be expected in real world,
Observation Periods Required
repeated counts.
Repeatabilities were found to improve somewhat (smaller The repeatability, 02, affects the amount of data collec-
2 ) in the second phase, suggesting that as a group the tion required to obtain a given precision. If one wants to
observers became more repeatable with additional expe- estimate, say, the mean number of hourly traffic conflicts at
rience. Also, mean conflict counts tended to decrease some- an intersection within a range of ±p percent and with con-
what, especially for the same-direction conflict categories. fidence 1 - c, the number of hours required is:
Results are detailed in Table G-9.
n = (100 t/p) 2 0 e2 /Y 2 . ( 2)
The residual variances, in general, were quite large and
represented the major contributors to the total variances in Here, Y is the hourly mean value and 02 is the hourly
conflict counts—typically, 50 to 90 percent or more. This variance (each is four times the values given in Appen-
probably means that the inherent variability in conflict dix 0, which are for 15-min counts); t is the statistic from
event rates is quite large. It is not conceivable that trained the normal distribution defined by c. For example, t =
observers count so erratically. 2.58, 1.96, 1.65, and 1.28 for a= 0.01, 0.05, 0.10, and
This finding can be put in better perspective by compar- 0.20, respectively (for large n).
ing the ratios, 0.2I, for various traffic events. For acci- Applications of this principle are given in Table 4. For
dents, which most believe to be distributed approximately same-direction conflicts, the requirements can be met in
P, 0-2/14 1. For the 15-min conflict counts obtained in about a day of observation, assuming the observer is ac-
this project, is in the range of 1.5 to 3.5, depending tively counting conflicts about half the time. For opposing
on the type examined (rear-end, opposing left turn, etc.). left-turn and summary cross-traffic categories, about one
For conflict opportunities the results indicated a range of week would be required, whereas nearly two weeks are
3 to 16 or more for the various types. Finally, analysis of needed for the individual cross-traffic categories for the
scattergrams and the like of traffic volume counts presented conditions stated (±50 percent with c= 0.10). (Using
in the Highway Capacity Manual (10) and the Traffic four times as much data would double all the precision
Engineering Handbook (11) yields values from 9 to.90 for (±25 percent rather than ±50 percent) according to the
O/tL. formula.) However, as described next, some categories
Coefficients of variation of the repeatability measure for (especially cross-traffic and opposing left-turn) are very
15-min counts, given in Table G-8, ranged from 73 to site-dependent; less observation would be required at sites
685 percent. The outstandingly bad conflict category, from with higher than average counts. -
a repeatability viewpoint, is the right-turn-on-red oppor-
tunity. All cross-traffic conflict types and opposing left- Site Characteristics
turn conflicts had CV's of more than 200 percent for Analyses of the basic parameters characterizing the inter-
15-min counts. sections in the field tests and their relationships to conflict
CV's for repeatability decrease as the observation period counts are described in Part 9 of Appendix G, with the
increases, according to V. This is, using a 1-hr count results given in Table G-10. The extra sites are studied
instead of a 15-min count would reduce the CV by half; explicitly in Part 11 of Appendix G.
and using 4-hr data sets would yield Cv's. only one-fourth No major differences between the first two phases of the
as large. Thus, the precision of an estimated mean count experiment were noted. The findings concerning the fac-
increases as longer count periods are used. tors are summarized in the following.
 12

TABLE 4 geometric configurations and traffic controls that, them-

ILLUSTRATIVE OBSERVATION REQUIREMENTS selevs, usually reflected the traffic volumes. For example,
Mean Hourly Hours of with higher volumes there tend to be more lanes and more
- Conflict Category Count Observation- sophisticated traffic control which, in turn, tend to reduce
7.16 4.6
traffic conflicts.
Left-Turn, Same Direction
However, different results are obtained if one examines
Right-Turn, Same Direction 4.89 5.1 on a pair basis the sites with the "same" characteristics-
Slow Vehicle 3.21 5.9 the so-called replicates in the experimental design. Here,
a strong correspondence is found between traffic conflicts
Opposing Left Turn 0.77 21.6
and traffic volumes. In calculating conflict rates, various
Right-Turn from Right 0.71 23.9 normalizing volumes were examined, such as total inter-
section volume, mainline volume, cross-traffic volume, and
Cross Traffic from Right 0.31 39.3
left-turning volume. The best agreement was achieved with
Left Turn from Right 0.59 24.5 mainline volume, in general.
Left Turn from Left 0.78 18.1
Analyses of variance were then conducted of various
average conflict count rates (using mainline volume) to de-
Cross Traffic from Left 0.39 30.0
termine significant site characteristics. Results are sum-
All Same Direction 15.48 3.4 marized in Tables 5 through 9. In each of these tables the
average conflict rates are given, first, for the 12 types of
All Cross Traffic from Left 0.82 20.0
sites. Then the AOV results are presented, including the
All Cross Traffic from Right 1.45 14.8 standard error pertaining to the set of averages.
a! Results based on data from second 3-week block of data collection, Table 5 indicates that typical cross-traffic conflict rates
which exhibited slightly lower, residual variances. can range from 0.18 to 4.43 per 1,000 mainline vehicles,
b/ Hours of data required to estimate mean hourly count within ± 50%
depending on the type of site. The only significant factor,
with 90% confidence. however, is the presence or absence of signalization. Other
things being equal, signalized intersections should experi-
1. Speed-No effect of speed limit on cross-traffic or ence only about one-tenth as many cross-traffic conflicts as
opposing left-turn conflicts; tendency, for more rear-end unsignalized intersections.
conflicts (except to turn right) on high-speed routes; tend- 'Same-direction conflict rates are much higher, as can be
ency for more conflict opportunities on high-speed routes. seen from Table 6. The most significant difference between
opportunities and more sites is due to the number of lanes on the mainline ap-
conflicts t 3-way intersections of nearly all types (where proach; 2-lane roads experience nearly 3 times as many as
movement is prmitted by geometrics of the intersection). 4-lane roads. It is also noteworthy that fewer same-
rear-end direction conflicts are observed at 3-way intersections than
conflicts of all types for signalized intersections, except in at 4-way intersections, other things being equal. The inter-
conjunction with right turns; more opposing left-turn con- action between intersection type and speed arises because,
flicts and opportunities at signalized intersections; fewer on 2-lane roads, speed is an important factor (more con-
cross-traffic conflicts and opportunities at signalized inter- flicts on high-speed roads); there is no speed relationship
sections. on 4-lane roads.
4. Two-lane vs. 4-lane, unsignalized intersections-More The conflicts related to left-turn movements are analyzed
rear-end conflicts of all types at 2-lane intersections; fewer in Table 7. Here, as with same-direction conflicts, 2-lane
cross-traffic conflicts at 2-lane intersections; no highly sig- roads have significantly higher conflict rates. Also, a simi-
nificant differences in opposing left-turn conflicts or in any lar interaction exists. Moreover, there is a marginally sig-
types of conflict opportunities.
nificant speed effect (more conflicts at high-speed inter-
on sections) and a marginally significant effect of the number
extra site data-No significant differences of any kind.
of legs (more conflicts at 4-way intersections). It is also
Examination of the individual sites and corresponding noteworthy, although not statistically significant, that so
conflict counts are presented in Part 10 of Appendix G; few conflicts per 1,000 vehicles are observed at low-speed,
further analysis of the four extra sites is in Part 11. 3-way, 4-lane, unsignalized intersections.
Table 8 indicates only that there are significantly fewer
Traffic Volume Effects
total conflict opportunities at signalized intersections than
The effects of traffic volumes on conflicts and conflict at others, as expected. The lack of other significant findings
rates are discussed in Part 10 of Appendix G, together with is due, in part, to the very large standard error, S, which
the analyses summarized in the following. is about half of the over-all average of 165.2 conflict
On an over-all basis, averaging across all sites, little cor- opportunities per 1,000 mainline vehicles.
relation could be found between traffic volumes or direc- Table 9 shows generally the same sort of results as given
tional movements and various conflict categories. Thus, in Table 6. (Table 6 deals with same-direction conflicts and
generally speaking, conflict counts were relatively indepen- Table 9 deals with total conflicts, most of which are same-
dent of traffic volumes. However, this result should not be direction conflicts.) The major effect is that 2-lane ap-
unexpected because the set of sites covered a variety of proaches have far more conflicts than 4-lane approaches.
 13

'0I
01
II a
44 04000400 0'
3 43 P.fl'3'.3U•) a' 0
0
fl'0Ui
0 (0
en
'4 _00
V
"<00 (3
0i"Ui 11
04747, V VVV
o&n.oa''.o
011 Ui<
01 Ui 74 0' Ui
a1 iOPi O I.I en a' en en '0 7'. .3 ((0
'"1 0100'3 0 '0 (N Ui en ('4
U4 0(01
74 43 eN 0 0' .3 00 iN z (0740
OI1 01 0'<3'0Oa'" 0 'UI I
a'en".
iii "4 C'4 40 00 0 00 '0 .3 C-) 01 7
44
'4401
iNOiNeninO,
Ui 0 .7 PI .3 0 0
(0
11 0
44 6
0
07400'0'0
'
00)
II
01
OWl ."P4C(IPIUiP'WUi
El) '0J Li 6(0 Pl0'0'3Ui'0N'
.0
0J
04 .31
incien a' CL1 .3
- 6 0 06
00
a'
0
140 7104C.iC'17'iC'1
LU 00
00
oOIi - 0 '1 11 7404047171047171
0 1-.
z 000). 04
0
00
'0 '0
0000 '0 '0
.-.0000 00
vv 0NON
0 0000
.".0000
(0,43 .344444444
'30
0443.3 .-.t.-1I0(0CU01'
1.1.1.1.0 7.
0000
Ii
0.0.00001.
>,0.04J444J4J4U
0 c1,z 00000
'.40,4 110 00 1.1.1.1.0
0.0.00001. '3
0,40
0 Ei00Z00000 '3
00
0.30 0.0.0LiLiuLiui pi
'0
'0 t'Li'Li,4007)' I
00 0043430)) 0 LU
(00(0
(U
E.00ZEiE4430U0 0) - 000 (0 (fl1-iO
1-4 044314
430)
(00
< Zc1 -3.-i-i
E <0

40.374 -

0 a
z
++a
.4 -'e'en 014
U. eNV VVVV '00a' Li
.3(071 'F- (4 0eneno---
0) Ui 0
Pl7"0
1. V
01 0 00 '0 en .3 .3 en '3
--
'0_Seq
U.I
0 WI '3 0471 r"'(0r" 74 LU-)
'3(01 40.3-"3e'1O'3
5-i00a'.3."Ui0N'
d-dddd- z O01.11
07!101 '0 en r.. en a' Ui 0 a'
00 43 1en'0e-Uir'.UiC9'3
ZO
-la I '0 (01 (N 0P.'0•'. 7'
007 74074 0411 W074'3Ui<'30' 0' '.7 00 a' '0 iN Ui 0 P4
00
0 l 7
0.0 0 IN 0 Ii
44 (0 0 Ui 0 - 0 0 0 0 en
00
(0 10401
0 enr-.cN 0
740000 04 001 C4Ui00 04
a' 06 06
'torn 0
a'
0
00
00
0 04 7474' 74 74 Ui
00
00
I 000.
0 C
0

0
0
N0N
0 ".00
0000
0
1.
144 0
0000
.._0000
004.1444444 '0 0
0,4.4.4.4 000
vVv
....0UUUU.'. U V
0.
IiO 0.0 .00001. '.7' 50)?
10000
0 0.0.000044
0.0.001.0441.
0000 44
'4) E(04Z00000 Li a Li
0.0.0441.444441
E4(OZ0000 0 (041+
43 a'
'3
, 0
14. 43
004
43 "I 04434310
E404431-lI40)I40 ('( 01

Day-of-Week and Time-of-Day Effects Time-of-day effects for severe conflicts were described
earlier (Fig. 3). More details are given in Appendix G,
Over-all, there were no clear-cut, uniform differences in
conflict counts by day of week-although Mondays may Figure G-3 and Table G15. There, it is shown that both
have experienced a few more conflicts of some types than morning and afternoon peaks exist for volumes and op-
did the other weekdays, and Fridays may have experienced portunities, as compared with severe conflicts which ex-.
more conflict opportunities of some types. Details are hibit only an afternoon peak. Other traffic conflicts tend to
provided in Table G-10. have both morning and afternoon peaks, but with the p.m.
14

TABLE 9
ANALYSIS OF VARIANCE OF TOTAL CONFLICTS

Site Averages, per 1,000 Main-line Vehicles

High Speed Low Speed
Intersection Type. 4-Way 3-Way 4-Way 3-Way

4-Lane, Vosignalized 15.97 13.45 18.04 13.92

1171i367. 12.11 11.55
.-.---------------
2-Lane, Unsignalized 48.50 31.87 29.60 27.71

AOV Results

Degrees of Sum of Mean

Factor Freedom Squares Square F-ratioJ

.T (Type) 2 2,304.8943 1,152.4472 37.61**

S (Speed) 1 82.3251 82.3251 2.69
N (No. Legs) 1 94.2084 94.2084 3.07
TS Interaction 2 188.1719 94.0659 3.07
TN Interaction 2 100.3903 50.1951 1.64
SN Interaction 1 18.7798 18.7798 <1
TSN Interaction 2 94.2355 47,1177 1.54
e (error) 367.6563 30.6380

- 5.45

a/ ** p <0.05

peak being far more pronounced. The observation of the heterogeneity of such things as total accidents and total
higher conflict rates in the afternoon is in agreement with conflicts.
general accident experience, implying that driving habits, When accidents of certain types were compared to con-
on the average, deteriorate late in the day. flicts of analogous types, much better relationships were
obtained (see Figs. 0-4 through 0-7). Opposing left-lurn
Accident Relationships accidents and cross-traffic accidents, particularly, yielded
Limited accident data for the intersections used in this good (significant) correlations with analogous conflicts, with
study provided some insight into conflict/accident relation- correlation coefficients of 0.619 and 0.667, respectively.
ships, discussed in Part 14 of Appendix G. Comparisons between accidents and analogous conflict
Over-all, correlation coefficients between total accidents opportunities were mostly unproductive. Most correlation
over a 3-year time period and several categories of con- coefficients were essentially zero (and most were negative).
flicts and opportunities at the experimental sites were rela- The exception was rear-end accidents, which had a high cor-
tively meaningless. Total traffic volumes correlated as well relation coefficient with mainline volumes (0.971), based
as anything. But, this should not be surprising because of on very limited data (see Fig. 0-8).

CHAPTER THREE

INTERPRETATION, APPRAISAL, APPLICATION

USES OF TRAFFIC. CONFLICTS verse accident history. It is not appropriate, however, for
identifying hazardous intersections because of the cost per
Applying the traffic conflicts technique is somewhat time- intersection required for its application. However, traffic
consuming, so it should not be used indiscriminantly. conflicts are well suited to confirming (or denying) sug-
Rather, the TCT should be applied only for one of several gestions that a specific site has an accident problem or has
well-defined reasons. inherent safety problems not yet illuminated by an exten-
The TCT is an excellent tool for diagnosing safety/ sive accident history. Typical sources of such suggestions
operational problems of intersections that have previously are citizen complaints, a prominent serious or fatal acci-
been singled out for attention, usually because of an ad- dent, or a short-term "rash" of accidents at a particular
 15

intersection. Also, in this category would be sites that tem- unusual traffic events or problems, and any observer-
porarily or recently have been modified, so that accident suggested remedial measures. Total traffic volume counts,
histories are not suitable (e.g., a construction site). turning movements, or classification counts may also be
Traffic conflicts are also applicable to before/after evalua- desirable information for the traffic engineer. However,
tions of intersection improvements, both on a site-specific such data would require extra observers or a modified
basis as well as to gather "research" data on counter- (alternating) observation schedule for the conflict observers.
measure effectiveness. One must be cautious, however, to It would be most efficient to observe conflicts when traffic
ensure that changes in conflict counts are causally related in is heavy, but not during congested, stop-and-go traffic.
a logical fashion to the type of improvement implemented. Typical periods would be the morning and evening peaks
and, in some locations, a midmorning peak. However, con-
CONFLICT CATEGORIES flicts can be observed at any time of the day, and adjusted
by means of mainline volume counts. Thus, evening and
The traffic conflict categories to be observed and re- weekend counting may be desirable if unusual problems are
corded in the field should be reliable, repeatable, and prac- suspected at such times, and if there are no employment
tical; and, at least, have face validity if not a strong acci- constraints on taking conflict counts at such times.
dent correlation. Following these guidelines, the conflict
categories that should be used are: right turn, same direc- APPLICATION OF CONFLICTS RESULTS
tion; left turn, same direction; slow vehicle, same direction;
The types and numbers of conflict counts encountered
opposing left turn; right turn from right; cross traffic from
depend on the characteristics of the intersections. The
right; left turn from right; cross traffic from left; and left
types of conflicts for various intersections (including those
turn from left. For each of these, secondary conflicts
recommended for general as well as special use) are shown
should also be observed and recorded. Also, simultane-
in Figure 4.
ously with the traffic conflicts, observers should count and
The numbers of conflict counts to be expected, or that
record the traffic volume on the leg being observed moving
are indicative of safety or operational problems, cannot be
toward the intersection.
stated unequivocally, at present. Sufficient research on this
Preliminary observations on a site, accident records, or
topic has not yet been accomplished. However, several
citizen complaints may indicate that other, more specialized
points are apparent. First, the counts, themselves, are not
categories might also be noted in certain instances. These
useful comparative indiëators. Even the limited number of
special categories include: right turn from left; right turn
intersections used in this project illustrated the extreme
on red; lane change; pedestrian; and pedalcycle.
variations in counts between nominally similar intersec-
The observers should always note any special occur-
tions. Counts should be normalized (divided) by traffic
rences, particularly the apparent causes of slow-vehicle
volumes, yielding conflict rates. The mainline volumes ap-
conflicts.
pear to be most appropriate for this purpose rather than
In analyzing conflict counts, certain categories should be
total volumes, cross-traffic volumes, etc. Tables 5 through
combined to obtain more strongly based figures. First, each
9 may provide some guidance as to average conflict rates
secondary conflict category should be summed with its
and standard errors for various types of intersections.
causative conflict category (e.g., combine right-turn, same-
High traffic conflict rates of specific types may suggest
direction secondary conflict with right-turn, same-direction
certain needs for intersection improvements. Improve-
conflicts). Then, the following sums should be created:
ments that might be considered are given in Tables 10
all same-direction conflicts; opposing left-turn and left-turn
through 12 for three intersection configurations. The un-
same-direction conflicts; all conflicts involving vehicles
signalized intersections appear to offer fewer options for
from the right; all conflicts involving vehicles from the left;
improvement, with the exception of the very broad option
and all conflicts involving cross traffic.
of adding signalization.
To evaluate intersection improvements, conflict counts
CONFLICT OBSERVATIONS
may be used in before/after comparisons, provided no
Traffic conflicts can be observed at an intersection by major changes in traffic volumes have occurred. The counts
either one or two persons. In either case, individuals should may be compared using standard statistical tests, such as
observe opposite legs of the intersection alternately. Traffic t-tests, provided transformations are first applied as dis-
counting boards should be used for the higher counts, but cussed in Part 4 of Appendix G.
less frequent events can be recorded manually.
TRAINING AND IMPLEMENTATION
A basic work segment of 30 min is recommended, based
on the experiences of the researchers of this study as well An agency intending to use the TCT should be aware
as those of several U.S. and foreign agencies. In each seg- that properly trained and experienced observers are neces-
ment, traffic conflicts should be observed for 20 mm. The sary for success. Otherwise, only inaccurate and unreliable
other 10 miii should be used for recording the counts and data can be expected. Available options are: (1) contract
other data, and moving to the opposing leg. such work with qualified consultants; or (2) train and
Other data should also be collected at the site either prior maintain traffic technicians inhouse. The latter may be
to or after a several-hour period of conflict observations. most cost effective if TCT use will be widespread; the
These data include a sketch of the intersection, photographs former may be more appropriate for occasional needs or
from each leg, an intersection inventory, comments on any unusual applications (e.g., nights or weekends).
16

Conflict Type
Right-Turn Left-Turn Cross
intersection lype - Rear-End - Crossing Crossing Traffic

0.
u 0)
C .
4, -c C C 0 O)C - • - 0 I-

. I.
'I . . -E i
0>
4,
c o.-000
3. 3 a...!.
a. E E E E E 4, 0 E
2 2 2 2 2
u2
i-
- vs 3 < vs vs vs vs vs 3 0- u- u. u LL Q. W Z 0
v v v v . * 6

4 Yes 2 All • • 0 0 • • V V V V • 6

4 No 4 All • • • 0 . • 0 . . 10

4 No 2 All • • • • • • • • • • • 10

3W Yes 4 W V 5

3 Yes 4 E • • S V . * 3

3 Yes 4 S V V V • 1

3 Yes 2 W • • 0 0 V • 4

3 Yes 2 .E • 0 0 V 0 3

3 Yes 2. S 0 V V V • 1

3 No 4W • • 0 0 0 0 6

3 No 4 E • • 0 • 4

3 No 4 S • . . . 4

3 No 2 W • • 5

3 No 2 E • 4

3 No 2 S • . • S 4

/ Orientation: W E • Applicable in general

1S * Applicable only with protected left turn
/ Excluding pedestrian, signal violations, V: Possible only with signal violation
and RTOR

Figure 4. Applicable conflicts for typical intersections.

Training concepts are discussed in Appendixes H and I. Finally, an agency that invests the training effort in its
There is no substitute for field practice and experience that team of observers should keep them in practice. Although
will accustom the trainees to the variety of real world hap- they certainly need not do traffic conflict counts constantly,
penings and help them develop a consistency in interpreta- they should do them a few days each month, as a mini-
tion. It is strongly recommended that several persons be mum. After any substantial layoff (e.g., wintertime), some
trained together, and comparative evaluations of their prog- retraining and practice may be worthwhile. To make the
ress be made. Although it is not inherently difficult to com- best use of its trained observers, the agency (say, a state
prehend and apply the operational definitions, some per- transportation department) should make them available to
sons will simply not view traffic dvents the same way as its districts as a service on an as-needed or as-requested
most, probably because of deep-s91ated ideas or behaviors. basis.
Others will soon become bored with the repetitive nature
of the task and become unsuitathe.
 17

TABLE 10
IMPROVEMENTS FOR SIGNALIZED, 4—LEG, 4—LANE INTERSECTIONS
Same
Direction Right Left
(Rear-End) o Turn .Turn Thru
1.
a Cross Cross Cross
I—
Traffic Traffic Traffic 00
0 U 0-
a aJ
I
00501.
S a
_ U

ed
0
sO ..,
U

sO
0.
E.
50
50
5 a 50 40 50 C
00 00 _d 00 ed .50
0.0)515 o a e a a a a a 0
0.
050000 0. 0 C 0 0 0 0 0
O50'l 50 Id. Id. A. 0
..I 00 55 3 E 0 5 5 0. Id.
Improvement

Left-Turn Bay X X X
Left-Turn Phase X x dx

Left-Turn Restriction N N X
Right-Turn Bay X X
Right-Turn Radius or N X X X
Roadway
Signal Cycle or Phase X X XX X X
Length
Actuated Signals X X x N X
Longer Amber or all Red N X X X X X
Clearance
RTOR Restrictions X X X
Pedestrian Barriers X
Pedestrian Phase X
Add Lanes XXX N X
Parking Restrictions NX
Install Median X
Improve Corner Sight X X
Distance
SpeedZone XXX X X
Advance Warning or XNX XN X
Sight Distance Control
Advance Street Name Sign XXX XN
Enforcement x x x x x

TABLE 11
IMPROVEMENTS FOR UNSIGNALIZED, 4—LEG
2—LANE INTERSECTIONS

Direction
Right Left
(Rear-End)
1. Turn Turn Thru
Cross Cross Cross
Traffic Traffic Traffic
a
I 0
00001. .-J U U U 0

O 5-. a .t 00 50
HIU ,5 .1 00 _1 00 .5 II

U.C)5155 0 a E 10 10 10 a
eds000U 0. 0 0 0 0 0 0 0
Improvement -' 00 510 .-5 H 0 Id. Id. P. SI Sd 0.

Add Signal XX X X X N X X
Left-Turn Bay N X
Right-Turn Bay X
Right-Turn Radius or X X X
Roadway
Pedestrian Barriers X
Add Lanes XXX X X
Parking Restrictions NX x
Improve Corner Sight X X X X N X
Distance
Speed Zone xN X X X X X X X
Advance Warning or XXX X N
Sight Distance to
Traffic Control
Advance Street Name Sign NXX X
Enforcement X X X X X
18

TABLE 12
IMPROVEMENT FOR UNSIGNALIZEDEG
2-LANE INTERSECTIONS
of
Same Direction Right Left
(Rear-End) Turn Turn
Cross Cross
0
O
C Traffic Traffic
CUCOU -J U =

00 00 U

.0)05 C C 'I
t000_J 0 0 0
lmr,rovement

Add Signal X X X X
Right-Turn Bay X
Right Turn Radius or X X
Roadway
Pedestrian Barriers X
AddLanes XX X -
Parking Restrictions X X
Improve Corner Sight X X X
Distance
SpeedZone XX X X X
Advance Warning or XX I X
Sight Distance to
Traffic Control
Advance Street Name Sign IX I
Enforcement I K
RTOR Restrictions X

a! Viewed from leg allowing thru and right-turn movements.

CHAPTER FOUR

CONCLUSIONS AND SUGGESTED RESEARCH

CONCLUSIONS Rear-end accidents at intersections seem to be more

strongly associated with mainline traffic volumes than with
The use of the traffic conflicts technique (TCT) at
rear-end conflicts, although observations of the latter may
intersections is most suitable for diagnosis, improvement
help to discover the reasons for rear-end accidents.
evaluation, and confirmation or denial of the presence of
The identification of "severe" conflicts, as distin--
safety hazards or operational problems at suspect locations.
guished from others, may be of general interest, but they
It is not recommended for routine hazardous location iden-
occur too infrequently to be of use as diagnostic or evalua-
tification because of the large amounts of data collection
tive measures.
that would be required.
The amount of data collection needed to obtain rea-
Traffic conflicts data should be viewed as supplements
sonably precise conflict-rate estimates depends on the type
to, not replacements of, accident data.
of conflict and the type of intersection, but is typically on
The recommended traffic conflicts data can be ob-
the order of a few hours to a few days.
tained reliably by traffic technicians with moderate training,
Traffic conflicts and traffic conflict rates (especially
a minimum of special abilities, and no equipment other
severe conflicts) increase substantially throughout the mid-
than a mechanical count board and watch.
afternoon and late afternoon.
Traffic conflicts, as stochastic traffic events, vary quite
The training of persons in the traffic conflicts tech-
markedly in number and rate from day to day, even under
nique should rely heavily on supervised and/or critically
nominally identical conditions, just as do other traffic events
reviewed field practice.
such as accidents and turning volumes. Thus, they are not
as repeatable as would be desirable.
RESEARCH NEEDS
Cross-traffic and opposing left-turn accidents are
usually the most prevalent and serious safety problems at Although the activities of this project led naturally to
intersections. The TCT is particularly useful for these many recommendations, most of these referred directly to
problems. ways and means of implementing the observation, record-
 19

ing, and analysis of traffic conflicts. Methodologies, defini- categories of traffic conflicts, dependent on site characteris-
tions, and the like, that are operationally feasible have been tics. This research provided a starting point, and suggested
determined. No further research along these lines is rec- that conflict rates rather than counts are most appropriate.
ommended. However, two needs that relate to the applica- What is needed is a much more extensive and diverse set
tion of the TCT at intersections are apparent. of intersections, at which conflict counts would be con-
One of these needs is to quantify the relationship between ducted and better estimates made of means and variances.
traffic conflicts of certain types and accidents of analogous In short, useful guidelines should be established for indi-
types. This project yielded suggestions concerning opposing vidual intersection types as to the expected and abnormal
left-turn and cross-traffic conflicts and the analogous acci-
conflict rates.
dents, and similar information about same-direction acci-
The present research was limited to intersection conflicts.
dents and traffic volumes at intersections. Much more work
applying the TCT at a diversity of intersections, combined However, the literature contains many examples of other
with accident data available from good accident records areas of application, including midblock locations, freeway
systems, is required. entrances and exits, weaving areas, construction zones, and
The other need, which is equal if not greater in impor- pedestrian crossings. Futher research is needed to clarify
tance, is to establish norms and "warrants" for various and standardize procedures to be used in such situations.

REFERENCES

Proc., First Workshop on Traffic Conflicts, Oslo "Highway Capacity Manual-19651" HRB Special
(1977) 138 pp. Report No. 87 (1965).
PERKINS, S. R., and HARRIS, J. I., "Traffic Conflict Institute of Transportation Engineers, Transportation
Characteristics—Accident Potential at Intersections." and Traffic Engineering Handbook. Prentice-Hall, Inc.
Highway Research Record No. 225 (1968) pp. 35-43. (1976).
PERKINS, S. R., "GMR Traffic Conflicts Technique PARKER, M. R., JR., ET AL., "Right-Turn-On-Red."
Procedures Manual." General Motors Research Pub- Virginia Highway and Transportation Research Coun-
lication, GMR-895 (Aug. 11, 1969). cil, Report VHTRC 76-R9 (Sept. 1975) 57 pp.
BAKER, W. T., "An Evaluation of the Traffic Conflicts PARKER, M. R., JR., ET AL., "An Evaluation of the
Technique." Highway Research Record No. 384 Movement of 14-Foot Wide Manufactured Housing
(1972) pp. 1-8. Units in Virginia." Virginia Highway and Transpor-
PUGH, D. E., "Traffic Conflicts in Washington State." tation Research Council, Report VHTRC 77-R28
Washington State Department of Highways (Apr. 18, (Nov. 1976).
1975). HAUER, E., "Indirect Measurement of Safety—The
PADDOCK, R. D., "The Conflict Technique Procedures Conflict Method." University of Toronto (Aug. 1977)
Manual." Ohio Department of Transportation (1975). 41 pp. (unpublished).
ALLEN, B. L., and SHIN, B. T., "A New Look at the HAYWARD, J. C., "Near-Miss Determination Through
Conceptual and Empirical Aspects of the Traffic Con- Use of a Scale of Danger." Highway Research Record
flicts Technique." McMaster Univ., Hamilton, On- No. 384 (1972) pp. 24-34.
tario. Can. (Feb. 1977) 104 pp. HAUER, E., "Design Considerations of Traffic Con-
GLENNON, J. C., ET AL., "Critique of the Traffic Con- flicts Surveys." Transportation Research Record No.
flicts Technique." Transportation Research Record 667 (1978) pp. 57-66.
No. 630 (1977) pp. 32-38. BEALL, G., "The Transformation from Entomological
HYDIU'J, C., "A Traffic Conflicts Technique for De- Field Experiments So That the Analysis of Variance
termining Risk." University of Lund (1977) 48 pp. Becomes Applicable." Biometrika, 32 (1942) p. 243.
20

APPENDIX A

LITERATURE, REVIEW AND BIBLIOGRAPHY

LITERATURE REVIEW Each category is discussed in the following with respect to

the type of information contained in it.
During Task 1 of this project, more than 60 documents
were obtained, reviewed, and summarized. The bibliogra-
Research Applications of TCT
phy in this appendix contains an international list of traffic
conflicts reports and discussions. The majority of researchers using TCT were concerned
In 1967, Perkins and Harris, of the General Motors Cor- with developing or refining procedures and techniques for
poration, published the first paper on traffic conflicts research purposes, although some agencies used the TCT
(A-39). A traffic conflict, paraphrasing their definition, both experimentally and operationally. There are four
was a potential accident situation in which a driver brakes major areas in which TCT's are used in research: predic-
or swerves to avoid a collision. tion of accidents, development of conflicts definitions, de-
The traffic conflicts technique (TCT) was originally de- velopment of TCT procedures, and simulation.
veloped to evaluate accident potential and operational de- It is generally agreed that conflicts and accidents are re-
ficiencies of intersections without depending on accident lated somehow. However, statistical relationships between
data for these analyses. It was believed that a direct rela- the two have not been good, even though statistically sig-
tionship existed between accidents and conflicts, and that nificant in some cases. This significance has been attributed
any physical or operational improvements to intersections to traffic volume dependence by some researchers, and
that reduced conflicts, would be substantiated by a result- some studies have shown that traffic volumes are actually
ing reduction in accidents. Much of the early conflicts re- better than conflicts at predicting accidents (A-li, A-33).
search was concerned with finding an exact relationship Not much research has been done to specifically define
between accidents and conflicts. Some researchers were conflicts. There are two basic versions, with many modifi-
discouraged by the lack of correlation between accidents cations, of conflicts definition: the General Motors (A-39)
and conflicts, and believed the TCT could not be used as and the time measured to collision (TMTC) (A-20). They
an operational tool because of this. contain similarities, but differ in the degree of sophistica-
Today, there are two basic schools of thought on the tion with which they are measured.
TCT. First, there are those researchers trying to correlate The procedures used to record traffic conflicts vary con-
accidents and conflicts. They have redefined conflicts and siderably and are discussed under "Recording Techniques."
the techniques used to measure them to improve the corre- Research in conflicts simulation, a formative area, has
lation. There has been some success; but, for the most part, not yet been very successful.
conflicts remain imperfect predictors of accidents, in large
part because accidents, themselves, are not good predictors Operational Uses of TCT
of accidents. Other researchers believe that accident pre-
diction, per ce, may not be as important as identifying po- Operationally, the TCT has enjoyed more success. Types
tential hazards and operational deficiencies. As a result of of operational uses include: before/after improvement
this thinking, the TCT has been used to justify and sub- studies, construction zones, freeway acceleration lanes,
stantiate many types of operational traffic engineering im- freeway lane drops, freeway weave areas, interchanges,
provements at a variety of locations. intersections, oversize loads, pedestrians, signing and traffic
A tabulation of many of the key findings of the litera- signals. In almost all of these areas, new or modified defi-
ture review is summarized in Table A-I. The findings are nitions were derived to fit the needs of the studies.
organized into 12 categories: The biggest difference between research and operational
uses of the TCT is that research studies are usually per-
Research applications of TCT. formed using more sophisticated equipment and techniques,
Operational uses of TCT. whereas operational data are manually recorded by a team
Type of document. of observers. Because many of these latter studies are done
TCT definition. by transportation departments that lack the necessary
Observation procedures. equipment, the manual observation method is preferred
Observers. because of its lower cost and ease of performance.
Training.
Recording techniques.
Type of Document
Analysis methods.
Relationships. Types of documents reviewed include: general discus-
Type of reporting organization. sion, research study, standard/official ruling/guideline and
Nationality. status report. The majority of the reports reviewed were
TABLE A-i
SUBJECT REFERENCE MATRIX
TRAFFIC CONFUCTS TECHNIQUE LITERATURE REVIEW

IIr ...U............D....a..l..DU.
u.
.
11MOMEMOMMUM
OMEMEM i••u••uuuu•umuau•u
MEMEMMIMMEME
00 ___ Dl.lUUNDU•DUUUU
___________DUU•UU••UU•IUUUDmUDNUUU•U___ EMMMMMMMIMM E013 OMEME no
er iuuuu•u•••••auu•••i
UI.. I.1U•UU•UUDUI•UU•.UIUUUDU.DUUaUU uum~mmmmj
UL4 .
LIUU.II.UUI•Da•.U.•D.UUDU.....___

_____________UU•NUUUIUUUUU•••UD•UUD•iUUSU• ......I......u....D..ua
SEEM
_____________lUU••U•UUUUUUDIUD••U•UlDDDU•• D•U•UUIUIUUQD••UUU••
___•••uuuuuuuaa•uu•a•••
'UUU•UU•••••U•IUUUUI•UUI•IIUIUI lill•UUIU•UUIUU•IUIIUID
_____________
___•UUI••I•••I•UU•U•UUDND
___ •UUUUD•Q••I•IINU•UDUD•

DUU••UURU•U•UUUDUIUDSUD•D•U
DuD•uu•Duuu•maDDu••••aD
IMME•••I•UIm•UD•UU•QIIUUU

tIrrI71.
u•uuuaumu•m•u••a•uuai
_________ aauu•ui•uuuiiusau•••
_________ UU•UUiUUUUUD•UDD•••D
o on
____•.U.D•DUU•D•••IUD•U•DUl•
PEMMMI
_____________ ...uui•.••au•uaau••u
_____'D•UUUUUU•UI•l•UUUUUDD• UUUDNUU•UUDDI•DUU•DUQU
 F..)
TABLE A-i (Continued) F..)

TRAFFIC CONFUCTS TECHNIQUE UTERATURE REVIEW

.... T!TI(

_____________!IH!!I1HHiiI!iiiIIP!I!ilIIIUPIiiIII

'•'uD•uuuuI•uuu••u•DlaiIaaaau
•••uiiva•uuuuuu•au•aui•
1IluD.uu...uuD......a..D...aaaD.
___••UUUU•I•D•UIUDU•N•IDM
NUIUU•••••UU•UUD•UD•
_____________uuauuuu•uu••ui•i•a•••uu•••a•a iuuau•u•uanuu••ii•••i•a
"UUINUUUU•URDUUUD•flUU•UUDDfllD END
______________U•UUUUUUU•U••DUUUUUU•U•DDD•••MINE•••U•U•••UUDU•UUUDU••QMD
______________ MINE
____•••U•UD••UUUUU•QUU•UDUD•
. ••••a••••••••••••a••I•••D••••
U••UUDUU•IIUUUUUUUUU••D
. "U•••UUUU•DlUll•U•UD•UIUUDUDQDU •••uiia•uuuauuuu•uu••nø•

_uuuunuu•uuuuuuu•ua••i•••u•u•••Moon uuuuu••iuuu•••uuuuu•au
____UDDU••••••••••UNSUUUUUUDDUUU___
.UD•.•UUUUD.UU•O.U•DDD•
UUUU•U•UUUQU•U•••D•D
_
977
___•UU•UU•U•II•U•UU•U•UD•
MEMO
_______ ___ .....ua...au.ua....DDua
END
 23

research studies. Many of the remainder, which are in- at locations other than intersections. (See the "Operational
cluded in Table A-i, were of the general discussion type Use" of TCT category.)
and were presented at the 1977 International Workshop on
Traffic Conflicts, held in Oslo (A-63). An extensive collec- Observation Procedures
tion of international research abstracts is contained in a
document from the Netherlands (A-67). This category consists of number of observers, location
of observers, time of observation, and days of week.
TCT Definition The GM procedure calls for two observers, located 100
to 300 ft (30 to 91 m) from the intersection, to observe
The basic types of definitions used in the studies include: vehicle movements for 10 hr (0730 to 1200 hr and 1245
General Motors (GM) (A-38, A-39), time measured to to 1815 hr) during Tuesday, Wednesday, or Thursday
collision (TMTC) (A-20), erratic maneuver, and other. (A-39). This procedure usually consists of counting con-
According to the GM definition, which was designed for flicts and traffic volume on two opposite approach legs dur-
use at intersections, "a traffic conflict occurs when a driver ing a day. Two observers count conflicts and volume for
takes evasive action, brakes or weaves to avoid a collision." a 15-min period. The next 15 min are spent recording con-
The observable evasive action consists of brakelight indi- flicts and moving to the opposite approach, where the count
cations and/or swerves or lane changes. Originally, there resumes again. The data collection forms used in the GM
were 24 different types of intersection conflicts in the GM technique included Counters Inventory of Existing High-
procedures. Some researchers believe that recording all way Features, Volume Count Sheet, Conflicts Count Sheet,
brakelight applications may not present a true indication of Analyst's Inventory of Improvements, Accident Data Sheet,
the operation problems of a location. In the United States, and Photographs Sheet.
four transportation departments that are active TCT users Forms and procedures used by U.S. agencies are basi-
have begun to classify conflicts by severity as routine, mod- cally the same as for the GM procedure, but with some
erate, or severe (A-8, A-34, A-40, A-49). A routine con- modifications to suit their needs. In general, one, two, or
flict involves precautionary braking or swerving with a three observers are used to observe conflicts, traffic volume,
small risk of collision. A moderate conflict involves con- or special activities (e.g., delay and compliance and ac-
trolled evasive action, but with a higher potential for ceptance of RTOR (A-37). Location of observers is gen-
collision. Rapid deceleration or a severe swerve with no ele- erally the same; that is, a location permitting a good view
ment of control would indicate a severe conflict. Discus- of the traffic movements of the location. Typically, this is
sions with the States of Ohio and Washington indicate that 200 to 300 ft (61 to 91 m) upstream of the location of
moderate and severe conflicts are not common occurrences, interest, and along the right edge of the traveled way.
however. Times of observation vary. Ohio follows the GM pro-
The other basic method of defining a conflict is illus- cedure, while the State of Washington observes conflicts
trated by the time measured to collision (TCTM), which is during peak periods (usually for 4 to 6 hr) on any day and
"the time required for two vehicles to collide if they con- at any time, including nights and weekends. Kentucky, on
tinue at their present speeds on the same path" (A-20). the other hand, usually counts continuously for 1 to 2 hr
This and related definitions are popular with foreign re- during a peak period (A-66). There is not much informa-
searchers (A-2, A-13, A-22, A-32, A-45). Because this tion regarding conflict counting on nights or weekends—
technique involves more precise measurements, more so- although during one Canadian study, four college students
phisticated equipment is normally used in the determina- observed conflicts from 0700 through 2100 hr on Monday
tion of these conflicts. This equipment includes: time-lapse through Friday (A-Il).
photography, videotape, and automatic measuring devices.
A numerical minimum TMTC would be approximately
Observers
½ sec, which is the driver reaction time (A-20), although
times of 1 sec (A-20) and 11/2 sec (A-22) have been used This category consists of qualifications and personal char-
for analysis purposes. acteristics of conflict observers. It was hoped that this cate-
According to the TMTC definition, a conflict is one step gory could give insights as to the best types of people hired
in a continuum of events that could lead to an accident. If as conflicts observers. As it turned out, very little informa-
two vehicles are not on a collision path, as determined by tion was obtained from the literature.
TMTC measurement techniques, a conflict has not oc- In research studies, college students were typically used
curred. This eliminates all events that drivers thought were to observe conflicts. Operationally, state organizations use
"near-misses," but did not qualify according to the TMTC available technicians who are trained as observers. Gen-
definition. erally, these people have some sort of previous traffic ob-
An erratic maneuver occurs when an unimpeded vehicle servation experience, such as spot speed studies or traffic
brakes or suddenly swerves. Because it is an unimpeded volume counts. Both Ohio and Washington have techni-
veheicle (has no vehicles directly ahead or rapidly over- cians who started conflicts observation with the original
taking it), this implies that there is no other vehicle caus- 1969 FHWA conflicts study (A-5).
ing the maneuver. Erratic maneuvers were included with
conflicts in three studies (A-50, A-51, A-57). Training
There were many types of other conflicts. Most of these
were designed to fit the special needs of studies performed The training category consists of: type, duration, and
24

use of audiovisual aides. None of the research studies mentally and operationally (A-9, A-15, A-22, A-26, A-30,
thoroughly outlines a training program, but some indicated A-50, A-52, A-64).
the type of training given. On-the-job-training appears to Reliability concerns the variation in conflict counts be-
be the most popular method of training conflicts observers tween observers. This is checked in order to determine the
(A-4, A-34, A-40, A-50). In this type of training, the uniformity in conflicts counts. Both repeatability and re-
trainee is assigned to an experienced observer who teaches liability should be tested continually throughout any type of
him the technique in the field, usually for about 2 weeks, training program. Reliability was mentioned in 10 studies
until he acquires the skill to be on his own. Videotape or by different researchers and was used both experimentally
film was used in four studies (A-is, A-22, A-29, A-52). and operationally (A-9, A-15, A-22, A-26, A-29, A-30,
Field training was used in two studies (A-il, A-52). A-32, A-36, A50, A-64).
Malaterre and Muhlrad (A-62) have recently written a There were 33 studies that mentioned accident relation-
training manual for conflicts observers (in French). ships. Some noted that the relationships were significant.
However, only two studies gave correlation coefficients
Recording Techniques (A-i, A-31). Eight studies mentioned conflicts/traffic vol-
Many types of recording techniques were used and in- ume relationships, but none presented correlation co-
clude manual counting (hand tally or counters), auto- efficients.
mated, photographs, time-lapse photography, and video-
tape. All of the studies that used on-site observers used BIBLIOGRAPHY AND REFERENCES
some form of manual counting, but most did not specify
whether it was by hand tally, mechanical counters, or both. A-I. AGENT, K. R., "Development of Warrants for Left-
Automated counters were used in five studies (A-13, Turn Phasing." Kentucky Department of Trans-
A-26, A-32, A-35, A-57), all of which involved experimen- portation, Research Report 456 (Aug. 1976) 28 pp.
tal uses of the TCT. The types of automated devices used ALLEN, B. L., and SHIN, B. T., "A New Look at the
include tapeswitches, loop detectors, and radar. Some stud- Conceptual and Emperical Aspects of the Traffic
ies used photographs as prescribed in the GM technique Conflicts Technique." McMaster University, Ham-
(A-34, A-40). Time-lapse photography was used in seven ilton, Ontario, Can. (Feb. 1977) 104 pp.
studies (A-20, A-26, A-32, A-46, A-51, A-53, A-55). Only AL-ASHARI, N., "Alternative Methods of Examin-
two involved operational use of the TCT (A-51, A-55). ing Correlation of Conflicts with Accidents." Michi-
Videotape was used in eight studies (A-2, A-9, A-12, A-22, gan Department of Highways and Transportation,
A-26, A-28, A-51, A-53, A-65) of which all but two (A-9, Traffic Engineering, pp. 34-36 (Oct. 1976).
A-51) involved research on the use of the TCT. AMUNDSEN, F. H., and LARSSEN, H. 0., "Traffic
In general, the use of the various technical recording Conflict Techniques Status in Norway." First Work-
devices was preferred in the research projects, and manual shop on Traffic Conflicts, Oslo, Norway, Proc.
observation was normally used in operational applications. (1977) pp. 34-39.
A-S. BAKER, W. T., "An Evaluation of the Traffic Con-
Analysis Methods flicts Technique." Highway Research Record No.
384 (1972) pp. 1-8.
The points of interest in this category include the use of BAKER, W. T., and GLAUZ, W. D., "The Traffic
special coding, computers, statistics, and graphs. Most of Conflicts Experience in the United States." First
the studies indicated that some type of statistical proce- Workshop on Traffic Conflicts, Oslo, Norway, Proc.
dures was used. Some of the statistical procedures include (1977) pp. 59-63.
linear regression (least squares), Spearman Rank Correla- CAMPBELL, R. E., and KING, L. E., "Rural Inter-
tion, Mann-Whitney U test, ANOVA, t-test, x2 and F-test. section Investigation for the Purpose of Evaluating
Use of computers was specifically indicated in some stud- the General Motors Traffic-Conflicts Technique."
ies. Some of the computer programs used include UTCS-1, Highway Research Board Special Report 107 (Aug.
SIMSCRIPT 11.5, OMNITAB II, SAS (MIT), BMDOZR 1969) pp. 60-69.
(UCLA), and GENPLOT (Ohio DOT). It is suspected CIMA, B. T., "An Evaluation of Freeway Merging
that computers were used for statistical purposes in more Safety as Influenced by Ramp Metering Control."
studies than were indicated. Transportation Research Record No. 630 (1977)
pp. 44-51.
Relationships CIMA, B. T., "Freeway Merging: Traffic Conflicts
This category includes measures of repeatability and re- Technique." Illinois Department of Transportation
liability (which are internal relationships), and external re- (Dec. 29, 1975) (unpublished).
lationships to accidents and volumes. Repeatability is a A-b. CLYDE, M. N., Untitled paper concerning traffic
measure of the day-to-day variation in conflict counts at a conflicts in Michigan, Traffic Conflicts Seminar,
particular site noted by the same observer. The purpose for Washington, D.C. (Mar. 20-21, 1974).
checking this is to determine if conflicts vary by day of the A-li. COOPER, P. J., "Predicting Intersection Accidents."
week and, also, to determine if the observer is consistent Ministry of Transport, Canada Road and Motor
with his counts. Repeatability was mentioned in seven Vehicle Traffic Safety Branch (Sept. 1973) 93 pp.
studies by different researchers and was used both experi- A-12. Danish Council on Road Safety Research. First
 25

Workshop on Traffic Conflicts, Oslo, Norway the Traffic Conflicts Technique." Unpublished the-
(1977). sis, Ohio State University (1975).
DENH0LM, C. J., "Feasibility Study for the Objec- MERILINNA, M. J., "Use of the Traffic Conflicts
tive Measurement of Conflicts at Intersections." Technique in Finnish Road Conditions." First
Plessey Radar Research Centre (1977). Workshop on Traffic Conflicts, Oslo, Norway, Proc.
GLENNON, J. C., ET AL., "Critique of the Traffic (1977) pp. 64-70.
Conflicts Technique." Transportation Research Rec- OLDER, S. J., and SHIPPEY, J., "Traffic Conflict
ordNo. 630 (1977) pp. 32-38. Studies in the United Kingdom." TRRL, First
GUTTINGER, V. A., "Conflict Observation Tech- Workshop on Traffic Conflicts; Oslo, Norway, Proc.
niques of the Work Done and On-Going Research (1977) pp. 1-15.
in Holland." First Workshop on Traffic Conflicts, OPPE, S., "What Task is a Traffic Conflict Tech-
Oslo, Norway, Proc. (1977) pp. 16-21. nique Intended For?" First Workshop on Traffic
HAKKERT, A. S., BALASHA, D., LIVNEH, M., and Conflicts, Oslo, Norway, Proc. (1977) pp. 111-116.
PRASHKER, J., "Irregularities in Traffic Flow as an PADDOCK, R. D., "The Conflict Technique Proce-
Estimate of Risk at Intersections." First Workshop dures Manual." Ohio Department of Transporta-
on Traffic Conflicts, Oslo, Norway, Proc. (1977) tion (1975).
pp. 71-86. PADDOCK, R. D., and SPENCE, D. E., "The Con-
HANSEN, H. K., "Brief Description of the Organiza- flicts Technique: An Accident Prediction Method."
tion." The Road Directorate, the Secretariat for Ohio Department of Transportation (Aug. 1973)
Safety Road Improvements-Denmark. First Work- 72 pp.
shop on Traffic Conflicts, Oslo, Norway (1977). PARKER, M. R., JR., ET AL., "An Evaluation of the
HAUER, E., "Indirect Measurement of Safety-The Movement of 14-Foot Wide Manufactured Housing
Conflict Method." University of Toronto (Aug. Units in Virginia." Virginia Highway and Trans-
1977) 41 pp. (Unpublished). portation Research Council, Report VHTRC 77-
HAUER, E., "The Traffic Conflicts Technique- R28 (Nov. 1976).
Fundamental Issues." University of Toronto (Jan. PARKER, M. R., JR., ET AL., "Right-Turn-On-Red."
1975) 28 pp. Virginia Highway and Transportation Research
HAYWARD, J. C., "Near-Miss Determination Through Council, Report VHTRC 76-R9 (Sept. 1975) 67 pp.
Use of a Scale of Danger." Highway Research Rec- PERKINS, S. R., "GMR Traffic Conflicts Technique
,ordNo. 384 (1972) pp. 24-34. Procedures Manual." General Motors Research
HERNANDEZ, V. L., "A Traffic Conflict Simulation." Publication, GMR-895 (Aug. 11, 1969).
(Dec. 10, 1977) (unpublished). PERKINS, S. R., and HARRIS, J. I., "Traffic Conflict
HYDN, C., "Relations Between Conflicts and Traf- Characteristics-Accident Potential at Intersec-
fic Accidents." University of Lund (May 17, 1975) tions." Highway Research Record 225 (1968)
36 pp. pp. 35-43.
HYDN, C., "A Traffic Conflicts Technique for De- A40. PUGH, D., "Relevancy of Conflict Studies to Safety
termining Risk." University of Lund (1977) 48 pp. Improvements at Intersections." Washington De-
HYDáN, C., "A Traffic Conflicts Technique for Ex- partment of Transportation (1977) 7 Pp,
amining Urban Intersections Problems." First PUGH, D. E., and HALPIN, T. J., "Traffic Conflicts
Workshop on Traffic Conflicts, Oslo, Norway, Proc. in Washington State." Washington State Depart-
(1977) pp. 87-98. ment of Highways (Mar. 20, 1974) 12 pp.
KALLBERG, H., KARHU, M., and PYYMAK!, M., "A PUGH, D. E., "Traffic Conflicts in Washington
Model for Simulating the Conflicts in City Traffic." State." Washington State Department of Highways
Technical Research Centre of Finland, Road and (Apr. 18, 1975).
Traffic Laboratory. First Workshop on Traffic Con- PUGH, W. L.,."A Study of Vehicular Accident Pre-
flicts, Oslo, Norway (1977). diction by the Use of Traffic Conflicts." Pennsyl-
KRAAY, J. H., "Some Notes About the Use of the vania State University, C.E. Department (Sept.
Traffic Conflict Method." First Workshop on Traf- 1976) 39 pp.
fic Conflicts, Oslo, Norway (1977). SHIMADA, J. K., "Hazardousness-A Concept for
LEISER, C. F., and PAINE, K. L., "A National Traffic Traffic Safety." JHK and Associates (Oct. 5, 1977)
Signal Safety Research Program." 3M Company 11 pp.
(1977) 7 pp. SHIPPEY, J., "Automatic Detection of Traffic Con-
LINDEROTH, U., and RINGHAGEN, L., "Traffic Study flicts." Transport and Road Research Laboratory
Techniques at Rural Intersections." First Workshop (1977).5 pp.
on Traffic Conflicts, Oslo, Norway, Proc. (1977) A46. SPICER, B. R., "A Pilot Study of Traffic Conflicts
pp. 40-46. at a Dual Carriageway Intersection." Road Re-
MALATERRE, G., and MUHLRAD, N, "A Conflict search Laboratory, TRRL Report LR410 (1971)
Technique." First Workshop on Traffic Conflicts, 7 pp.
Oslo, Norway, Proc. (1977) pp. 47-58. A-47. SPICER, B. R., "A Traffic Conflict Study at an Inter-
MERILINNA, M. J., "A Study of the Reliability of section on the Andoversford By-Pass." Transport
26

and Road Research Laboratory, TRRL Report Drops." Public Roads, Vol. 39, No. 1, pp. 1-6
LR520 (1972) 8pp. (June 1975).
A-48. ZAIDEL, D., KATZ, A., and ALGARLSHI, A., "A Case GRAHAM, J. L., and SHARP, M. C., "Effects of Ta-
Study of Vehicle-Pedestrian Conflict." Transporta- per Lengths on Traffic Operations in Construction
tion Research Institute, Technion-Israel Institute of Zones." Midwest Research Institute (Dec. 1977)
Technology, First Workshop on Traffic Conflicts, 33 pp.
Oslo, Norway (1977). SEYMOUR, W. M., DEEN, R. C., and HAVENS, J. H.,
A-49. ZEGEER, C. V., "Effectiveness of Green-Extension "Traffic Control for Maintenance on High-Speed
of Systems at High Speed Intersections." Kentucky Highways." Research Report 380, Kentucky De-
Department of Transportation, Research Report partment of Transportation (Dec. 1973).
472 (May 1977) 31 pp. A-59. TAYLOR, J. I., and THOMPSON, H. T., "Determining
A-50. ZEGEER, C. V., and DEEN, R., "Traffic Conflicts as Hazardousness of Spot Locations." Transportation
a Diagnostic Tool in Highway Safety." Transpor- Research Record No. 630 (1977) pp. 38-43.
tation Research Record No. 667 (1978) pp. 48-57. A-60. CLAYTON, M. E., and DEEN, R. C., "Evaluation of
A-5 1. ZEGEER, C. V., "Traffic Conflicts, Erratic Maneu- Urban Intersections Using Traffic Conflicts Mea-
vers, and Near-Miss Accidents: State-of-the-Art." sures." Kentucky Department of Transportation,
Kentucky Department of Transportation, Research Research Report 476 (Aug. 1977) 24 pp..
Report 484 (Nov. 1977) 24 pp. A-61. HAUER, E., "Design Considerations of Traffic Con-
A-52. ZIMMERMANN, G., ZIMOLONG, B., and ERKE, H., flicts Surveys." Transportation Research Record
"The Development of the Traffic Conflicts Tech- No. 667 (1978) pp. 57-66.
nique in the Federal Republic of Germany." De- A-62. MALATERRE, G., and MUHLRAD, N., "Guide de
partment of Applied Psychology, Braunschweigh Formation Aux Conflicts de Trafic." ONSER (Feb.
Technical University, First Workshop on Traffic 1977) 15 pp.
Conflicts, Oslo, Norway, Proc. (1977) pp. 99-110. A-63. Proc., First Workshop on Traffic Conflicts, Oslo,
A-53. COOPER, P., "State-of-the-Art Report on Traffic Norway (1977) 138 pp.
Conflicts Research in Canada." First Workshop on A-64. GLAUZ, W. D., "Discussion of 'Design Considera-
Traffic Conflicts, Oslo, Norway (1977). tions of Traffic Conflict Surveys.'" Transportation
A-54. BLUNDEN,W. R., and MUNRO, R. D., "Report on Research Record No. 667 (1978) pp. 65-66.
the Study of Traffic Conflicts and Accident Expo- A-65. ALLEN, B. L., SHIN, B. T., and COOPER, P. J., "An-
sure." Australian Department of Transportation alysis of Traffic Conflicts and Collisions." Trans-
(June 1976) 61 pp. portation Research Record 667 (1978) pp. 67-74.
A-55. PIGMAN, J. G., SEYMOUR, W. M., AGENT, K. R., A-66. ZEGEER, C. V., "Development of a Traffic Conflicts
and CORNETTE, D. L., "An Operational Analysis of Procedure for Kentucky." Kentucky Department
the 1-64, 1-65, and 1-71 Route Junction at Louis- of Transportation, Research Report 490 (Jan.
yule." Kentucky Bureau of Highways, Research 1978) 26 pp.
Report 326 (Apr. 1972) 5 pp. A-67. VAN DEN HONDEL, M., and KRAAY, J. H., Review of
A-56. PIGMAN, J. G., and AGENT, K. R., "Raised Pave- Traffic Conflicts Technique Studies. Institute for
ment Markers as a Traffic Control Measure at Lane Road Safety Research SWOV (1979) 112 pp.

APPENDIX B
THE STATE OF PRACTICE OF TRAFFIC CONFLICT ANALYSIS

Appendix A provides a general review of traffic conflict conflicts technique, who are called on from time to time to
use. That use, which is spread broadly throughout the apply the technique to assist in resolving operational prob-
world, includes rather fundamental research, research ap- lems. In this context, operational use is essentially confined
plications, and routine operational applications. This appen- to certain states in the United States, although many other
dix focuses on the routine operational use, although it states and foreign countries use traffic conflicts for research
cannot be totally distinguished from some research applica- or for resolving occasional operational problems.
tions. The U.S. operational experience originated in 1969,
By routine operational use, it is meant that an agency has when FHWA awarded contracts to the States of Washing-
a person or persons trained and experienced in the traffic ton, Ohio, and Virginia to aid in the evaluation of the tech-
 27

nique (4). These states had observers trained, and then flicts, they have developed the field procedures into a fairly
had them apply the technique to 886 intersection approach standardized routine. First, a supervisor may go to the
legs before intersection improvements were made, and to intersection, make a sketch, find suitable parking locations,
420 intersection approach legs after improvements. This etc. Then a techniciah counts conflicts during every other
initial activity has persisted, in varying degrees, to the 15-min period. Moreover, he alternates counting between
present. opposite approach legs. As such, four 15-min counts are
obtained for each leg in 4 hr. During the 15-min breaks
WASHINGTON from counting, the technician records his counts, takes
The State of Washington has been most active in apply- Polaroid color pictures looking toward the intersection
ing traffic conflicts routinely. The following comments are from each leg (as part of the documentation for the inter-
based on their documentation (5), plus heavy emphasis on section), and takes a careful intersection inventory accord-
an on-site review in 1978 with engineers and traffic tech- ing to the GM procedure. He is strongly encouraged to
nicians of the Washington Department of Highways. Since take notes on anything that may be potentially of interest
the FHWA contract, they have collected an additional 515 concerning the intersection and traffic in and around the
day-counts. All of this work was considered "operational," area.
and not "research," and was funded internally out of "plan- In the analysis process, three kinds of data are used:
ning" money, at about $35,000 per year. This means of (1) conflict counts; (2) the traffic volume and turning
funding has one distinct advantage—district engineers can counts (usually collected by a second technician); and
call for a conflict count at an intersection of interest, with- (3) the vehicle counts, as described by GM (basically,
out having to spend district money for it. conflict-opportunity counts).
Their normal mode of operation is to use accident data Traffic conflicts are subdivided into three categories of
to locate intersections with a potential for improvement. severity. Severity rating 1 is described as a routine conflict.
The traffic conflicts technique is then applied to help in Severity 3 is a near accident, typified, for example, by a
verifying or discounting the accident experience. In this locked wheel skid. Severity level 2 is intermediate to these
same way, traffic conflicts counts are often used in response two extremes. They are unsure of the viability of the se-
to public complaints about potentially hazardous inter- verity rating concept. Subjectively, they believe it should
sections. They have found it very useful in dealing with be advantageous. However, they have noted that they very
such concerns in a formal way either to confirm or dis- rarely observe conflicts of seventies 2 or 3, so have not had
count the perceived hazardousness. They determine the sufficient data for extensive analysis.
type of improvement to be implemented by using the traffic In analyzing traffic conflicts counts, data of various lev-
conflicts counts augmented by such things as: (1) pro- els of detail are used. At the finest level, 24 different types
fessional opinion, (2) on-site observation by a traffic en- of traffic conflicts could be observed, but data at this fine
gineer, and (3) improvements that have already been level of detail are not analyzed very often. More com-
programmed (that may or may not be related to safety monly, they reduce it to 13 levels for purposes of making
improvements). In general, priorities are determined for engineering interpretations of the operations of the inter-
safety improvements by placing intersections into four cate- section. For computer analysis purposes, these are reduced
gories, defined by the presence or absence of signalization to six types of conflicts. These are later reduced from six
and the presence or absence of channelization. Then, for categories to three—which are rear-end, cross-traffic, and
each category, a rating system is used, based on the num- other. In the final analysis, however, priorities for improve-
ber of conflicts per hour for that type of intersection. ments at intersections are determined from the grand total
Washington also uses traffic conflicts as a means of evaluat- of all conflicts.
ing improvements. They caution, however, that the method Washington currently has three technicians trained to do
is not always appropriate if, for example, an improvement traffic conflicts counts, one or two of whom have been do-
is made that is not directly related to traffic conflicts. In ing it for 8 or 9 years. They have recruited other tech-
other words, caution must be used in interpreting changes nicians as conflict observers from time to time, but not
in traffic conflict counts in view of the type of improve- recently.
ments made. Most observers in Washington were existing "planning
Over-all, the traffic conflicts technique is considered as technicians" who already had much traffic experience at
simply a formal way in which a traffic technician can ob- such things as counting and classification counting. Thus,
serve traffic operations at an intersection and report back their training experience was essentially on-the-job practice
to the traffic engineer, who normally does not have time to with an experienced observer. Although training time
personally perform extended studies of individual inter- varied in practice, a typical time was 2 weeks, for persons
sections. who already were trained traffic planning technicians.
Washington uses a modified General Motors technique.
Counting is performed for 4 to 6 hr at an intersection, at OHIO
times when the problems are expected to be the worst:
Normally, this is during weekday daylight hours, but, in The State of Ohio has also continued to use traffic con-
theory, could be on a Friday night or on a Saturday or flicts to solve day-to-day operational problems. As of 1978,
Sunday. they had applied the technique at about 600 intersections
Although they use the GM operational definitions of con- throughout the State. Most of these were rural locations,
28

and the mainline average daily traffic volumes averaged the data in a form that is easy to use, and they also include
5,200. before/after data if data were recorded in this manner.
Requests for conflict counts are received from Ohio's Predicted accident confidence intervals for four confidence
DOT safety departments, which include high accident loca- levels are also provided. The appropriate printouts are then
tion requests and district traffic departments, as well as sent to each field district.
from local authorities. Conflict counts are also taken in Ohio currently has eight technicians in the group, of
conjunction with the signal installation program. The ap- whom five are willing to work 12-hr days and travel. Two
plications have routinely involved only three of Ohio's 12 technicians have been with the program since the FHWA
field districts— Ohio uses conflict counts in three ways: study. To train a new observer, the supervisor initially dis-
before vs. after studies (70 percent); location diagnosis, cusses the TCT with him, has him read the procedures
which includes intersections and freeway gore areas manual, and discusses the manual with him. After com-
(25 percent); and as a research tool (5 percent). pleting this, the new observer is assigned to an experienced
The conflict data supplement accident, traffic volume, technician for on-the-job training until he learns the TCT,
and other data to determine the specific improvements at usually a few weeks.
a location. For example, in a before/after study of an
intersection that became signalized, the TCT may not be VIRGINIA
able to show a justification of the installation. Part of the
Virginia was the third state (along with Washington and
problem is that Ohio (and probably other TCT users as
Ohio) participating in the 1970 FHWA study (4). Unlike
well) has not been able to derive any "conflict warrants,"
the other two, however, Virginia did not continue to use
which indicate specific improvements based on specific
the technique routinely after the study was completed. On
conflicts.
the other hand, Virginia continued to use traffic conflicts
For all practical purposes, Ohio uses the GM conflicts
as a research tool for guidance in making policy decisions.
procedure, except that conflicts are now recorded as rou-
Two well-documented examples are the evaluation of right-
tine, moderate, and severe. Ohio believes that recording
turn-on-red policy (12) and the analysis of hazards in
conflicts by severity has not been of much use so far, be-
transporting oversize loads (13). In both cases, traffic con-
cause very few moderate and severe conflicts have been
flicts were a major (but not the only) contributor to the
recorded.
policy decisions.
Traffic conflicts and volumes are recorded during 10-hr
days, starting at 0730 hr, to ensure collection during the
KENTUCKY
morning, afternoon, and midday peaks—on Tuesdays,
Wednesdays, and Thursdays. The conflicts team consists Although Kentucky did not receive the same impetus of
of two people, although one person is usually able to re- federal funding that the previously mentioned three states
cord both conflicts and volumes. Counts are usually taken did, it has nevertheless been very active in applying traffic
on the two opposite approaches of the road with the highest conflicts to a variety of research and operational problems.
traffic volume. Since 1972 Kentucky has published about a dozen major
As part of the volume counting procedure, a sample reports on this or a closely related subject (e.g., erratic
lane(s) on each approach is chosen. The sample lane(s) maneuvers).
is generally a through-lane. The number of vehicles that Kentucky is using the basic GM technique with modifi-
stop, slow, or proceed through undisturbed are recorded. It cations. However, as opposed to the work in Washington
is also recorded whether the stopped vehicle's brake lights and Ohio, the Kentucky work is all being performed by
operated, did not operate, or were not observed. A turning the Division of Research for research purposes or, more
movement by the stopped vehicle is also recorded. These accurately, for solving operational problems through re-
data are used as measures of traffic flow efficiency. search. One of the major studies that they performed
Each observation period is 15 mm. The next 15 min are developed warrants for a left-turn bay or a left-turn pro-
used to record data, including a physical inventory and tected phase. TCT has not been developed and used state-
photographs, and move to the opposite approach where wide on an operational basis, although a manual is under
observation resumes again. Photographs of all approaches preparation for statewide distribution to foster wider use.
are taken at distances of 200, 600, and 1,000 ft. A total In using the technique, Kentucky counts not only traffic
of 12 forms is used to collect and assemble data so that conflicts but a number of other vehicle actions as well.
they can be arranged and analyzed by computer. The Typically, three observers are used at an intersection—one
forms include: (1) Conflict Field Notes; '(2) Traffic Con- at each of two opposing legs counting conflicts and other
flicts Header Coding Form; (3) Physical Inventory; (4) maneuvers—with the third observer doing a traffic count.
Intersection Inventory; (5) Comments or Special Data; They count continuously in 15-min blocks for 1 or 2 hr.
(6) Intersection Count Data; (7) Approach Conflicts Typically, these counting periods are the morning or eve-
Counts; (8) Approach Volume Counts; (9) Accident Data; ning peaks, or perhaps a noon rush hour.
(10) Traffic Conflicts Data Sheet, Accident Data; ( 11 ) In- In addition to the GM definitions, they observe and count
ventory of Improvements; and (12) Photographic and weaves (lane changes) that are really not conflicts, but sim-
Phasing Diagram. ply conflict opportunities. As noted in their reports, they
After the collected data are computer coded, printouts have developed an extensive list of possible types of weaves,
of the assembled data are made. These printouts present conflicts, and erratic maneuvers, which their observers re-
 29

fer to for proper encoding under one of the columns titled point out that traffic conflict counts should not be used
"other." They record a number of other situations that are alone, but only in conjunction with other measures of
definitely not conflicts, such as red-light violations, traffic diagnosis.
back-up congestion, and slow moving vehicles. Also, in The Division of Research hires technicians that are used
their lengthy list of "others" they include such items as on a variety of research projects. As such, they are not
right-turn-on-red without stop, and numerous pedestrian necessarily of the same background as technicians that
and bicycle behaviors classified as "erratic maneuvers." Fi- might be employed for operational duties in a state trans-
nally, their interpretation of a conflict such as a "slow for portation department. For example, all had finished high
left turn" conflict is broader than that of the GM definition, school and typically had some college education. These
in that they do not require the two vehicles to be "traveling persons do a number of types of work to support the re-
as a pair." search efforts, such as doing simple calculations, operating
Although the observers are instructed to designate each desk calculators, plotting graphs, etc. They also do a va-
conflict as routine, moderate, or severe, they have not made riety of types of traffic data collection in the field. (Also,
great use of these distinctions. much of their traffic conflict counting was performed by the
The Kentucky analysis activities are less structured than engineers.)
those in Ohio or in Washington, primarily because they are The first traffic conflicts counting in Kentucky was per-
using traffic conflicts as a research tool rather than an op- formed by people who were mostly self-taught, with the
erational tool. Thus, they do not use special programs and guidance of FHWA. They, in turn, have passed on their
routine analysis methods. Rather, they tend to apply what- training to their technicians. This training is primarily done
ever statistical techniques seem appropriate for the particu- in the field, on site, with "hands-on" experience.
lar study.
Most of their analyses are based on conflict numbers OTHER
rather than on conflict rates. They often examine indi- Although other states may be using traffic conflicts (an
vidual conflict types, but also usually look at total conflicts unpublished TRB committee report based on a 1974 sur-
(which they define as the total of all events recorded, minus vey states that nine states use 'conflict analysis" either
the weaves). They doubt whether this kind of total is too often or routinely), details are not documented. It is prob-
meaningful, however. For example, this total includes the able that many agencies use some form of traffic- conflict
number of times when back-ups or congestion occurs. They analysis, but perhaps do not follow the GM or other formal
suspect that these phenomena probably will not correlate definitions.
well with accidents, but rather are indicative of operational Exceptions may be the States of Michigan and Kansas.
problems. They expect that some categories of conflicts Each of these states has supported formal traffic conflict
will correlate well with certain types of accidents and thus training programs for persons throughout the state. This
believe that the analyses should emphasize those categories training, provided by FHWA, follows closely the GM defi-
(e.g., they emphasize opposing left-turn conflicts when nitions. The extent to which this training has been followed
evaluating the need for a left-turn phase). Finally, they by operational use is not known, however.

APPENDIX C

PH ILOSOPH ICAL CONSIDERATIONS IN TRAFFIC CONFLICT DEFI N ITIONS

What is a traffic conflict? In the broadest sense, a traffic related. It can be argued that an accident is simply a
conflict is a traffic event involving the interaction of two conflict where the evasive action was too little or too late.
vehicles, where one or both drivers may have to take eva- The basic problem in defining traffic conflicts, opera-
sive action to avoid a collision. The importance of looking tionally, results when the several desirable attributes of the
at some objective measure of traffic conflicts (interactions definition are examined. These attributes are:
between vehicles) is not that traffic conflicts cause acci-
Safety-relatedness--At least in a conceptual sense,
dents, but that they should be symptomatic of the same
conflicts should be statistically related to accidents (abso-
things that cause (or contribute to) accidents. Although lute measure).
here is no rational reason why traffic accidents and some Site-relatedness—They should be useful in diagnosing
measure of traffic conflicts should be causally related, there problem locations or measuring the effectiveness of a state
is every reason to believe that they could be statistically improvement (comparative measure).
30

Reliability—The definition should provide minimum Attempts to improve the likelihood of obtaining a good
variation between observers on how they record the same correlation have generally aimed to change the definition of
event. Reliability is dependent on the explicitness of con- conflicts, to bring them "closer" to accidents. One current,
flict definitions, technique tractability (not too much or too strong argument is that a proper definition of a traffic con-
little to observe), and observer knowledge (training). flict must ensure that every accident be preceded by con-
Repeatability—The definition should result in an ac- flict (7, 14). Any definition that does not satisfy this re-
ceptable level of variation in repeated observations by the quirement is unsatisfactory, according to this argument. It
same observer at the same site. This statistical fluctuation is, however, a debatable argument, because it can lead to
has an important impact on determining meaningful sample unrealistic demands on the operational techniques needed.
sizes. Much of the current theoretical and applied research in
Practicality—Reliable, repeatable, safety-related, and traffic conflicts, especially the Canadian and European
site-related data should be obtainable in a reasonable time work, deals with definitions relating to the nearness that
with reasonable resource expense. two interacting vehicles come to a collision. For example,
The last three requirements represent the major focus of whenever two vehicles enter an intersection from legs
this research effort. The second is also dealt with, to the 90 deg apart, they miss colliding (usually) by some mea-
extent that the limited amounts of data collected permit. surable amount. Commonly, time is used as the measure
The relationship between traffic conflicts and accidents was of the nearness to a collision.
not central to this research; rather, it is implicitly assumed Figure C-i shows, conceptually, how definitions of traffic
that there is some correspondence between the observation conflicts could be developed on this basis. This is termed
of evasive actions and the presence of safety (or opera- the nearness-to-collision concept. (Note: this is not the
tional) deficiencies of an intersection. The determination concept adopted in this project, but it is discussed to clarify
of this correspondence, quantitatively, is left. primarily to better the definitions finally adopted.) Figure C-i portrays
other researchers. many events (traffic interactions) via a distribution of some
Nevertheless, the keen interest in this question (the re- measure of nearness to collision. Most such . interactions
lationship between conflicts and accidents) expressed by are so devoid of collision potential that they would not be
researchers and practitioners alike demands that this proj- of interest—they are not to be defined as conflicts. How-
ect not ignore the subject. This appendix brings together ever, some threshold value of "nearness" is selected, such
the concerns and opinions expressed, and how they were that all interactions yielding smaller values are called con-
dealt with in this project, in developing operational defini- flicts. Often, these are further subdivided into moderate or
tions of traffic conflicts. severe conflicts (sometimes termed "near misses"). Fur-
First, it is in order to briefly examine why previous sta- ther, of course, the extreme situations where the measure
tistical research has been fraught with difficulties. Much is zero or less are actually collisions.
work has been done (see, for instance, the review by Glen-
non et al. (8)) in attempting to find strong correlations
between conflicts and accidents. Most of the work has been
either unfruitful or misleading for one or more of the
following reasons: Collisions
Near Misses orSerious Conflicts
Not all accidents are reported. Moderate Conflicts
Routine or Minor Conflicts
Not all accidents (in varying proportions) happen at Not conflicts
the times conflicts are measured.
Not all accidents are of the types representative of the
conflicts (e.g., single vehicle accidents).
Accident reporting levels vary tremendously from lo-
cation to location.
Not all accident data used by previous researchers
were statistically repeatable.
Not all conflict data used by previous researchers
were statistically repeatable.
Some traffic conflict definitions result in truncated
counts (e.g., do not measure events on a continuum that
includes accidents as an extreme).
Various interpretations of conflict definitions have
been used by different researchers.
Conflict definitions have not been precise enough to
ensure interobserver reliability.
Training of conflict observers has been insufficient in
some cases.
No attempt has been made to account for major NEARNESS TO COLUSION

site parameters that might influence the accident/conflict Figure C-I. Frequency distribution of traffic events, illustrating
relationship. the nearness-to-collision concept of traffic conflicts.
 31

The distribution of events can also be viewed in a cu- this case, however, vehicle C slows abruptly at time T4 ,
mulative sense, as in Figure C-2. Its scale is purposely before proceeding across the intersection at time T5. Its
compatible with Figure C-I. It demonstrates, however, that PE is T5 - T2. Thus, vehicles B and C were faced with
using this approach clearly forces all accidents to be pre- identical situations. Vehicle C reacted; vehicle B did not.
ceded by conflicts, by definition. It also suggests the rela- Yet, according to the PE values, vehicle B had a traffic
tive rarity of not only collisions but of conflicts as well, conflict whereas vehicle C did not (Note that if the GM
depending on the threshold selected. definitions were used (3), vehicle C would have a conflict
A highly desirable feature of such a curve is that the and vehicle B would not—just the opposite result.)
definition of conflicts can be represented as a point on the Still another approach is shown in Figure C-6. This
cumulative frequency curve. Thus, if this curve could be figure shows the gap time (GT), also studied by Allen (7).
generated for a given location, a direct relationship could It is the time interval between when two crossing vehicles
be drawn between the conflict measure and accidents. In would arrive at the point where their paths cross, if neither
other words, accidents and conflicts could be defined as two makes an evasive maneuver. It is therefore similar to
points in the same continuum. That this frequency distribu- TMTC, except an actual collision course is not necessary.
tion could be predicted for any intersection, however, is an A conflict is said to occur when the GT is less than some
accomplishment of questionable practicality. This curve is threshold. (Note that GT can be negative.) Figure C-6
a function of: (1) intersection volume and its distribution shows vehicle A again making left turns in front of on-
between the two roadways; (2) percentage of turning ma- coming vehicles. Vehicles B and C, representing two differ-
neuvers; (3) directional distribution; (4) hourly distribu- ent situations, both have GT values of zero for this illustra-
tion; (5) traffic mix; (6) number of lanes; (7) type of tion. Vehicle B missed the collision by simply lifting his
traffic control; (8) sight distance limitations, and numerous foot off the accelerator and coasting through the inter-
other factors. section. Vehicle C, however, was traveling much faster and
A specific illustration of the foregoing is the time mea- reacted by braking severely. Both would be conflicts ac-
sured to collision (TMTC), first proposed by Hayward cording to the GT definition; but only vehicle B would be
(15) as a measure of "near misses," but not specifically in conflictusing PE, and only vehicle C would be with GM
for conflicts. It can be visualized with the aid of the space- definitions.
time diagram in Figure C-3. Vehicles A and B are in a Both the GT and PE definitions suffer further from their
leader—follower situation when vehicle A changes speed. inability to be generalized to car-following vehicle inter-
One then continuously determines, for vehicle B, the time
remaining before a collision, assuming "they continue at
their present speeds and on the same path." The minimum
of the values so obtained is the TMTC. In this highly 100

idealized example, the time remaining is initially T3 - T1,-

but it steadily decreases until time T. when it has its mini-
mum value; T - T2, which is the correct TMTC.
The definition is more difficult to apply in real situations.
(Hayward used movie film, reduced it with a film analyzer, 80

and did the calculations on a computer.) Furthermore, a

fundamental philosophical concern arises with cross-traffic
situations, as shown in Figure C-4. In this case, vehicle A
is an opposing vehicle approaching the intersection, which
is idealized as a point, for simplicity in this example. At
time T, it slows to make a left turn in front of the on-
coming vehicle B. At time T2 it completes the movement,
and at T3 vehicle B reaches the intersection. In this case,
there is no conflict by thi definition, no matter how short
the time T3 - T2, because the vehicles were not actually
on a collision course—they missed one another without any
response from vehicle B.
This particular problem is eliminated by use of the post
encroachment (PE) time as defined by Allen (7). In effect,
it is the clearance interval between two vehicles with cross- 20
ing paths; any value less than some threshold is defined to
be a conflict. However, it is easy to illustrate how it, too,
can lead to misleading or inconsistent results. For exam-
ple, Figure C-S shows two alternative traffic events. The
first involves vehicles A and B, and is identical to that
shown in Figure C-4. The PE is T3 - T2. The alternative
event involves the same initial situation - vehicle A mak- NEARNESS TO COLUSON

ing a left turn in front of another vehicle (vehicle Q. In Figure C-2. Cumulative frequency distribution of traffic events.
32

uJ
U

T1 12 T3
TIME TIME
Figure C-3. Time measured to col- Figure C-4. Time measured to collision—
lision—car following. cross traffic.

Intersection ".__

TIME
Figure C-5. Post encroachment time Figure C-6. Gap time.

actions, because there is no specific point in space that can the definition to the actual events and respóñses they
be used for making time references. Perhaps, the best all- elicited. The former has certain ideological advantages but
around definition that resolves these problems is that of is not as adaptable to an observational methodology.
Hydén (9). His measure is TO (time of accident, in A second difficulty arises because of a fundamental dif-
Swedish), and is somewhat like TMTC. It is different in ference between cross-traffic or opposing conflicts, and
two important respects, however. First, it is not necessarily same-direction or car-following conflicts. For same-
a minimum time that must be determined by analysis of all direction conflicts, if a lead vehicle slows, the follower
possible times during an approach to a collision situation. must brake or swerve at some time, with some greater or
Rather, it is specifically the time from when the offended lesser degree of severity; otherwise, there will be a colli-
vehicle begins an evasive maneuver until a "collision" sion. Such conflicts occur repeatedly in routine driving.
would have occurred in the absence of the maneuver. Sec- For cross-traffic or opposing conflicts, the amount of
ondly, it is not necessary that the vehicles actually be on time that the vehicles will each occupy the same space is
a collision course. A "conflict" can still occur if the pro- small, and the chance that they will do it simultaneously is
jected arrival interval (in effect, the gap time) is less than remote. Thus, any given situation may or may not require
a predetermined threshold. evasive action. Furthermore, unnecessary evasive action
All of the foregoing discussion suggests that there are frequently may be taken in some situations, while no eva-
several fundamental difficulties in reaching an ideal defini- sive action may be taken in others of a more critical nature
tion of a traffic conflict. One difficulty is in deciding (closer or nearer miss).
whether one should focus on hypothetical movements or Another, more philosophical problem is to separate a
actual movements. That is, should one examine what theoretical definition of some physical phenomenon from
would happen if a vehicle were to do (or not to do) thus the methodology used to measure that phenomenon. The
or so in response to a traffic situation, or should one limit situation is a little analogous to the often quoted physics
 33

paradox concerning the definition of sound: "if in a large a desirable definition: safety-relatedness, site-relatedness,
forest, a tree falls to the ground, and there is no one around reliability, repeatability, and practicality. As stated pre-,
to witness it, does it create a sound"? One answer to that, viously, this research focused on the last three of these, and
of course, is that in theory it does create a sound (pressure did not deal explicitly with the first. However, it is neces-
waves were created in the air), even though no one was sary to forestall a trap that could arise if the first is ignored.
able to observe (hear) it. One could, for example, readily show that a very good (and
In dealing with traffic conflicts it seems wise to propose perhaps "best") measure from the viewpoint of the last
a global definition that recognizes at least a theoretical rela- four attributes would simply be traffic volume. It is cer-
tionship to accidents, such as through a continuum of tainly a practical (cheap) measure; it should be as repeat-
events as shown in Figures c-i and C-2. In practice, how- able as any other traffic measure; observers can easily be
ever, some relaxation from this definition may be necessary taught how to all count about the same; and it is related to
to make the methodology practical. That is, operational site geometrics, traffic control devices, etc.
descriptions may be quite acceptable, even if it is known However, volume is not necessarily a good measure of
that they may sonietimes result in errors. The question is safety, and certainly does not have much face validity for
really one of a tradeoff between a very rigorous definition such application. Certainly, it is difficult to believe that
that is extremely difficult to apply consistently and accu- safety funds could be justified solely on a traffic volume
rately in practice (and thus, very prone to observer error), basis. Therefore, safety-relatedness must be brought into
or a definition that is easier to apply operationally but is the decision-making process of the research study some-
"wrong" on occasion. Probably the best answer is to use way. This was done implicitly, by the choice of definitions
the definition that, at a given cost, yields the least total of traffic events that qualify as traffic conflicts. Such defini-
error (or, alternatively, the most precise estimates). To tions are based on events or situations that are strongly
illustrate this further, Table C-i gives a number of sources suggestive of having an accident potential. Nevertheless, it
for errors and the kinds of things to which each is sensitive. was not possible within the scope of this project to prove
Finally, one returns again to the five basic attributes of such a relationship rigorously.

TABLE C-i
SOURCES OF ERROR

Training Definition Technique Saiple Observer

Source Sensitivity Sensitivity Sensitivity Sensitivity Error

One kind of conflict is X

recorded as another.

A conflict is not observed

because of too much activity.

A conflict is missed because

observer is distracted.

Subjective judgment of ob- X X

server records some conflicts

that are not, and not others

that are.

Random variation. X

Systematic variation (daily

variations).
34

APPENDIX D
THEORETICAL FRAMEWORK FOR THE ACCIDENT-CONFLICT RELATIONSHIPS

One of the appealing aspects of TCT methods is a per- rather, they apparently fit the more general negative bi-
ceived association between conflicts and accidents. (Alter- nomial probability distribution function so that V(C) =
natively, both may be perceived as alternative measures of K 0, where K is greater than 1. In other words, the con-
an idealization like "hazard," if one prefers.) At this time, flict count distribution has a larger relative variance than
adequate controlled experimental data necessary to quan- the accident distribution. This does not necessarily make
tify such an association do not exist. Nevertheless, enough C inferior to A as an estimator, however, because ps>>
empirical evidence is available to allow a tentative discus- so that may be less than o/ji. (Also, of course, n0
can be increased "without limit," for all practical purposes).
sion that may be useful in clarifying the issues involved.
The "extra" dispersion in conflict observations presum-
This statistical discussion uses the following terminology:
ably arises from three kinds of sources:
A= the number of accidents during a
The short sampling unit for conflicts means that time
year at the site of interest;
of day, day of week, seasonal, etc., temporal influences may
= the "true" mean number of annual
not be averaged out for conflict counts as they are for ac-
accidents at the site of interest—an
cident observations. Thus, systematic level effects on C due
idealization (that is, the yearly av-
to short-time effects are often included in 02(C) data. This
erage that would result if one could
variance inflation is not a theoretical problem, although it is
perform the ideal experiment of
a practical problem until such effects are identified by
collecting accident data at an un-
changing site, with unchanging experimentation.
traffic characteristics, for a very In counting conflicts, there are contributions to 0 2 (C)
large number of years—thus, /hA is from observer errors as well as possible inputs from un-
a theoretical value); settled conflict definitions. Although such effects may be
V(A) = the true variance in the number of reduced by thorough training, etc., "measurement error" is
annual accidents—also an idealiza- intrinsic.
tion; Traffic conflicts are, themselves, more variable than
A = the sample mean number of annual Poisson. Substantial inherent variation would remain even
accidents, based on n 4 years of if "perfect" observers counted them at a given location
accident data; under ideal, duplicated conditions (same time of day, day
2 (A) = the sample variance of annual ac- of week, season, etc.).
cidents, based on nA years of data; The precision of C is determined by 0.2 (C) =
C = the number of traffic conflicts ob- n0 = K1 0 /n0, theoretically. Now despite the appearance
served at the site of interest during of the factor, K, the relative or percentage error in C is
a short period (e.g., I day); and much less than the corresponding error in A for practical
0,V(C),,o.2(C) = corresponding quantities for con- values of sample size. Mathematically,
flict counts, where the sample con-
sists of nc short-term (e.g., daily) 1/CVK0.2 (C)/nc < l/AVo(A)/n (D-1)
periods.
For example, if C= 30 A (e.g., the daily conflict count is
The objective of this appendix is to compare the worth 30 times the annual accident rate), the percentage error in
of two candidate estimates of A: A = the mean of the C based on one observation is about one-quarter that in a
sample of annual accidents themselves, and C = the mean corresponding A, for K = 2. (Of course, a sample estimate
of a sample of daily conflict counts. could not actually be built on one observation, but the
It is reasonable to assume that accidents have a Poisson example is used to compare C and A without introducing
distribution. This implies V(A) =JA. Thus, A is an un- the complications of sample size.) Stated differently, it
biased estimator of 14A with precision determined by 0_2(A) would take about 16 times as many years of accident data
= 0-2(A)/n4 = ILAInA, theoretically. In practice, nA will as days of conflict data to equalize the relative errors of
seldom be very large, because conditions at a site will the two.
seldom remain unchanged for many years. (Of course, one In principle, therefore, conflicts, C, are almost bound to
could use a sampling unit smaller than a year, but then 14A be more precise than accidents, A, as estimates of ILA (see
would be correspondingly depressed so that nothing essen- Table D-1 for a numerical display). Of course, the two
tial would be altered.) measures must be related structurally. In the remainder of
Traffic conflicts do not satisfy a Poisson distribution; the discussion, it will be assumed for convenience that the
 35

TABLE D-1
HYPOTHETICAL COMPARISON OF ACCIDENTS VS. CONFLICT
COUNTS AS PREDICTORS OF ACCIDENTS

Relative Errors in Accident Prediction (Percent)

From From Conflict Counts with K 2I

b/ Accidentsc/ Pc 10 A PC 30A PC
1 100.0 44.7 25.8 14.1

5 44.7 20.0 11.5 6.3

10 31.6 14.1 8.2 4.5

20 22.4 10.0 5.8 3.2

From Conflict Counts with K = 3

1 100.0 54.8 31.6 17.3

5 44.7 24.5 14.1 7.7.

10 31.6 17.3 10.0 5.5

20 22.4 12.2 7.1 3.9

a/ A sample size of one (e.g., 1 year or 1 day, as appropriate) is used to

illustrate the two methods. The estimated CV of either method would

decrease proportional to 1/ n for a real sample.

b/ The expected mean number of accidents.

ci The ratio 0 (A)/PA l(/T

d/ The ratio 0(C)/PC where K = the negative binomial inflation

in conflict variance.

relationship is of the form uA = . j (i.e., the true mean nary sampling error in an observed C), and (2) observer
accident rate is proportional to the true conflict mean with errors in counting conflicts. These errors cause bias in the
proportionately constant ). - estimation of the relationship between /xA and pe. To re-
Now, Table D-1 compares the precision of A and C as if move the bias caused by the second type of source, it is
$ were known (i.e., the precision of C in Table D-1 reflects necessary to run controlled experiments on conflict observ-
only conflict variability itself, and does not reflect any un- ing; no amount of historical conflict counts will address the
certainty arising from lack of knowledge of ). In prac- issue.
tice, the equation = $, (or whatever equation is neces- Table D-2 gives comparisons of uA and lA c where hypo-
sary) has to be "discovered." For example, an appropriate thetical observer errors are responsible for some of the
data set might yield a regression analysis that estimates $. variance of C. In this table, despite the presence of ob-
In the regression analysis the variances of A and C affect server errors, it is always true that conflicts are superior to
the precision of the estimation, but also the different sources accidents as estimators of p.s. However, on the basis of the
of variation in C affect the $ estimation differently. In par- work performed under this project, the hypothetical ob-
ticular, measurement error in observing C biases the $ esti- server errors in Table D-2 are too large. Indeed, the results
mate (b, say) (see, for example, Statistical Methods, G. W. of this research indicate that the observer errors are gen-
Snedecor and W. G. Cochran, Iowa State University Press, erally of negligible magnitude if the observers are well
6th edition, 1967, for a detailed discussion of the impact trained.
of measurement errors on regression estimates). In other Comparison of Tables D-1 and D-2 indicates that, gen-
words, observer mistakes in counting C's do not merely erally speaking, conflicts are better than accidents at pre-
make b more uncertain, but also make b systematically dicting jA, unless the observer errors in conflicts become
wrong, no matter whatthe sample size of the (A,C) data. sizeable; and available controlled data on observer errors
Therefore, in addition to a data set of (A,C) values, it is indicate that they are not sizeable. However, historical data
necessary to have a data set that measures the observer sometimes seem to display large observer errors, and there-
contributions to variation in C. fore the quantification of observer contributions to varia-
In summary, C has two kinds of liabilities as an estima- tion in conflict counts should receive high priority in the
tor of p.j: (1) intrinsic variation in conflicts (i.e., ordi- future.
36

TABLE D-2
HYPOTHETICAL COMPARISON OF ACCIDENTS VS. CONFLICT
COUNTS AS PREDICTORS OF ACCIDENTS GIVEN
MEASUREMENT ERRORS IN CONFLICT OBSERVATION

Relative Errors in Accident Prediction (Percent)

From From Conflict Counts with K 2J

Error!' UA Accidents PC 10 A SC = JUPA wc 100 A

0e2 1/4 PC 1 100.0 47.4 27.4 15.0

5 44.7 21.2 12.2 6.7

10 31.6 15.0 8.7 4.7

20 22.4 10.6 6.1 3.4

Ce2 1/2 lJc 1 100.0 50.0 28.9 15.8

5 44.7 22.4 12.9 7.1

10 31.6 15.8 9.1 5.0

20 22.4 11.2 6.5 3.5

Ce2 PC 1 100.0 54.8 31.6 17.3

5 44.7 24.5 14.1 7.7

10 31.6 17.3 10.0 5.5,

20 22.4 12.2 7.1 3.9

a! The measurement or observer error in conflict Counts.

b/ Assumes that intrinsic variation in Conflicts results in CCZ = 2C'

The hypothetical measurement variances add to CC2 so that for

Ce2 5c for ecamp1e, the total CC2 = 3c

APPENDIX E

CONFLICT DEFINITIONS USED IN FIELD STUDIES

This appendix deals with traffic conflict definitions. It which will form the basis for specific operational defi-
begins with a generalized or global definition-a frame- nitions:
work into which most operational definitions may be
placed. Then 13 basic types of intersection conflict situa- A traffic conflict is a traffic event involving two or
tions are described. This appendix continues with a discus- more road users, in which one user performs some atypi-
cal or unusual action, such as a change in direction or
sion of alternative operational definitions of conflicts for speed, that places another user in jeopardy of a collision.
each situation that was used in the observer training and unless an evasive maneuver is undertaken.
field testing.
First, a generalized definition of a traffic conflict is given, Generally speaking, the road users are motor vehicles, but
 37

the definition is broader in that they could also be pedes- With these constraints, it is obvious that the operational
trians or cyclists. definitions must encompass readily observable events. It
The action of the first user is atypical or unusual in that was judged highly unlikely, on the basis of the research and
it is not an action that every road user, or the typical road operational experiences throughout the world in the last
user, would perform under the same circumstances, al- 10 years, that relatively unskilled persons could be trained
though it need not necessarily be an infrequent or extreme to consistently "observe" traffic events, even of a near-miss
action. An example might be precautionary braking by a variety, unless some reaction to the event was elicited in
motorist driving through an intersection, even though there one of the road users. Thus, the general approach taken
is no cross traffic. This restriction does, however, rule out by General Motors (2) was adopted. The traffic event
certain types of movements that all (or nearly all) users must elicit an evasive maneuver (braking or swerving) by
initiate under the same conditions. Examples here are the offended driver.
stopping for a stop sign or red traffic signal indication, or An intersection traffic conflict can then be described,
reducing speed to negotiate a turn in the roadway. Thus, operationally, as a traffic event involving several distinct
the definition implicitly excludes actions that are in com- stages:
pliance with a traffic control device or that are required in One vehicle makes some sort of unusual, atypical, or
response to the roadway geometrics. unexpected maneuver.
Within the context of this general definition, which is A second vehicle is placed in jeopardy of a collision.
conceptual in nature, it is not necessary that there actually The second vehicle reacts by braking or swerving.
be an evasive maneuver or that there actually be an im- The second vehicle then continues to proceed through
pending collision. It suffices that the instigating action or the intersection.
maneuver threatens another user with the possibility of a
collision and, thereby, places the user in the position of The last stage is necessary to convince the observer that the
probably taking some evasive maneuver. Clearly, however, second vehicle was, indeed, responding to the offending
many collisions occur in which there are no evasive ma- maneuver and not, for example, to a traffic control device.
neuvers; they would be included as extreme cases under Within this framework a basic set of operational defini-
this broad definition. Further, there are often "near miss" tions can be stated, corresponding to different types of in-
situations in which a second driver either is unaware of a stigating maneuvers. Subsequently, alternate or modified
collision potential, or is unusually adept at estimating time definitions are also described that were used in the com-
intervals and clearances, and chooses not to make an eva- parative field evaluations. Over-all, there were 13 inter-
sive maneuver; these situations are also included under the section conflict situations that appeared to be potentially
broad definition. useful in pinpointing operational or safety deficiencies.
To further clarify this general definition, counter exam- They are given in Table E-1. The following paragraphs
ples may be given. For example, the definition would ex- describe each one. Note that all are described from the
clude "evasive maneuvers" that are strictly precautionary viewpoint (direction of travel) of a driver that is being
in nature. For example, it would not include braking or offended or conflicted with, rather than that of the road
swerving (lane changing) of a through vehicle in response user instigating the conflict situation:
to an anticipated opposing left turn, perhaps instigated by Left turn, same direction—A left-turn, same-direction
the opposing driver turning his wheels (but not encroach- conflict situation occurs when an instigating vehicle slows
ing on the lane of the through vehicle). Likewise, it would to make a left turn, thus placing a following, conflicted
not include braking or swerving occasioned by the presence vehicle in jeopardy of a rear-end collision. The conflicted
of a stopped vehicle on a cross street, which may "threaten" vehicle brakes or swerves, then continues through the
to encroach but does not actually do so. Another general intersection (see Fig. E-1).
class of exclusions is violations such as "run red light" and Right turn, same direction—A right-turn, same-
"run stop sign," unless such violations occur in the presence direction conflict situation occurs when an instigating ve-
of a through-vehicle that is placed in jeopardy of a hicle slows to make a right turn, thus placing a following,
collision. conflicted vehicle in jeopardy of a rear-end collision. The
Adopting the general definition as a basis for practical conflicted vehicle brakes or swerves, then continues through
operational definitions requires that certain assumptions be the intersection (see Fig. E-2).
made. Throughout the project it has become, and re- Slow vehicle, same direction—A slow-vehicle, same-
mained, clear that, if they are to be implemented widely direction conflict situation occurs when an instigating ve-
in the United States, any operational definitions must avoid hicle slows while approaching or passing through an inter-
or minimize the use of sophisticated equipment or that re- section, thus placing a following vehicle in jeopardy of a
quire painstaking measurements, whether in the field or rear-end collision. The following vehicle brakes or swerves,
later in the office. Thus, the operational definitions must be then continues through the intersection (see Fig. E-3).
suitable for application by human observers. Moreover, it The reason for the vehicle's slowness may or may not be
is unlikely that such definitions would be used frequently evident, but it could simply be a precautionary action or as
if they required highly educated and experienced traffic the result of some congestion or other cause beyond the
engineers as observers; the definitions should be amenable intersection. If, however, a vehicle slows while approach-
to use by persons such as traffic technicians, with suitable ing or passing through an intersection because of another
training. vehicle that is approaching or within the intersection, the
38

TABLE E-1 traffic-from-right conflict situation occurs when an instigat-

INTERSECTION CONFLICT SITUATIONS ing vehicle approaching from the right makes a left turn,
thus placing the conflicted vehicle in jeopardy of a broad-
Left turn, same direction side collision. The conflicted vehicle brakes or swerves,
then continues through the. intersection (see Fig. E-7).
Right turn, same direction
Thru cross traffic from right-A thru, cross-traffic-
Slow vehicle, .same direction from-right conflict situation occurs when an instigating ve-
hicle approaching from the right crosses in front of the
Lane Change
conflicted vehicle, thus placing it in jeopardy of a broad-
Opposing left turn side collision. The conflicted vehicle brakes or swerves,
then continues through the intersection (see Fig. E-8).
Right turn cross traffic, from right
Right turn, cross traffic from left-The right-turn,
Left turn cross traffic, from right cross-traffic-from-left conflict situation is rather unusual.
It occurs when an instigating vehicle approaching from the
Thru cross traffic, from right left makes a right turn across the centerof the roadway and
Right turn cross traffic, from left into an opposing lane, thus placing a conflicted vehicle in
that lane in jeopardy of a head-on collision. The conflicted
Left turn cross traffic, from left vehicle brakes or swerves, then continues through the inter-
Thru cross traffic, from left
section (see Fig. E-9).
Left turn, cross traffic from left-A left-turn, cross-
Opposing right turn on red (during protected left traffic-from-left conflict situation occurs when an instigat-
turn phase)
ing vehicle approaching from the left makes a left turn,
Pedestrian thus placing a conflicted vehicle in jeopardy of a broadside
or rear-end collision. The conflicted vehicle brakes or
slowing vehicle is itself a conflicted vehicle responding to swerves, then continues through the intersection (see
some other conflict situation. In this case, a vehicle fol- Fig. E-10).
lowing the slowing vehicle is said to face not a slow-vehicle, Thru cross traffic from left-A thru, cross-traffic
same-direction situation, but rather a secondary conflict from-left conflict situation occurs when an instigating ve-
situation, which is described subsequently. hicle approaching from the left crosses in front of a con-
Lane change-A lane-change conflict situation oc- flicted vehicle, thus placing it in jeopardy of a broadside
curs when an instigating vehicle changes from one lane to collision. The conflicted vehicle brakes or swerves, then
another, thus placing a following, conflicted vehicle in the continues through the intersection (see Fig. E-ll).
new lane in jeopardy of a rear-end sideswipe collision. The Opposing right turn on red-An opposing right-turn-
conflicted vehicle brakes or swerves, then continues through on-red conflict situation can occur only at a signalized inter-
the intersection (see Fig. E-4). However, if the lane section that includes a protected left-turn phase. The situa-
change is made by a vehicle because it is in jeopardy, it- tion occurs when an oncoming vehicle makes a right turn
self, of a rear-end collision with another vehicle, the fol- on red during the protected left-turn phase, thus placing
lowing vehicle in the new lane is said to be faced not with a left turning, conflicted vehicle (which has the right-of-
a lane-change conflict situation, but with a secondary con- way) in jeopardy of a broadside or rear-end collision. The
flict situation (described later). conflicted vehicle brakes or swerves, then continues the left-
By convention, in the following conflict situations the turn movement through the intersection (see Fig. E-12).
conflicted vehicle is presumed to have the right-of-way, and Pedestrian-A pedestrian conflict situation occurs
this right-of-way is threatened by some other road user. when a pedestrian (the instigating road user) crosses in
Situations such as when a "conflicted" vehicle is in jeop- front of a vehicle that has the right-of-way, thus creating
ardy of a collision because it is running a red light, for a potential collision situation. The vehicle brakes or
example, are not treated as traffic conflicts. swerves, then continues through the intersection. Any such
Opposing left turn-An opposing left-turn conflict crossing on the near side or far side of the intersection (see
situation occurs when an oncoming vehicle makes a left Figs. E-13 and E-14) is liable to be a conflict situation.
turn, thus placing the conflicted vehicle in jeopardy of a However, pedestrian movements on the right and left sides
head-on or broadside collision. The conflicted vehicle of the intersection are not considered liable to create con-
brakes or swerves, then continues through the intersection flict situations if such movements have the right-of-way,
(see Fig. E-5). such as during a "walk" phase.
Right turn, cross traffic from right-A right-turn, Secondary Conflict Situations-In any of the fore-
cross-traffic-from-right situation occurs when an instigat- going 13 conflict situations, it is possible that when the
ing vehicle approaching from the right makes a right turn, conflicted vehicle makes an evasive maneuver, it places yet
thus placing the conflicted vehicle in jeopardy of a broad- another road user in jeopardy of a collision. This type of
side or rear-end collision. The conflicted vehicle brakes traffic event is called a secondary conflict situation (it is
or swerves, then continues through the intersection (see comparable to the GM-defined "previous conflict"). Nearly
Fig. E-6). always, the secondary conflict situation will appear the
Left turn, cross traffic from right-A left-turn, cross- same as a slow-vehicle, same-direction conflict situation or
 39

..
IH I M-1 r

re E-1. Left-turn, same-direction conflict Figure E-2. Right-turn, same-direction conflict.

Figure E-3. Slow-vehicle, same-direction conflict. FigureE-4. Lane-change conflict.

40

HH . . _ A -.
-

Figure E-5. Opposing left-turn conflict. Figure E-6. Right-turn, cross-traffic-from-right conflict.

Figure E-7. Left-turn, cross-traffic-from-right conflict. Figure E-8. Thru, cross-traffic-from-right conflict.
 41

gure E-9. Right-turn, cross-traffic-from-left conflict. Figure E-10. Left-turn, cross-traffic-from-left conflict.

gure E-i1. Thru, cross-traffic-from-left conflict. Figure E-12. Opposing right-turn-on-red conflict.
 42

Figure E-13. Pedestrian, far-side conflict. Figure E-14. Pedestrian near-side conflict.

.IIIL___

Figure E-15. Slow-vehicle, same-direction secondary conflict. Figure E-16. Rig/it-turn, cross-traffic-from-right secondary
conflict.
 43

a lane-change situation, as described previously. The dif- For conflict situations involving traveling in the same
ference, of course, is that the secondary conflict situation is direction (situations 1, 2, and 3, plus secondary conflicts),
the result of an instigating vehicle that slowed or swerved the original GM work specified that the vehicles must be
(changed lanes) in response to some other conflict situa- traveling as a pair (i.e., a "car-following" situation must
tion (see examples in .Figs. E-15 through E-18). exist). In actual practice, however, some users prefer to
In all of the foregoing operational definitions it is neces- include all situations where a second vehicle brakes or
sary that the conflicted vehicle, the one that is placed in swerves (changes lanes), whether it was in close following
jeopardy of .a collision, actually takes an evasive maneuver, position or whether it came upon the leading vehicle sev-
as evidenced by obvious braking or swerving. In most cases eral seconds later. For these types of conflict situations,
the braking will be observed as brake-light indications, al- then, the less restrictive definition includes these later ar-
though a noticeable "diving" of the vehicle in the absence riving vehicles, whereas the baseline definition agrees with
of brake lights is also acceptable evidence of an evasive the GM definitions and includes only paired-vehicle
maneuver. situations.
Alternate operational definitions were also used in the For the other types of conflict situations, the GM study
field tests to determine their value relative to the baseline suggested counting "vehicles." Many users do this, believ-
definition. For each of the 13 conflict situations (14, in- ing that such data may be useful. An alternate terminology
cluding secondary conflicts), one less restrictive and one is suggested, "opportunities." Thus, for example, all lane
more restrictive definition was examined. These definitions changes on the approach to the intersection represent op-
were developed on the basis of observations of actual prac- portunities for traffic conflicts as previously described; they
tice and on evidence in the literature. become such only if a conflicted vehicle is relatively close

II I II I

I—

__ 11 P 1HH:
Figure E-17. Left-turn, cross-traffic-from-left secondary Figure E-18. Pedestrian, far-side secondary conflict.
conflict. .
44

and it reacts by braking or swerving. The less restrictive Sweden. A time-to-collision measurement is used, defined
definition includes situations where there is no reaction as the time interval from when a conflicted vehicle reacts
(even though the instigating action created an opportunity). (brakes or swerves) until a collision (or a near miss) would
So much for less restrictive definitions. The other option have occurred had there been no reaction. Specifically,
is to include only a subset of the baseline traffic conflicts
Hydén defined a conflict as serious if the time-to-collision
within a more restrictive definition. This subset consists
value was less than the threshold of 1.5 sec, as determined
only of traffic situations that lead to conflicts that exceed
some threshold level of "severity." subjectively by trained observers. This approach is adapt-
Of the numerous approaches toward development of de- able to all types of conflict situations previously described,
scriptive definitions of severe conflicts developed in the and was reportedly capable of being applied in the field.
United States and elsewhere, it was determined that the Table E-2 summarizes the total set of conflict definitions
most promising was that developed by Hydén (9) in used in the field tests.

TABLE E-2
FIELD-TESTED CONFLICT DEFINITIONS

"24' Brake/Swerve 'Severe

Conflict Situation Opportunity All Pair Only Conflicts

Left Turn, Same Direction X. x x

Right Turn, Same Direction X X X
Slow Vehicle, Same Direction X X X
Lanechange X X
Opposing Left Turn X X X
Right Turn, from Right X X x
Left Turn, from Right X X x
Cross Traffic, from Right X X X
Right Turn, from Left X X x
Left Turn, from Left X X x
Cross Traffic, from Left X X X
Opposing RTOR X X X
Pedestrian x x x
- Secondary X x
 45

APPENDIX F
FIELD STUDIES

This appendix provides details about the field studies. site was observed by each person, and each person had
They were designed to test the various operational defini- been paired with each of the other 11 at least twice.
tions of traffic conflicts presented in Appendix E for their The experimental design for the conflict observations had
repeatability, reliability, and practicality, as well as for pro- as primary goals: (1) to estimate repeatability (i.e., the
viding information on the site-relatedness of the various variance of a single observer when nominally replicating
types of conflicts. This appendix has four components: a conflict observation); and (2) to estimate reliability (i.e.,
(1) formal experimental design, (2) observer recruiting the variance of conflict observations due to differences be-
and training, (3) field study procedures used in these tween observers. These goals amount to being able to esti-
tests, and (4) actual conduct of the studies. mate the residual variance, 0-0 2 , and the observer-observer
variance, 0-2
EXPERIMENTAL DESIGN The design also enables achievement of the secondary
goals of estimating: (1) factor effects, where the factors
A preliminary experimental design developed for the are type (Tl = 4-lane unsignalized, T2 = 4-lane signalized,
field tests envisioned using four independent variables— T3 = 2-lane unsignalized), number of legs (Nl = 4-way,
number of intersection legs (3 or 4), number of lanes on N2 = 3-way), and speed (Si = 40 mph or more, S2 =
the approach legs of interest (2 or 4), traffic control (sig- less than 40 mph); (2) site or location, L, effects (i.e.,
nalized or stop sign), and speed limit (high or low). The variation in conflict levels between nominally identical
factor, speed limit, was selected because traffic speeds were physical locations); (3) variation in conflict levels between
expected to have a greater influence on traffic conflicts than days of week; (4) variation in conflict levels between op-
a designation of urban or rural, based on preliminary ob- posite legs of sites; and (5) temporal effects. There are
servations in earlier tasks of the project. Traffic volume, three kinds of temporal effects, in addition to day of week,
which was suggested as an independent variable in the estimable from the data set: short-term (I 5-min intervals),
project work plan, was retained, but as an implicit rather morning vs. afternoon; and long-term (first 3 weeks vs.
than an explicit parameter. It was decided to use actually second 3 weeks vs. third 3 weeks).
measured traffic volumes rather than a simple "high" or The foregoing sets of goals were met by attributing,
"low" designation applied to the average annual daily traffic through statistical analysis, the variations in conflict counts
(AADT). Moreover, retaining the actual volume enables to the factors mentioned. These "components of variance"
much more flexibility in interpreting the results obtained. have two kinds of interpretations according to the physical
The originally conceived plan would have used 24 inter- nature of the effect being described. For example,
sections, of which half would have 4 legs and half would really is a variance in the usual sense, because the 12 ob-
have 3 legs. Also, they would be equally split between sig- servers used in the field studies are regarded as representa-
nalized and stop-sign controlled, as well as between 2-lane tives of a theoretical population of potential observers. In
and 4-lane. However, it was anticipated that 16 of the other words, the 12 observers are used to estimate the
intersections would be on lower speed routes, while 8 general variability imparted to conflict counts by human
would be on higher speed routes, in general conformance observers.
with the mix of intersections expected to be found A factor such as speed also produces a variance com-
nationwide. ponent (0-S2, say), but the interpretation is different be-
The final experimental design was predicated on the ex- cause the two levels, Sl and S2, are not representatives of
istence of suitable sites. Table F-i presents the design a speed level "population," but are, of course, the two spe-
framework finally developed. It contains 24 sites that form cific speed levels selected for evaluation in the experiment.
the basic experiment (the first three rows), plus 4 sites that Thus, a-S2 is not itself pertinent; rather, the specific dif-
form a subsidiary experiment. The following discussion ferences reflected by Si vs. S2 (and their statistical sig-
concentrates on the basic experiment. nificances) are relevant in the assessment of the speed
Before presenting the statistical aspects of the design, an factor. Following customary nomenclature, however, all
overview is in order. The basic experiment included 24 factors are labeled in terms of variance components, for
sites, as stated. These sites were observed by 12 persons. convenience.
They worked in pairs, with the 2 persons in the pair alter- Although the reliability derived from a-02 is a primary ex-
nating approach legs every half-hour. After a half day perimental objective, the explicit assessment of specific ob-
(morning or afternoon), during which 6 sites were ob- server differences is also of some interest in the research
served, the pairings were rotated, and another 6 sites were because, for example, this variance might arise primarily
visited. This process was continued on a 4-day per week from one unique observer vs. 11 relatively homogeneous
basis, so that after 3 weeks (12 days or 24 half-days) each observers, etc. Therefore, the experimental design also al-
 46

TABLE F-i what one could predict from knowing the observer, time,
EXPERIMENTAL DESIGN FRAMEWORK and leg at a particular site. This variation in turn arises,
theoretically, from the countless unobserved variables op-
High Speed Low Speed erating continuously on the traffic situation. Thus, the term
Experiment Lanes Signalization 4-Way 3-Way 3-Way
"repeatability" is appropriate, because 02 describes an ir-
Basic 4 No X X X X reducible lower limit to the precision obtainable in conflict
observations. (Theoretically, of course, future experiments
4 Yes X X X X
could incorporate new factors that would explain part of
2 No X X X K cr 2 )

Repeatability and reliability estimations are the primary

Subsidiary 2 Yes X - X -
goals of the experimental analysis. However, the conflicts
experiment also measures factor effects, because the fac-
a! Each X represents two physical sites, each with two legs or approaches tors T, N, S,and L form a nested factorial experiment (L
nested in TNS cells), except that only 6 of the 24 sites can
being observed.
be visited in any one half-day.
Thus the data set is a 3 . 23 factorial, with the restric-
lows for estimation of specific observer differences, as dis- tion that only 6 cells can be observed at a time. Four such
cussed in the following. periods are represented by four half-days of the week, so
For convenience, label the situation at a site during a that every "one-half week" (4 mornings or 4 afternoons)
half-day (morning or afternoon) a "visit." At each visit, is a complete 3 23 factorial in "blocks" of six. This ar-
a particular pair of observers measures conflicts under a rangement is balanced Plan 6.10, CRC Handbook of
unique set of conditions. Among the 12 observers there are Tables for Probability and Statistics, p. 70) and, thus, all
66 possible observer pairs, and the construction of each effects are retrievable in the AOV. A factor stage is shown
possible pair yields, by definition, a balanced incomplete in Figure F-2. Such a stage consists of 3 weeks of half-
block (BIB) design. During a 3-week segment of the con- days. Of course, the factor stage analysis measures day
flict experiment this 66-visit BIB is itself replicated twice. effects and week effects in addition to T, N, S, and L effects.
(Twelve pairs/day X 12 days = 144; 144/66 = 2 + re- The only looseness in the design of the factor stages
mainder.) It is relatively immaterial to the determination arises from the fact that different observers produce the
of o- how the visits are distributed among the sites, be- readings in each of the stages of the design. This was not
cause the only point is that all possible pairs of observers anticipated to be a problem because: (1) in effect, the
are run. various observers are randomly assigned to the locations,
Next, the estimation of the repeatability, 0 c2, is consid- (2) over the whole experiment the locations are all seen
ered. Ideally, is the variance of replicated observations by all of the observers, and (3) factor effects small enough
(i.e., the (mean) dispersion of counts by one observer at to be smothered by observer variations are too small to be
a specific site under the same conditions). Of course, iden- physically useful, anyway.
tical traffic conditions do not occur in practice, so two suc- Finally, the entire set of field trials consists simply of
cessive Tuesdays at 9:00 am., for example, are replicates 6 replicates of the factor stage (three 3-week periods of
only nominally. The actual occurrence of conflicts will be half-days—morning and afternoon). The 3-week periods
variable between two such periods. In practice, therefore, or "phases" are useful because: (1) a long-term trend
O ( includes two effects—the variability in an observer's could arise in the data either physically or because of ad-
ability to properly recognize traffic conflicts, and conflict ditional learning effects among the observers; and (2) the
generation at a given site and time. middle phase uses Tuesday through Friday (instead of
Technically, o is estimated as the residual variance Monday through Thursday) as the days of the week, so
from factorial subsets (see Fig. F-I). Each 1-hr observa- that eventually all S weekdays are observed.
tion at a site is one-half of a 23 factorial' design over the The foregoing discussion pertained to the basic experi-
factors observer, time interval, and leg. Thus, each 4 hr at ment using 24 sites. The subsidiary experiment involved
a site produces 16 observations obtained in four "blocks."
Each visit thus yields 9 deg of freedom for the estimation
of o• 2 .
Perhaps it is worthwhile to belabor 0. 2 somewhat. At
Time Leg
any site an average difference between the observer counts, Observer Interval 1 2
the leg counts, and the time counts can be observed in the
data. The variance,
1 x -
arises from the inability of these A
average effects to exactly predict each observed conflict 2 -' x
count. That is, is the variance of the deviations be-
1 - x
tween the observed counts and counts predicted from B
knowing the average effects of observer, leg, and time at 2 x -
each site. By definition, this discrepancy arises because the
Figure F-I. Experimental design 'of a 1-hr observation
true traffic at a point in time is variable over and above block.
 47

Day

Monday Tuesday Wednesday Thursday

oiooJ 0000 0001 0010

Week I
0011 0111 0110 0101
1010 1001 1000 1100
1101 1110 1111 1011
2001 2010 2100, 2000
2110 2101 2011 2111

Week 2 0010 0001 0000 0100

0101 0110 0111 0011
1001 1010 1100 1000
1110 1101 1011 1111
2100 2000 2001 2010
2011 2111 2110 2101

Week 3 0001 0010 0100 0000

0110 0101 0011 0111
1100 1000 1001 1010
1.011 1111 111.0 . 1101
2010 2001 2000 2100
2101 2110 2111 2011

a! Digits represent the levels of the factors Type (T), Number of Legs

(N), Speed (5), Location (L) where, in general, '0' stands for the

first level of a factor, "1" for the second level, etc. Thus, 0100

= 1st type, 2nd number, 1st speed, 1st location.

Figure F-2. Factor stage of conflicts experiment.

four additional sites, all of which were 4-way, signalized colleges, and many persons were contacted who had ap-
intersections on 2-lane streets. Two were on high-speed plied for a subprofessional job at MRI during the last 6
routes and two were on low-speed routes (see Table F-I). months or so (including stenographers, etc.).
The additional sites enabled an estimation of the conflict The preliminary screening process involved obtaining
rates at these kinds of sites, and also provided data on the answers to the following four major questions:
specific alternate observers, which is useful because they Are you 18 years old or older?
may have a higher interobserver variance than the 12 Do you have an automobile and a valid operator's
observers used in the main experiment.
license?
Do you have references that will indicate you are a
OBSERVER RECRUITING AND TRAINING very dependable, trustworthy person that has the self-
The first step in this process was to attract a large num- discipline to carry out assignments with minimal super-
ber of applicants. It was the intent of the research team to vision?
attract as broad a spectrum of applicants as possible. It Are you available for work from May 30th through
was expected that the major source of applicants would be August 11th, to work between the hours of 7 a.m. and
respondents to newspaper advertisements. To this end, a 6 p.m. 4 weekdays per week?
working relationship was developed with the Missouri Em- Persons who could respond affirmatively to these questions
ployment Office. A newspaper advertisement, placed to were then eligible for a personal interview.
run for three consecutive weeks, was worded so as to at- More than 80 persons were invited for a personal inter-
tract persons from all walks of life, not just students. Re- view, and most responded to the invitation. All respon-
spondents were directed to apply to the Missouri Employ- dents were interviewed both by a professional interviewer
ment Office. Applications were screened there, and those from. MRI personnel services office and by a member of the
that passed the screening processes were referred to MRI. technical staff. In addition to confirming that all answers to
In addition to this procedure, several others were used. the previous questions were affirmative, each applicant was
The job opening was, of course, posted within MRI, as is also rated regarding the following criteria:
normal practice. This resulted in some applicants attracted
by word-of-mouth. Also, most of the retirement homes and Expresses strong interest in the type of work.
complexes in the Kansas City Metropolitan Area were con- A mature attitude, with indications of a sense of
tacted. Notices were posted in the local high schools and responsibility.
48

An interest in outdoor work. eliminated cross-traffic problems to better focus on this par-
Not overly concerned about the potential routine ticular group of traffic conflicts. At the end of the second
nature of the work. day the intent was to introduce another set of traffic con-
5; Some record of prior work experience. flicts. However, there were numerous questions requiring
All applicants were placed into four rating categories: special consideration, either to further illustrate the basic
top prospects, secondary prospects, possible prospects, and definitions involved or to clarify special cases that required
unsuitable prospects. References were then contacted, es- interpretation. This experience was extremely beneficial,
pecially prior employers, of all persons in the top prospect and was followed throughout the remainder of the training
category as well as many persons in the secondary prospect program.
category. Following this, job offers were made to nearly The third day it rained much of the morning, requiring
two dozen applicants. Of these, 19 accepted and began the more time to be spent in the classroom than originally an-
training program. One person dropped out early in the ticipated. This time was used to observe traffic conflicts
program, for personal reasons. Another was terminated at that had previously been recorded on videotape, both in
the conclusion of the training program because, although Ohio during an earlier field trip there, as well as at a num-
he was very conscientious and hard working, he found him- ber of intersections in the Kansas City area. A short test
self unable to agree philosophically with the traffic conflict was also administered using the videotape. The second set
definitions and, therefore, did not produce results compati- of traffic conflicts was then introduced which consisted of
ble with those of the other observers. all the cross-traffic movements with the exception of the
The 17 remaining persons formed the crew of conflict right-turn-from-left traffic conflict. In the afternoon the
observers. They came from a variety of backgrounds, with remaining types of traffic conflicts, including secondary
none of them having traffic engineering credentials. All conflicts, were introduced. Over-all, about half of the third
were relatively young (early 20's) and most were students, day was devoted to field observations, again in groups of
although one was a secondary education teacher and an- four or five trainees with on-site supervision.
other a college graduate without a job. There were 6 fe- The fourth day included more supervised small group
males and 11 males. Moreover, the group contained a work in the field. Rear-end conflicts were considered in
mixture of racial and ethnic backgrounds. more detail, distinguishing between those that satisfy the
The training program covered a 9-day period. Three of "paired-vehicle concept" included in the General Motors
the persons were singled out for an additional 1 to 2 days definition as well as the more general definition where the
of supervised field practice to clear up a few isolated diffi- offended vehicle need not have been "following" the lead
culties. The basic course outline is shown in Figure F-3. vehicle. This distinction was initially difficult for many of
This outline is essentially the same as the one developed the trainees to comprehend. Further testing, using video-
prior to beginning the training program, and was adhered tapes, as well as more conventional classroom question-
to for the most part. However, there were some departures and-answer-type of tests, was also accomplished on the
from the original schedule for a variety of reasons, such as fourth day.
rain showers during a scheduled field trip, perceived needs Subsequently, only one new traffic conflict concept was
for more (or less) classroom presentation, and extra dis- presented—the "serious conflict." As indicated in Appen-
cussion on various topics. Following are a number of com- dix E, the Swedish definition was used for this concept.
ments concerning the schedule and the activities therein. Training was facilitated using a videotape obtained from
The first half-day was low key. It was a time for intro- the Lund Institute of Technology. The tape, prepared for
ductions and orientations, and was concluded by showing training purposes, had to be rerecorded on a cassette to
film, viewgraphs, and slides dealing with transportation and enable playback on a U.S. version of the Sony Eumatic.
traffic safety in general. Following a presentation and review of the traffic events
The afternoon of the first day covered traffic volume on this tape (there were 41 traffic conflicts, each of which
counting, with about an hour and a half of field practice had been carefully analyzed by the Lund Institute and the
in this skill. During this time the trainees also observed corresponding "time to collision" calculated), several hours
traffic movements in an unstructured way, on a small group were devoted to field observation in Kansas City's Country
basis, with special attention directed toward driver actions Club Plaza area. Traffic flow patterns there are somewhat
that might be indicative of unsafe practices. The purpose similar to those on the Swedish videotape. There is rela-
of this activity was to provide a common basis for intro- tively little traffic control in this area, with a moderate
ducing the more formal definitions, beginning the next day. amount of pedestrian traffic and some bicycle traffic. It
Most of the second day was taken up with an introduc- was anticipated that a number of serious conflicts would be
tion to the subject of traffic conflicts, with an emphasis on seen. In fact, however, there were fewer than expected.
a subset of conflicts that were termed as "Group 1." This Also at this time, a procedure was instituted that was
group consisted of all the rear-end type of conflicts plus the followed throughout much of the remainder of the train-
opposing left turn. The material was presented using a film ing. That procedure consisted of pairing the trainees—
and 35-mm slides, together with discussions of some of the different pairs every half-day—and directing the members
fine points. About half the day, however, was spent in the of each pair to simultaneously observe and record the same
field where the trainees worked in groups of four or five flow. Furthermore, two pairs were sent to the same inter-
persons, under the purview of an MRI staff member, at sig- section to alternatively observe opposite legs. This pro-
nalized intersections. The use of signalized intersections cedure allowed estimates of interobserver reliability and de-
 49

Period Content Period Content

Day I - Morning Day 3 - Afternoon

(Lecture and discussion) Welcome (Review, lecture, group Remaining types of conflicts

Introductory remarks field practice)

Hiring details Day 4 - Morning

Introduction and orientation (Lecture, discussion, small Details of rear end conflicts

to the course group field observations) (paired-vehicle concept)

Objectives and plans Day 4 - Afternoon

Discussion of the (Review, testing, video

transportation system tape, small group field

Day 1 - Afternoon observations)

(Lecturing and preliminary Volume counting

Day5
field activities) Turning movements
(Review, orientation, field Serious conflicts
Data' recording
work in independent pairs)' Field procedures
Use of count boards
Day6
Field work in volume
(Field work in independent All conflicts, including
counting
pairs using field procedures) serious conflicts.
Recognition of unsafe
Day7
maneuvers
Day 2 - Morning (Field work in independent Full work day,

(Lecture and group Rear end and opposing left pairs)

field practice) turn conflicts DsyB

Day 2 - Afternoon (Bus tour, review, fully Tour of experimental field

(Discussion and group Review of morning activities independent field work) sites

field practice) More field recording Individual,review and

Introduction to cross-traffic consultation

conflicts Day 9

Day 3 - Morning (Fully independent field

(Testing, review, and All conflicts discussed to date work, final review, project

group field practice) assignments.

Figure F-3. Intersection conflicts--outline for pro jeôt training course.

termination of which trainees might be having difficulties tions, again using the independent pair concept. This work-
with certain concepts. day included the morning peak-hour.
On the sixth day, the trainees operated essentially with- On the eighth day, the morning was devoted to a bus
out supervision, including the evening peak-hour. The seri- tour to orient the observers to the locationsand routes serv-
ous conflict concepts were also reviewed, and the trainees ing the 28 field sites to be used in the experiment.
were tested using a selected number of sequences from the This time was also used to review, individually with the
Swedish tape displayed in a random order. The trainees trainees any particular difficuties or problems they had
correctly identified 81 percent of the conflicts as being been having in traffic conflicts observation. This review
either serious or not serious. was facilitated by a statistical analysis of each trainee's field
The seventh day was totally devoted to field observa- work from Day 3 through Day 7. All field work was turned
50

in and the MRI technical staff scored the work on a com- FIELD STUDY PROCEDURES
parative basis. That is, for a given location and approxi-
mate time period, the mean number of conflicts and op- The intersections for the field studies were selected to
portunities was determined. Those observers whose data provide a wide range of intersection types. A primary con-
entry in any particular data field was more than one stan- sideration, however, was that the test intersections have
dard deviation above or below the group mean were then relatively simple signalization and geometrics. Intersec-
identified. Situations in which the standard error was rela- tions without sophisticated signal phasing and channeliza-
tively high indicated a need to review the basic concept tion tend to produce more conflicts than intersections with
with the whole class. On the other hand, if the standard separated traffic movements. Also, intersections in the
error was relatively low, but one or two individuals con- Kansas City metropolitan area, if of basic design, would
sistently departed from the group mean, those individuals tend to operate similarly to the same types of intersections
were singled out for special consultation. in other parts of the United States. By emphasizing the
more basic and common geometrics and signalization, the
To supplement this comparative testing procedure and to
findings should be of more universal value.
provide a final basis for rating each of the trainees, the
The following general guidelines were considered for a
afternoon of the eighth day was devoted to rather inten-
sive comparative field observation and testing. This con- possible test site: minimum pedestrian traffic, no unusual
sight restrictions (horizontal, vertical, and cross-corner
sisted of assigning half of the trainees to each of two legs
sight distance); no appreciable grade (downgrades pro-
of an intersection for simultaneous but independent obser-
duce more brake-light applications); no parking restric-
vation and recording. The data from this exercise were
tions during just a portion of the workday ("no parking
analyzed very carefully for group and/or individual prob-
lems. For the most part, the findings confirmed those ob- any time" restrictions were acceptable); no one-way streets;
no extreme turning volumes; available observation loca-
tained earlier—that one of the individuals consistently
tions on the two studyapproaches; no oblique or offset
scored nearly every type of conflict in a unique fashion,
intersections; no appreciable driveway traffic; no unusual
whereas the remainder were relatively consistent with only
traffic peaking; no unusually heavy truck or bus traffic;
a few minor exceptions. It was subsequently decided to
minimal congestion; and no freeway ramps. Obviously,
terminate the services of that individual but to retain the
some compromises were required. The final selections are
remainder, with• a small amount of additional individual
shown in Figure F-4 superimposed on the experimental
attention for a few.
plan.
The ninth day consisted of fully independent traffic con- Of the 24 regular test sites, 12 were signalized. Eleven
flict observations without supervision in the morning, fol- of these intersections operated with fixed-time signal con-
lowing all the field procedures described subsequently. In trollers, the other used a traffic-actuated signal on the cross
the afternoon, final questions were resolved, and individual street. The signalized intersections did not have extreme
assignments for the field studies were made. cycle splits. That is, the two green phases in each signal

High Speed (> 40 mph) Low Speed (< 40 mph)

4-Way 3-Way 4-Way 3-Way

4
1. K-7 - 87th St. 3. K-7 - K-lO 5. Holmes Rd. - 7. Antioch Ave. -
79th St. 99th Terr.
Unsignalize 2. US 40 - Lee's 4. US 40 - Stadium 6. Mission Rd. - State Line Rd. -
Summit Rd. Dr. 89th St. 85th St.
4-Lane
Bannister Rd. - Troost Ave. - 13. Oak St. - 15. Santa Fe Dr. -
James A. Reed Rd. 82nd Terr. Volker Blvd. 83rd St.
Signalized
0. Troost Ave. - US 50 - 4. Volker Blvd. - 16. Broadway Ave. -
89th St. Elmwood Ave. Oak St. Pershing Rd.

7. M-291 - K-150 - 1. Wornall Rd. - 23. Lee Blvd. -

Woods Chapel Rd. State Line Rd. Minor Dr. 85th Terr.
nsignalize
8. K-iSO - 0. State Line Rd. - 2. Nail Ave. - 24. Mission Rd. -
Quivers Rd. Blue Ridge Blvd. 99th St.
________ ________________ 99th St.
2-Lane Ext.
5. M-291 - 7. 75th St. -
Colburn Rd. Switzer Rd.
Signalized
6. Noland Rd. - Wornall Rd. -
US 50 55th St.
Figure F-4. Array of Sites selected (observed street named first, followed by cross street).
 51

cycle were approximately equal. Two intersections had hour, but they usually work individually (i.e., one observer
separate left-turn phases. The unsignalized test sites were per intersection). If traffic volume is counted by a state,
all two-way stop intersections, controlled either by signing it is usually done by a separate observer. For this project,
or signing in conjunction with flashing signals. There were two observers were assigned to each intersection so that the
no uncontrolled, yield, or 4 -way stop intersections. reliability of conflict types and numbers could be compared
Potential test sites were all in the portion of the met- between observers.
ropolitan area south of the Missouri River, which is the A typical intersection diagram is shown in Figure F-7.
city's north-south divider. Besides providing a geographic In addition to providing geometric and traffic control de-
boundary, the Missouri River is also a natural traffic bar- tails, the diagrams also indicate the approximate observer
rier with only a few crossing points. Inasmuch as the con- locations.
flict observers worked at two sites per day, it would be Four types of conflict forms were used in the study; they
difficult for them to travel between sites separated by the are shown in Figures F-8 through F-il. The four forms
river in the travel time allotted. More importantly, the treat three types of intersections—unsignalized, 4-leg; sig-
majority of the population of the Kansas City metropolitan nalized, 4-leg; and 3-leg. The 3-leg intersections require
area is located south of the river. This means that the separate forms for the "left" and "right" legs. Each of the
streets and highways south of the river tend to carry higher forms illustrates the conflicts that are common to the ap-
traffic volumes, which, in turn, could create larger num- propriate type of intersections. The forms also contain
bers of traffic conflicts. (Sites with volumes so low as to blank columns for field use, so that other types of conflicts
virtually assure few or no traffic conflicts were purposely occurring frequently at a particular test site can be added.
not selected.) The types of information contained in a completed con-
Of the 28 test sites, 11 were in Kansas and 17 were in flict form are now explained. The intersections were coded
Missouri, and all were within a 20-mile radius of Midwest from 1 through 28 for purposes of the research study (see
Research Institute. The Missouri sites were located in Kan- Fig. F-4). This is the "Location Number." The name of
sas City, Missouri, as well as in two other jurisdictions. The the intersection is included in the "Description." All inter-
Kansas sites were in several different communities, and two sections were named such that the street on which, the ob-
of them were in rural, unincorporated areas. servers were stationed was given first. For example, at the
The design called for the conflict observers to work four Holmes Road-79th Street intersection, the observers would
of five weekdays each week. Although conflicts were ulti- be stationed on Holmes Road, with 79th Street being the
mately observed on all weekdays, the majority of the work cross street. The "Leg Numbers" correspond to directions
was concentrated on Tuesdays, Wednesdays, and Thurs- of the compass. The north, east, south, and west ap-
days. This was done in order to collect a large amount of proaches are numbered 1, 3, 5, and 7, respectively. The
data on what were considered "typical" weekdays, as well "Day" of the week and corresponding "Date" are next.
as to ensure data collection during morning and afternoon The "Time Period" was indicated as either a.m. or p.m.
peaks, while maintaining a 40-hr work week. Each observer wrote his name and observer number in the
Each of the 12 regular observers worked daily at two of final space. The count start time was the starting time of
the 24 regular test sites. There were two observers at each each 15-min observation period. (Any observation period
test site, and each observer had a different partner for the not equal to 15 min would be specially noted.) The total
morning and afternoon periods. Thus, a total of 12 inter- 15-min opportunity and conflict counts were written in the
sections was observed each day. The remaining five ob- appropriate blocks. Any notes or comments were to be
servers were given daily, special assignments. These people written on the back of the form.
observed conflicts at the extra locations, replaced regular The supplementary information contained in the Traffic
observers in emergency situations (sickness, car trouble), Volume Counts Form (Fig. F-12) is the same as that on
performed make-up work, and did project-related office the conflicts forms. It was learned during the training ses-
work. sions that volume counts agreed reasonably well among
The 10-hr workday started at 0700 hr and ended at observers. For this reason, and as a matter of simplicity,
1800 hr. Four hours of observation occurred at each the observers each counted traffic volume on only two ap-
morning and afternoon location. The remainder of the proaches of the intersection. In other words, duplication of
workday was allocated to travel time, preparation time, and this counting was not deemed necessary).
scheduled breaks. Figure F-5 shows a detailed daily work Daily site visits were made by at least one member of the
schedule. The materials needed by each observer are con- project staff. During this time, questions were answered
tained in Figure F-6. Each morning and afternoon con- and supplies replenished. The observers were provided
tains eight half-hour count periods. Of each half-hour pe- with the telephone numbers of the project staff and were
riod, 15 min were spent observing conflicts, which were urged to telephone if any problems needed immediate at-
then recorded; 5 min were spent counting traffic volumes, tention. On days when it rained, observation stopped until
which were then recorded; and the remaining time was the roads appeared to be drying. If in doubt, the observers
used to move to the opposite approach of the intersection. telephoned a staff member.
The two people at each intersection observed conflicts The individual site visits provided good contact, but only
and traffic volumes on the same street and at the same time, on a one-to-one basis. Also, the observers had little oppor-
but from opposite approaches. State transportation depart- tunity to maintain contact with the group after the training
ment conflict observers also change approaches every half- program. For these reasons, a newsletter was distributed to
 OBSERVATION PROCEDURES -LA
Field Needs - Each Observer
Outline
City map with location of test sites
Weekly location schedule
7:00 - 7:15 Preparation at site
7:15 - 9:15 Observation Procedures
Observation
9:15- 9:45 Count Board
Break
9:45 - 11:45 Observation Appropriate data forms
11:45 - 13:15 Lunch break and move Conflicts sheets (2/day/location)
13:15 - 13:30 Preparation at site Volume sheets (2/day/location)
13:30 - 15:30 Observation Obtain next week's date forms from supervisor on last work day of week
15:30 - 16:00 Break List of project supervisors, including reporting procedures
Complete set of intersection detail drawings
16:00 - 18:00 Observation
Tablet
Pencils
Detailed Schedule Note book binder
I.D. card
0700 Locate observation position; check intersection drawing to be Time piece (watch)
sure of observation position. Folding chair (optional)

Observe operational characteristics of the intersection (to get MILITARY TIME

a 'feel" for traffic movements and interactions)
Complete upper portion of conflicts and volume forms (location, 12:00 Midnight 0000 Hours 12:00 Noon 1200 Hours
description, etc.) (One sheet for each approach (leg) observed, 1:00 0100 1:00 1300
i.e., 2 conflict forms and 2 traffic volume forms per location) 2:00 0200 2:00 1400
"Zero" count board 3:00 0300 3:00 1500
4:00 0400 4:00 1600
0715 S cart Conflicts observation 5:00 0500 5:00 1700
0730 Record.conflicts; "zero count board 6:00 0600 6:00 1800
Count traffic volume for 5 minutes (make sure to record starting time) 7:00 0700 7:00 1900
Record Traffic volume 8:00 0800 8:00 2000
Move to opposite approach 9:00 0900 9:00 2100
0745 Start conflict observation, traffic volume count, etc. 10:00 1000 10:00 2200
Move to opposite approach 11:00 1100 11:00 2300
0815 Start conflicts observation, etc.
Continue throughout the work day Reporting Procedures

Note: A 5 minute deviation is allowed in Count start timas, but conflicts If unable to report to work on any given day
must be observed for 15 minutes each period and traffic volumes Telephone designated supervisor before 0600 hrs of that day
must be counted for 5 minutes each period. Give name
Give intersection number and name
Figure F-5. Detailed, daily work schedule. State whether or not you will be working the following day

Figure F-6. Field needs.

 53

' £-

2. US 40-LEE5 SUMMIT RD.,

CONFLICTS FORM I

U)
:3

IL

LEE'5 SUMMIT RD.

SERVICE
STATION

IIMut
0 4.5
KANSAS
CITY

Figure F-7. Typical intersection diagrams—site 2.

all observers. Any general statements concerning policy The first two phases (6 weeks) of the study were rela-
and procedures were included in the newsletters, as well as tively free of operational problems, except for occasional
other items of interest. Two newsletters are shown in illness or car trouble. One day was rained out, but this day
Figures F-13 and F-14. was made up using extra observers. Initially, some of the
observers were incorrectly completing the supplementary
CONDUCT OF THE FIELD STUDIES information blocks on the conflict count sheets. Typical
The traffic conflict field studies began on June 12, 1978 examples were wrong location numbers and leg numbers.
and ended (formally) on August 10, 1978. The basic study The most troublesome problem was when two observers
framework consisted of observing conflicts at 24 intersec- incorrectly indicated they were at the same approach at the
tions during three separate 3-week observation phases. same time. This happened about five times. When this
Figure F-iS is the basic observation schedule for the first occurred, each observer was asked which approach he had
3 weeks. It covers the assignments of the 12 regular ob- started on that day. Usually, one or both of the observers
servers. The schedules for the other 6 weeks were similar. would remember. If neither did, other data for that inter-
 UNSIGNALIZED 4-LEG INTERSECTION CONFLICTS
MRI N2
8
Location Number Description l.eg Number
Day Dote ________________________ Time Period Observer, No. 7
Opportunity -0, Conflict - C. Paired Vehicle - PV. Secondary Conflict - SC. Severe Conflict -®
5

OUN
ItART - ---

MEN
___MEMO
mom .................UU.....-.-...
SME
ANN
ANN
ANN
ANN
MME 0MErME ••U••• ..U......................
-•••a••
MEN
'otes and Lomments on Bock of Form
••••••••••••••••••••••

Figure F-8. Conflict form; unsignalized, 4-leg intersections.

uIIIIIIIIIIII
LOC L D DATE T OBS DUR
Midwest Research Institute Project 4456D. Form I
 SIGNALIZED 4-LEG INTERSECTION CONFLICTS
MRI
Location Number Description Leg Number

Day Date ________________________ Time Period ______________________ Observer. No.

Opportunity -0. Conflict - C, Paired Vehicle - PV. Secondary Conflict - SC, Severe Conflict -c

HI I.
Turn

1
.
STAr

MEN
___..U....UE..U..U.a..IU.aUUNURUI•a
___•uu•uu•u•uaumu•uuM•ua•RmUUUUa
___•••••••••••••••••••••••••U••u••U•
___•UUURU••U••lU•NUilUNUUU•NUU••l
___••R..••uUu•uu••uuUUUUU•RUi
___•U•iUU•RUlUUlUUU.0
ANN U MINIMEMEM
.UUR.
___a•ai•••suuuuu•s•uuu••mi•u••ii•
___•auuau•uu•uu•uuui•RNiUUUaa.0u.SU.
___UUUU•SU•RlU•I•UU•RU•R•iUR•lU
IMMINIMMINIM UURUUUUUUUUUUUU•UU
___•IUI•UU•I•RUNU•
___••auU..U.•••••I••lU•U.lNNa..U.
ENINNE
--.....u....•.•••.••••-•••••..•••
Notes and Comments on Bøck of Form

Figure F-9. Conflict form; signalized, 4-leg intersections.

[HILIIIIHII
LOC . 0 DATE T OBS OUR
Midwest Research Institute Prolect 4456D, Form II
 3—LEG INTERSECTION CONFLICTS (LEFT LEG
MRI N2
Lfl

8
Location Number Description Leg Number
Day Date ______________________ Time Period
Opportunity - 0, Conflict - C, Paired Vehicle - PV. Secondary Conflict - SC, Severe Conflict -
Observer, No._______________
73
7 5
It

COUNT
START
TIME

m•u•u••uuu••u•uuuuu••uuuu•u•u•

...........-..............-...
••••••••••••u•••••••••••••••••
••••••••••••u••••••••••u••••••••

...-..........................-
...u...........-................
UUU•U•UUl•UN••UU••UU•U••U•

..........u.......m.............
...............u......-.........
UU•U•••UU••U••••UU••SUU••UU••

Notes and Comments on Back of Form

Figure F-JO. Conflict form; 3-leg intersections (left leg).

IIIHIIIIIIIII
LOC L D DATE T OBS DUR
Midwest Research Institute Project 4456D. Form Ill
 3-LEG INTERSECTION CONFLICTS (RIGHT LEG)
MRI 8 N 2 Lit

Location Number Dewription Leg Number

73
Day Date ______________________ Time Penod Observer, No._______________
Right
Opportunity - 0, Conflict - C. Paired Vehicle - PV. Secondary Conflict - SC, Severe Conflict -<)
5

COUNT
START
TIME
MENEM
U••U••U•••UU•••iURUi•UU••UUU
u............—..................
___•u••uuuu••u•uamuu•u•u•u••••uuu
___...........................--.
___•U•••••U•U•UU••UUURUUUUUUUUU
___•...............u.u...u......a.
-
•••uuuuuu•••u•mam•u••u•uuuuuu•u•

___•NUUU•l•UUUUNl•U•UUN••U••Ui
___••••••••U•••i•R•UUN•U•U••RU••

--.................................
___•U.-......U....U.......-..U....
___•ui•ii•••uu•uuu••iu•ui•••uumun
Notes and Comments on Back of Form

Figure F11. Conflict form; 3-leg intersections (right leg).

LIIIHIHIHI
LOc L D DATE T OBS DUR
Midwest Research Institute Project 4456D; Form IV
MRIS
TRAFFIC VOLUME COUNTS 1 I
82
IlIllIHhlItHI
LOC L D DATE I OBS DUR __________
Location Number Description Leg Numbers_,._.
Day ______ __
Date ________ __
Time Period ________ Observer, No. __ 6 4
1®
Approach (Leg) Number Approach (Leg) Number
COUNT
START $
TIME (
(MIUTARY) I I

Total
Notes and Comments on back at torm Midwest Research Institute Project 4456 D, Form V
Figure F-12. Traffic volume Counts form.
 59

section were examined for similarities that would indicate THE CONFLICT OBSERVER
the starting position of each observer.
Vol. 1, No. 1 June 13, 1978
Occasionally, an observer would fail to record data for
a particular time period. To help alleviate these types of (An occasional publication to keep far-flung participants informed.)
problems, a procedure was instituted whereby each ob-
server was asked to check his partner's forms for any in- First day is complete, and all survived (Maybe a little sunburn)
correct or missing data. The partner then initialled the
We're having some difficulty 'rotating' the count boards. Hence-
checked forms. This procedure was instituted in the third forth, we will assign these for several days at a time, according to
week of data collection and eliminated many of these prob- major needs. Everybody please bring these to the picnic on Thursday,
lems, which were not serious, but were time-consuming to for the next assignnnt.

correct. Did you hear about the radio interviews? KCMO (Radio 81) interviewed
Occasionally, the observers had to be reminded of cor- Coleman, Riscoe, and Glauz Monday morning, and broadcast it during
the day. You might be prepared for more of the same. People have
rect observation technique procedures including count start noticed you.
times and conflict definitions. Definitions that had to be
reviewed included conflict opportunities at signalized inter- A reminder. At signalized intersections, watch for red light viola-
tions and record these as "opportunities" of the appropriate type.
sections, red-light violations, and right-turn-from-left op- (Conflicts, if they occur.)
portunities. This sort of review probably would also be
Several have seen 'serious" conflicts. Let us know of any unusual
needed from time to time in an operational setting. happenings, so we can share these.
Although local governments were contacted during the
final site selection process to confirm that no planned con- Some confusion? Although you should report to your sites at 0700
hours and 1315 hours, first counts do not begin until 0715 hours
struction activities would interfere with the field studies, and 1330 hours. It will help if we can be rather consistent about
two unexpected situations arose during the field studies that this.

required changes in the data collection plan. In the first, Don't forget the picnic Thursday noon.
routine maintenance of a bridge located four blocks from
a test site uncovered severe deterioration of the structure, Flash First accident: 7:35 a.m. at 150 Highway and State Line.
Patti and John were witnesses.
necessitating its closure for extensive repairs. This closure
drastically changed the traffic patterns along that road. Be- Figure F-13. Sample newsletter No. 1.

THE CONFLICT OBSERVER

Vol. 1, No. 2 June 15, 1978

(An occasional publication to keep far-flung participants informed.)

To call MA! after (or before) normal working hours, try 753-7604 or 753-7605
if 753-7600 does not work. Person who answers can transfer call to Clauz
(x467), Migletz (573) or others.

Jsff is in the hospital today having knee surgery (minor, we hope). He ex-
pects to be back on the job Monday.

Miriam has, among other things, reviewed all the data for the first two days.
It looks quite good. A few minor slipups, such as missing names, dates,
times, sites, etc. Won't mention any names, but the numbers are 01, 07, 08,
10, 12, 17, and 18. -

Speaking of data, we are collecting over 100 pages of it each day: Enough
to make any statistician drool.

We've heard about several morei

'serius conflicts. Rick saw one involving
four vehicles: •- :.

There has been one last-minute schedule change, so far. There will probably
be more. Eric and Jeff went Out to 75th and Switzer, as assigned, only to
find the road under construction. Cleverly, they called in and got another
assignment.

Joan was the first victim of car troubles. Had a flat on the way to the
site. Was able to get there without much lost time, however.

We will have a visitor next week (Tuesday). Mr. David Witheford from
Washington, D.C., who monitors this project, will be here to observe field
activities. We will probably get Out to several different sites.

Figure F-14. Sample newsletter No. 2.

60

SITE ASSIGNMENTS Observer visibility is a legitimate concern. The data col-

Observer Number
lection procedures were similar to those used by the vari-
ous transportation departments visited before the start of
14 18 18 23 23
the conflict observer training program. Of particular im-
Monday AM 5 5 4 4 11. 11 14
6/12 PM 20 3 20 3 6 10 6 10 15 21 15 21 portance in these procedures was the location of the con-
1 8 8 1 10 15 15 10
flict observer relative to the intersection. Ideally, each ob-
Tuesday AM 19 22 22 19
6/13 PM 14 2 7 11 14 11 17 24 17 24 2 7 server should be hidden from the view of upstream traffic,
or at least be in a position that attracts very little attention.
Wednesday AM 9 2 7 16 16 9 20 21 2 7 21 20
6/14 NM 8 12 18 23 12 1 8 13 13 23 18 1 But, in these experiments (and in real world, operational
Thursday AM 3 17 6 13 6 12 24 3 12 17 24 13
application), some of the test sites provided good observa-
6/15 PM 22 16 19 5 9 16 19 4 22 9 5 4 tion positions, while others did not.
Monday AM 21 20 3 6 15 10 20 6 3 21 10 15 Photographs of conflict observers were taken at many
6/19 NM 19 22 2 7 13 12 7 2 13 12 19 22 test sites to show examples of locations with good and bad
Tuesday AM 14 7 24 2 17 24 11 7 2 11 17 14 observation positions. The photographs were taken at both
6/20 NM 9 9 16 16 18 18 23 23 3 3 6 6 urban and rural locations. A number are shown in the fol-
Wednesday AM 8 13 8 13 12 18 12 18 23 1 23 1 lowing figures. Before discussing the photographs, a few
6/21 NM 5 4 4 5 10 15 15 10 17 24 24 17 points should be made. During the summer, the tempera-
Thursday AM 19 9 16 22 19 22 5 4 5 4 9 16 ture in Kansas City can easily reach and exceed 100 F
6/22 PM 21 11 14 20 20 21 8 1 11 14 1 8 (38 C). Because of this, a prime concern of the observers
Monday AM 12 22 2 7 22 13 12 19 19 7 2 13 was to find an observation location that would not only
6/26 PM 11 5 14 18 14 23 4 11 23 5 4 18 satisfy data collection requirements but would also pro-
Tuesday AM 6 23 3 9 18 23 3 16 6 18 9 16 vide some shade. If possible, they preferred not to sit in
6/27 PM 1 8 10 15 22 19 8 15 10 1 19 22 their cars, but in chairs located as discreetly and safely as
Wednesday AM 5 4 10 15 17 24 15 10 17 24 5 4 practical. Some locations did not provide very good "hid-
6/28 PM 21 20 2 7 9 2 20 16 7 16 9 21 ing" places; and at such locations the observers sitting in
Thursday AM 21 21 20 20 1 1 8 8 11 11 14 14 lawn chairs with count boards did attract some attention.
6/29 PM 12 17 12 17 24 3 24 3 6 13 6 13 Figures F-16 through F-21 are photograhs (unposed) of
Figure F-IS. Site assign,nents for first 3 weeks. the observers at six intersections.
Figure F-16 shows two observation positions on a typi-
cal rural highway, where conflicts were observed from a
car. Figure F-17 shows observers at a rural intersection of
2-lane highways with some roadside development. The ob-
server on the west leg is exposed but relatively inconspicu-
ous, whereas the one on the east leg is on the service station
property.
Figures F-18 and F-19 show intersection locations in the
cause of this, that test site was abandoned and another was urban environment with partly or mostly industrial or busi-
chosen in its place. The new site, number 29, was substi- ness development. The observer locations are quite well
tuted for the site number 10. The new site had all of the hidden in Figure F-18. The observer on the west leg is
criteria needed to fit that classification. Fortunately, the behind the billboard and the one on the east leg is behind
bridge was not closed until the end of the sixth week. the large shrub. (The latter is very visible, however, to
Therefore, two phases had been completed by that time. motorists who have passed the observer but are, for exam-
The other unexpected situation occurred during the fi- ple, stopped by the traffic signal.) An unusual, but accept-
nal 3-week phase. The local gas company had to repair a able, situation is depicted in Figure F-19. At this par-
gas line near site number 8. Although all construction was ticular location the normal observation position is totally
performed off the roadway, a lane was closed to provide exposed and only 8 to 10 ft from the travelled way. On the
additional working room. The workers stated that the work north side of the street, howe,'er, a hill offered an excellent
would take about a week to complete. However, the gas vantage point and was unobtrusive.
company personnel did not work every day, and gave no
Figure F-20 shows a typical residential setting. On the
indication of which days they would be working. Substi-
south leg the observer is visible but well off the roadway.
tute sites had to be scheduled in case construction was
On the north leg the observer (behind the tree) is not visi-
being performed. The repairs took over 3 weeks.
ble to the motorists until they pass the tree.
Also during the final 3-week period, two rain days oc-
curred in a row. Very little data were collected during this Figure F-21 depicts a "worst" condition. This busy
time. Because of the rain days and the gas-line repair, the intersection is near the downtown area, by Kansas City's
entire schedule had to be changed. After the scheduled Crown Center complex. The observer on the north leg,
data collection period was complete, a few of the observers particularly, is totally exposed, a situation necessitated by
worked as much as two additional weeks to make up lost the lack of suitable surroundings. Again, the operational
observation time. The final 3-week segment of the study application of the technique must allow for such locations
was completed, although not as planned. because they exist in the typical, real world.
 61

44c

a. South Leg

LIMIT

50

Figure F-16. Observer locations at site 1.

62

4- T i Ij1I 1I1Ifl

FINA

LIN

'-• .-- c

a. West Leg

- -,-- -

I
u. East Leg
Figure F-I?. Observer locations at site /9.
 63

a. West Leg

. L

>:L

b. East Leg
Figure F-18. Observer locations at site 12.
64

a. West Leg

b. East Leg
Figure F-19. Observer locations at site 29.
 65

North Leg

-
S
..
ur
S.
-

South Leg
Figure F-20. Observer locations at site 22.
66

4 '4

4 p

a. North Leg

: '

b. South Leg
Figure F-21. Observer stations at site 16.
 67

APPENDIX G
DATA ANALYSIS PROCESS AND RESULTS

This appendix consists of 14 parts. The first part de- reviewed by the principal investigator or his assistant for
scribes the generalprocess of the data analysis in non- notes, special occurrences, problems, and needs for special
statistical terms. The process involved the effectively si- handling. After this checking, the books (one book for
multaneous examinations of many conflict definitions and each day of operations) were transferred to the MRI Com-
groupings, which are given in Part 2. The formal statisti- putation Center, where the data were keypunched and veri-
cal models are given in Part 3, and the application of these fied. The punched cards were sorted by site and leg and
models using data manipulations and transformations is in listed for a comparative review. This review process was
Part 4. aimed at detecting additional coding errors that might be-
In Part 5 some of the conflict types that can readily be come obvious by noting marked differences between counts
dismissed from further consideration are presented, and the of several observers at the same site and leg. Typical er-
decisions justified. A particular concept, severe conflicts, is rors detected were incorrect coding of site number, leg
dealt with at greater length in Part 6. number, and (occasionally) of certain conflict types.
The major focus of the project, the determination of ob- An example of a portion of a listing is shown in Figure
server reliability and repeatability, is the subject of Part 7. G-1. In this figure is displayed a sample of the data from
This discussion is continued into Part 8, which deals spe- several days of observations at sites 16 and 17. The first
cifically with the "extra" observers. (left) portion of the data is the heading information (site,
Part 9 examines the various factors descriptive of the leg, date, observer, time, etc.) coded on each sheet. The
intersections in the main experiment and how they are rest of the data are the 15-min conflict counts in the same
related to the conflict counts. Then, the individual sites, order as written on the data sheets, two digits per count.
themselves, are compared in Part 10, and some observa- The 5-min traffic volume data are in a similar format,
tions are made. Part 11 deals, in a similar way, with the except allowance is made for 3-digit counts.
extra sites—those not a part of the main experiment. This step completed the initial process of data checking.
The traffic volume data are given in Part 12. (They are The subsequent analyses can (and did) occasionally un-
utilized in Parts 10, 11, 13, and 14.) Part 13 examines the cover additional errors—through statistical tabulations of
time-of-day trends and their implications. Finally, a cur- frequency distributions, means, variances, etc.
sory look at related accident data is included in Part 14. The statistical analyses, themselves, were of a varied na-
ture. Clearly, there was a tremendous amount of data, and
PART 1—THE ANALYSIS PROCESS the opportunities for analysis were almost unlimited. The
4,000-plus data sheets included over 8,000 15-min conflict
Bridging the gap between the recording of data in the counts and a like number of volume counts, distributed
field and the attainment of meaningful interpretations of among 17 observers, 36-plus days, and 29 sites. The con-
the experiments was a multiple step process. This process flict counts are in as many as 30 to 40 categories, which can
began with the collection and review of the data, and con- (and should) be combined in numerous ways for mean-
tinued with encoding, error checking, data processing, and ingful analysis and interpretation. Thus, it was necessary
statistical analysis. These steps are described in the to structure the analysis process to ensure that efforts were
following. directed most efficiently.
The recording forms were collected nearly every day to The structuring was guided by the basic objectives of the
facilitate early detection and correction of errors. The data project, which can be restated as:
sheets were first scanned for completeness in the field, by
each observer's partner, and initialed. (This step was ini- Develop standardized procedures.
tiated part way through the field tests and did help in re- Develop cost-effective procedures.
ducing certain types of coding errors and oversights.) The Minimize interobserver variance (reliability).
data sheets also received a quick check by the collecting Minimize an observer's variance at a site (repeat-
supervisor to ascertain that no sheets were missing and to ability).
learn of any special problems. Then, in the office, all data Optimize procedural ability to diagnose intersection
sheets were sorted and logged in by observer and time of operations/safety.
day. Typically, about 120 pages of data were recorded and Optimize procedural ability to evaluate intersection
collected each workday. By the end of the field tests the improvements.
total number of data sheets exceeded 4,000. Consideration of the project objectives and contempla-
The sorted data sheets were placed in notebooks, and all tion of the data structure as well as the practical applica-
"heading" information was encoded for keypunching (see tions lead to several basic guidelines, including the
App. F for sample data forms). The sheets were again following:
68

416130802203151630 I 1 1 4 22 1
416130802203151730 1 1 1 1 4 1 1
416130802205151330 1 1 3 1
416130802205151430 1 2
416130802205151600 2 1 2
416130802205151700 1 2 1 1 1 3
416130809203151400 1 1 1
416130809203151500 1 2 1
416130809203151630 2 3 1 1 2 1 1 1
416130809203151730 2 1 1 3
416130809205151330 1
416130809205151430 1 1 1
416130809205151600 1 2
416130809205151700 2 1 1 2
316540803108150116 1 12 .99 2 113
316540803108150815 2 3 99 1 19 1 1 C
316540803108150945 1 3 36 1 12
316540803108151045 1 1 2 59 3 114 1
316540803118150745 1 63 2 19
316540803118150845 1 1 55 1 19
316540803118151015 3 34 17
316540803118151115 1 3 32 14
316530802203151330 1 1 2 17 2 15
316530802203151430 4 24 19 1
316530802203151602 1 2 23 12
316530802203151700 1 3 21 1 9
316530802205151400 1 36 2 11
316530802205151500 28 1 14 1
316530802205151630 28 9
316530802205151730 18 24
316530809203151330 1 14 26 20
316530809203151430 1 23 28
316530809203151600 1 1 2 29 21
316530809203151705 1 2 1 2 39 33
316530809205151400 1 21 17
316530809205151500 1 25 3 10
316530809205151630 1 30 5
111120808104150745 32 1 3 13 17 9 19 6 6
117120808104150845 21 1 4 2 10 13 2 5 6 3 3
117120808104151015 2 1 9 7 3 6 3 5
117120808104151115 2 1 1 1 11 II 5 IS 1 1
117120808106150715 3 1 3 1 3 13 9 19 5 5
117120808106150815 1 1 1 1 3 11 3 7 4 '
117120808106150945 2 2 2 1 8 14 4 12 2 2
117120808106151045 1 4. 11 3 9 5 3
117140803206151330 2 2 1 1 11 9 4 7 2 4
117140803206151430 3 1 1 10 13 11 13 6 5
117140803206151600 4 3 6 3 1 9 18 5 1 24 1 5 9 1
117140803206151700 2 2 1 5 12 7 6 5
117140803211151400 1 22 1 7 11 4 7 2 4
117140803211151500 21 7 13 9 8 4 2
117140803211151630 21 2 3 6 15 6 11 1 3
117140803211151730 1 15 2 15 5 3 6 3
117520808104150715 44 2 1 2 3 1 17 8 4 4

Figure G-1. Sample of listed conflict data.

All reliability and repeatability analyses must be con- conducted to achieve meaningful diagnoses or to identify
ducted in light of practical feasibility. This means that the the effectiveness of appropriate countermeasures.
analysis must be done in consideration of some reasonable Each stage of the procedure attempted to collapse the
maximum observation period. (This also raises the ques- total data set (types of conflicts or conflict definitions) for
tion of needed precision (e.g., 50 percent or 75 percent obtaining better measures of reliability and repeatibility.
vis-a-vis the 95 percent or 99 percent levels more com- The collapsing was checked for sensitivity to (1) site
monly employed by statisticians). parameters (lanes, legs, speeds, traffic volumes) and (2)
A very possible result of the study is that conflicts identifying potential countermeasures. In other words, the
may only be good for diagnosing certain kinds of problems final categories could depend on the kind of intersection.
(or identifying certain kinds of countermeasures) at cer-
tain kinds of intersections, under certain conditions, rather Stage 1
than being a universal tool.
Conflict categories will probably be collapsed or ag- Check for generally unreliable or inefficient conflict cate-
gregated, but observation categories and analysis cate- gories and either eliminate or combine with other appro-
gories need not be collapsed in the same manner. There priate categories. These are categories of conflicts where
may not be any advantage to combining or separating cer- occurrences are so infrequent as to be worthless in a diag-
tain categories for observation, but there may be great ad- nostic procedure. Possibilities include accidents (observed);
vantages to combining certain categories in various ways for unusual conflict types such as pedestrian conflicts, red-light
different analysis purposes (e.g., left-turn, same-direction violations, etc.; severe conflicts; right-turn-on-red conflicts;
leg 1 with opposing left-turn leg 5). lane-change conflicts; and several types of secondary
With these points in mind, the analysis was conducted in conflicts.
several sequential stages. The stages were, aimed at arriv- This stage was conducted manually (i.e., without formal
ing at optimum conflict categories that yield most reliable statistical analysis), by scanning data sheets and printouts,
and repeatable results from cost-effective conflict studies determining which of the foregoing possibilities (by site)
 69

show little promise of being cost effective. However, loss (e.g., "paired-vehicle" and "other" same-direction conflicts;
of sensitivity for diagnosis was also checked to identify opportunities and actual conflicts for cross-traffic move-
appropriate countermeasures, and categories were re- ments; and secondary conflicts corresponding to each type
adjusted if necessary. of instigating conflict). These 36 types are itemized in
Table G-1, along with the many collapsed categories
Stage 2 described next.
Categories 38 through 40 combine "paired-vehicle" and
Conduct reliability and repeatability analyses on remain- "other" same-direction conflicts. (Category 37 was not
ing conflict count categories and record results. This was used.) For example, if Y1 identifies counts in the first cate-
done via computer as outlined subsequently. gory, Y2 the counts in the second category, and so on, then
Y38= Y1 + Y 2 (G-la)
Stage 3
Y39 =Y4 +Y5 (G-lb)
Collapse categories where reliability and/or repeatability Y40 = Y7 + Y8 (0-Ic)
measures are not strong. Possibilities include those previ-
ously mentioned; combining paired-vehicle with other like, All rear-end categories (not paired) were collapsed into
same-direction conflicts; combining all or some different, Y41, all rear-end paired-vehicle categories into Y42, and all
same-direction conflicts; and combining some or all cross- rear-end secondary conflicts into Y43. That is,
traffic counts. Improved reliability or repeatability of the Y41 =Y1 +Y4 +Y7 (G-2a)
collapsed categories was also examined. On the basis of Y42 = Y2 + Y5 + Y8 (G-2b)
these analyses, an optimum category set was reconstituted Y43 =Y3 +Y 3 +Y9 (G-20
based on the first three stages.
These were further combined as:
Stage 4 Y44 = Y41 + Y42 (G-3a)
Y45 = Y41 + Y42 + Y43 (G-3b)
Reconstitute categories specifically as they relate to
countermeasure diagnosis (these will not be mutually ex- Moreover, these were combined with the corresponding
clusive nor necessarily the same groupings as derived in lane-change conflicts to yield the following:
stage 3). An example is to combine left-turn, same-
Y46 = Y44 + Y11 (G-4a)
direction conflicts with opposing left-turn conflicts viewed
Y47 = Y43 + Y 2 (G-4b)
from the opposite leg, in diagnosing need for a left-turn
Y48 = Y45 + Y]1 + Y12 (G-4c)
bay.
Similar combinations were created for cross-traffic conflicts
Stage 5 (see Table G-l). Thus,
Finalize optimum analysis category sets which may vary Y49 = Y18 + Y22 + Y28 (G-Sa)
by intersection types. Also, finalize optimum observation Y50 = Y17 + Y23 + Y29 (0-5b)
category sets which may also vary by intersection types. Y51 = Y)8 + YN + Y3() (G-5c)
These need not be the same sets, provided the analysis Y52 = Y50 + Y51 (0-5d)
categories can be derived from the observation categories. Y53 = Y10 + Y25 + Y31 (0-5e)
Y54 = Y20 + Y2 + Y32 (0-5f)
Stage 6 = Y21 + Y 27 + Y33 (G-5g)
Y56 = Y54 + Y55 (G-5h)
Determine expected values of the category sets as related
to site parameters. Then, rather major groupings were created. Category 57
consisted of all opportunities:
PART 2—DESCRIPTION OF CONFLICT CATEGORIES Y57 = Y10 + Y15 + Y10 + Y19 + Y 22 + Y25
+Y+Y31 +Y34 (0-6)
The operational definitions of the traffic conflicts ob-
served in the field are given in Appendix E. The forms on Category 58 contained all conflicts (other than secondary),
which conflict counts meeting these definitions were re- and category 59 was all the secondary conflicts:
corded are in Appendix F. This portion of Appendix G
Yss = Y1 + Y5 + Y4 + Y5 + Y 7 + Y8 + Y11
summarizes these definitions from the analysis viewpoint,
+ Y14 + Y17 + Y 29 + Y 23 + Y2
together with derived (collapsed) categories developed
+ Y59 + Y 95 + .Y35 (G-7a)
during the multistage process previously described.
= Y3 + Y6 + Y9 + Y12 + Y15 + Y18
There were 36 different types of traffic conflicts record-
+ Y51 + Y 54 + Y 27 + Y30 + Y33 + Y 30 (0-7b)
able during the field experiments. They can be grouped in
various ways. From one perspective, they can be divided Also, these were all combined into a single, total conflicts
into rear-end (or same-direction) conflicts, opposing con- category:
flicts, cross-traffic conflicts, and special conflicts. On the (0-8)
Yos = rS + Yso
other hand, the basic movements leading to particular con-
flict types can be subdivided, as described in Appendix E Finally, slow-for-left-turn and opposing left-turn cate-
TABLE 0-1
CONFLICT CATEGORIES
1 Left Turn Same Direction Conflict 36 Right Turn on Red Secondary Conflict

37 Not Used
2 Left Turn Same Direction Paired Vehicle Conflict

3 Left Turn Same Direction Secondary Conflict 38 Left Turn Same Direction Conflict Plus Paired Vehicle

Conflict
4 Right Turn Same Direction Conflict

5 Right Turn Same Direction Paired Vehicle Conflict 39 Right Turn Same Direction Conflict Plus Paired Vehicle

Conflict
6 Right Turn Same Direction Secondary Conflict

7 Slow Vehicle Conflict 40 Slow Vehicle Conflict Plus Paired Vehicle Conflict

41 Rear-End Conflict
8 Slow Vehicle Paired Vehicle Conflict
42 Rear-End Paired Vehicle Conflict
9 Slow Vehicle Secondary Conflict
43 Rear-End Secondary Conflict
10 Lane Change Opportunity
44 Rear-End Conflict Plus Paired Vehicle Conflict
11 Lane Change Conflict
45 Rear-End Conflict, Pairod Vehicle and Secondary Conflict
12 Lane Change Secondary Conflict
46 Same Direction Conflict Plus Paired Vehicle Conflict
13 Opposing Left Turn Opportunity
47 Same Direction Secondary Conflict
14 Opposing Left Turn Conflict
48 Same Direction Conflict, Paired Vehicle and Secondary
15 Opposing Left Turn Secondary Conflict
Conflict
16 Right Turn from Right Opportunity
49 All Cross Traffic from Right Opportunities
17 Right Turn from Right Conflict
50 All Cross Traffic from Right Conflicts
18 Right Turn from Right Secondary Conflict
51 All Cross Traffic from Right Secondary Conflicts
19 Left Turn from Left opportunity
52 All Cross Traffic from Right Conflicts Plus Secondary
20 Left Turn from Left Conflict
Conflicts
21 Left Turn from Left Secondary Conflict
53 All Cross Traffic from Left Opportunities
22 Left Turn from Right Opportunity
54 All Cross Traffic from Left Conflicts
23 Left Turn from Right Conflict
55 All Cross Traffic from Left Secondary Conflicts
24 Left Turn from Right Secondary Conflict
56 All Cross Traffic from Left Conflicts Plus Secondary
25 Cross Traffic from Left Opportunity
Conflicts
26 Cross Traffic from Left Conflict
57 All Opportunities
27 Cross Traffic from Left Secondary Conflict
58 All Conflicts Plus Paired Vehicle Conflicts
28 Cross Traffic from Right Opportunity
59 All Secondary Conflicts
29 Cross Traffic from Right Conflict
60 All Conflicts, Paired Vehicle and Secondary Conflicts
30 Cross Traffic from Right Secondary Conflict
61 Slow for Left Turn and Opposing Left Turn Conflicts
31 Right Turn from Left Opportunity
Plus Paired Vehicle Conflicts
32 Right Turn from Left Conflict
62 Slow for Left Turn and Opposing Left Turn Secondary
33 Right Turn from Left Secondary Conflict
Conflicts
34 Right Turn on Red Opportunity
63 Slow for Left Turn and Opposing Left Turn Conflicts,
35 Right Turn on Red Conflict
Paired Vehicle and Secondary Conflicts
 71

gories were combined to yield: was used. This program obtained 0-2 and 0-2 correctly and
automatically. To obtain the other variances, dummy vari-
Y01 = Y1 + Y2 + Y14 (G-9a)
ables were introduced into the multiple regression equa-
Y62 = Y3 + Y15 (G-9b)
tion. For example, consider the computation of the ob-
Y63 = Y61 + Y62 (G-9c)
server variance (reliability). One observer was arbitrarily
selected as a reference (it is immaterial to the final results
PART 3—STATISTICAL MODELS as to which one is the reference). To note the effect of a
Fundamentally, the data were viewed from an analysis second observer, a dummy variable, X2 , was introduced.
of variance (AOV) framework. The total variance, .,,2, of It has the value, 1, if the particular conflict count was ob-
all the 15-min counts of a given conflict category, Y, was tained by observer 2, and is zero otherwise. Also, a dummy
represented as the sum of the variances attributable to iden- variable, X2, was introduced to identify counts of observer
tified (and controlled) variables plus an error term, 0 2, 3, etc. Then, the regression equation takes the form,
interpreted as the practical repeatability. Mathematically, Y = 82X2 + $1 X3 + $4 X4 +. . . + $12X12 +.
cr112 =
cr02 + °t2 + cr02 +
0_N 2 O•9 + 0.12
2
+ + op + UR + cre 2 (G-10) where succeeding terms account for the other design fac-
tors. The values of $2 - 18121 which are calculated by the
where:
program, represent the average difference between the
= observer variance (reliability)—the variation due counts of the corresponding observers and the first (ref-
to systematic biases between observers; erence) observer.
= the variance between short (15-mm) time inter- Having obtained the $'s, they can be used in a second
vals at a site; (manual) step to obtain the variance. Specifically,
oD2 = the variance between days of week at a site. 12
0v2 = the variance between 3-leg and 4-leg sites; 0.2={12 $.2_ ($.)2
12 }/12. 1I-12- 2/N
= the variance between low-speed and high-speed
sites;
o-; = the variance between 2-lane vs. 4-lane unsignal- where N is the total number of 15-min counting intervals
ized intersections. included in the analysis.
0-,2 = the variance between legs at a site; Simultaneously, a totally analogous procedure was used
for the factor effects such as time of day, day of week, num-
0_ c2 = the variance between signalized and unsignalized
ber of legs, etc. That is, other dummy variables were intro-
traffic control at 4-lane intersections; duced into the regression equation (X's) and the corre-
= the variance between "replicate" sites of nominally sponding $'s were obtained. Then, the variances were
the same type (same speed, number of lanes, and calculated from the $'s (although some of these variances
traffic control); and were, indeed, calculated, often the $'s themselves and their
o2 = residual variance, or error, which is the "repeat- levels of significance were the quantities of interest).
ability" sought by the project. This is the variance
of repeated observations by the same observer PART 4—DATA MANIPULATIONS AND
under (theoretically) identical conditions (same TRANSFORMATIONS
physical site, same time of day and day of week,
etc.). Thus, 0-02 describes the variation in a count The experimental design (App. F) and the analysis
not attributable to known factors. It results be- process are written, for simplicity, as if each type of con-
cause (a) conflict counts actually differ from week flict category (the Y's) is observable at every intersection.
to week as traffic conditions are not really identi- In fact, some conflict categories, by definition, can occur
cal, and (b) observers make errors in counting at only a subset of the sites. Some examples are: oppos-
conflicts. ing right-turn-on-the-red conflict can only occur at sig-
nalized intersections with a protected left-turn phase; most
For various reasons, the actual computations were not cross-traffic conflicts can occur only at unsignalized inter-
carried out in "textbook" fashion. The complexity of the sections (unless there is a signal violation); and conflicts
mathematical model and the need to perform essentially involving turning movements can occur only on one of the
the same analysis many times (i.e., for 63 conflict cate- two observed legs of a 34eg intersection.
gories—Y's—and for each of the three phases of the Because of these inherent limitations, most of the Y's
9-week experiment) precluded manual calculations. (A were analyzed by using an appropriate subset of the data
few preliminary hand calculations were made to help de- base, and with corresponding but modified regression equa-
termine the final model and analysis approach.) Recourse tions. Special care was also required in the process of col-
was made to computerized computations, using the Statis- lapsing counts into groups, so that only sensible combina-
tical Package for the Social Sciences (SPSS) software. tions (from the standpoint of the individual intersection
Because the SPSS AOV routine could not handle the characteristics) were created.
number of factors in this model, a two-step process was Another special manipulation was required to properly
used. First, the SPSS multiple linear regression program distinguish between legitimate zero values for some counts
72

and occasional missing data values. The latter were identi- nating those categories that were so infrequent that, by
fled manually and coded as —1, which could be recognized themselves, were not practical in diagnostic procedures.
as "missing data" by the SPSS software. First, a number of conflict categories were essentially
Mathematical transformations were also required. Con- "eliminated" before the field experiments began, based on
flict counts exhibit a variance that tends to change with the preliminary observations and professional judgment. They
mean. That is, counts with larger mean values tend to have were eliminated in the sense that, although the observers
larger variances. Most statistical techniques (and, particu- were trained to recognize them, their recording was on a
larly, AOV) require a constant variance. To meet this re- "write-in" basis; the recording form did not contain ex-
quirement, a variance-stabilizing transformation is required. plicit reference to them. A major example is pedestrian
Conflict count data were examined to determine their conflicts. Most of the intersections in the experiment car-
apparent statistical distribution. The data appear to be best ried only minor pedestrian traffic, and none were in the
fit by the negative binomial distribution, which can be con- CBD or near other areas of heavy pedestrian concentration.
sidered as a generalization of the Poisson distribution. This A related example is pedalcycle conflicts. Kansas City (and
finding is in agreement with much of the recent literature most of the United States) does not experience heavy
(Hauer (16)). The proper transformation for such data pedalcycle volumes, so only passing reference to these ve-
is (17): hicles was made in the training program. Just a few pe-
destrian (and only a couple of pedalcycle) conflicts were
= XI' ln{Vl + X 2 Y1 + XVY} (G-13) noted during the entire study, so they were not analyzed.
where X = var(Y)/Y1, for each count, j, in conflict cate- Certain other conflict categories, which were recorded
gory, i. routinely for some types of intersections, do not normally
All analyses were performed with both the transformed occur at others. Most cross-traffic conflicts cannot happen
and untransformed versions of the data. The analyses of at signalized intersections unless there is a signal violation.
the untransformed data provide proper estimates of the A few were noted, but they were not frequent enough to
magnitudes of the various quantities of interest (a-'s and justify analysis. Similarly, certain turning movements can-
$'s), and these are the data (which have direct physical not occur on the data forms unless by exception. They,
interpretations) that are tabulated in this report. The too, were relatively rare with one clear exception. One sig-
analyses of the transformed data provide proper estimates nalized T-intersection contained a commercial driveway,
of the significance of the effects. However, their numeri- which could be viewed geometrically as a fourth leg. The
cal values are not directly meaningful physically, so are not driveway traffic was not signal controlled nor was it par-
tabulated herein; only their level of significance is referred ticularly heavy. Nevertheless, most observers noted several
to. conflicts involving driveway vehicles, clearly indicating an
The field data were collected in three phases, as de- operational problem at this intersection.
scribed in Appendix F. The intent was to analyze each Another situation of a "write-in" nature was observed
phase independently, as a separate experiment, and then to accidents. Interestingly, during the approximately 4,000
perform special analyses to examine differences between observer-hours (which, because they worked in pairs,
phases. (Special analyses were also to be performed for amounted to 2,000 observation-hours), six accidents were
the extra observers, extra sites, etc., as noted previously.) witnessed. This is an average of one accident in 333 hr of
In actuality, only the first two phases were completely observation, clearly not enough to be of practical utility,
analyzed, primarily because of missing data problems. As but probably more than most would anticipate.
noted in Appendix F, data collection proceeded very There were 36 basic conflict categories recorded rou-
smoothly during the first phase, with few problems en- tinely during the field tests. Stage 1 analyses indicated that
countered and relatively little missing data. In the second some should be dropped as useful concepts, simply because
phase, quite a bit of data were missing because of rain. they are so rare as to be impractical observational mea-
Then, in the third phase, one site was lost completely sures. Table 0-2 documents these decisions. Essentially,
(bridge repair) and another was lost most of the time (gas the tabulated conflicts each occurred, at most, only about
company construction work). Moreover, rain caused the once for every 8 observer-hours of observation, equivalent
total loss of two consecutive days of data. Although much to about 2 workdays. This observation rate is not consid-
of the missing data (but not all) was ultimately collected ered useful in a practical sense.
on a make-up basis, they were usually obtained by alternate The two major exceptions to this 8-hr limitation in
observers or by regular observers on a schedule not meeting Table G-2 are the right-turn-from-left opportunities and
the original plan. Because of the great number of changes the right-turn-on-red opportunities. Further examination
all of these factors would impose on the analysis process, of these data indicated that, in addition to the relative
it was decided to use portions of these data to fill "holes" rarity of these events, the interobserver variance was un-
in the first and second phases, and retain the rest for the usually high. That is, in retrospect, the observers clearly
various special analyses. did not all interpret these events in the same way. A few
observers were responsible (probably erroneously) for the
PART 5—RARELY OCCURRING CONFLICT TYPES majority of the counts. The definitions of these events are
apparently difficult, conceptually, so probably should not
As stated earlier, the first stage of the analysis was an be recommended (especially because the events themselves
examination of the conflict counts for the purpose of elimi- are fairly rare).
 73

TABLE G-2
CONFLICT TYPES THAT OCCUR RARELY (17)

Total Observation Hours

Conflict Number and Type Observer-Hours Per Conflict

Lane Change Conflicts 554 6.4

Lane Change Secondary Conflicts 554 62.5
15. Opposing Left Turn Secondary Conflicts 420 13.5
18. Right Turn from Right Secondary Conflicts -420 11.6
21. Left Turn from Left Secondary Conflicts 282 8.3
24. Left Turn from Right Secondary Conflicts 282 11.2
27. Cross Traffic from Left Secondary Conflicts 190 23.8
Cross Traffic from Right Secondary Conflicts 190 15.9
Right Turn from Left Opportunities 236 4.1.J
Right Turn from Left Conflicts 236 33.3
Right Turn from Left Secondary Conflicts 236 250.0
Right Turn on Red Opportunities 47 3.OJ
Right Turn on Red Conflicts 47 9.4
Right Turn on Red Secondary Conflicts 47

a/ Based on first 3-week phase of observations.

b/ Most were observed by just a few observers, indicating poor reliability
as well as a low occurrence rate.

PART 6—SEVERE CONFLICTS TABLE 0-3

Another relatively rarely observed phenomenon was a SEVERE CONFLICTS BY SITE
severe conflict. However, because the concept of utilizing Signalization Speed Legs Severe Conflicts./
Site Lanes
serious conflicts is so compelling and, in fact, the formal
practice of many Europeans and some of the states (al- 1 4 No High 4 4
2 4 No High 4 4
though more informally, in the latter instance), special 3 4 No High 3 2*
attention and analyses were devoted to these events. 4 4 No High 3 5
5 4 No Low 4 4
Severe conflicts, defined basically by using the Swedish 6 4 No Low' 4 6
time-to-collision concept (see App. E) were recorded 7 4 No Low 3 . 1
8 4 No Low 3 7*
throughout the field experiments. A grand total of 104
9 4 Yes High 4 5
severe conflicts was noted, of which 96 occurred at the 10 4 Yes High 4 1
main 24 locations and 8 occurred at the extra sites. Severe 11 4 Yes High 3 2
12 4 Yes High 3 6
conflicts were relatively rare—they averaged about one per 13 4 Yes Low 4 2*
18 observer-hours of observation. 14 4 Yes Low 4 3
15 4 Yes Low 3 6*
The distribution of severe conflicts among sites is given 4 Yes Low 3 - 6
16
in Table G-3. (The asterisks identify the six accidents ob- 17 2 No High 4 4
4 6*
served, which are included in the severe conflicts counts.) 18 2 No High
19 2 No High 3 2*
Chi-square analyses show that there are no significant dif- 20 2 No High 3 3
ferences in the counts attributable to the type of site. 21 2 No Low 4 3
22 2 No Low 4 8
Tables 0-4 through G-6 indicate how the severe con- No Low 3 1
23 2
flicts were distributed among observers, time periods, and 24 2 No Low 3 5
days of the week. Chi-square analyses were performed on 25 2 Yes High 4 2
26 2 Yes High 4 1
these data, taking into account the relative frequencies of 27 2 Yes Low 4 4
observation (e.g., observers 13, 14, and 18 did conflict 28 2 Yes Low 4 1

counts on only about two-thirds as many days as the others, a/ Asterisks identify accidents
and observer 17 did only a few counts; counts were taken
on nine Tuesdays, Wednesdays, and Thursdays, but only
on six Mondays and three Fridays). The analyses showed
that there were significant differences among observers—
with numbers 4, 1, and 15 seeing too many. There were
significant time effects, with substantially more severe con- TABLE 0-4
flicts recorded in the afternoon and, especially, late after-
SEVERE CONFLICTS BY OBSERVER
noon periods. There were no significant differences by day
of week. Observer Conflicts Observer Conflicts Observer Conflicts
Severe conflicts were also examined to determine if they
4 13 5
were distributed among "types" in the same way as regu- 6 14 3
lar conflicts. For this purpose, four groupings were used: 3 3 9 4 15 13
4 15
rear-end or same-direction conflicts; opposing left-turn con-
flicts; cross-traffic-from-right conflicts; and cross-traffic- 6 7 - - - 12 5
74

TABLE 0-5 TABLE G-6

SEVERE CONFLICTS BY TIME OF DAY SEVERE CONFLICTS BY DAY
OF WEEK
Time Period Conflicts Time Period Conflicts
Conflicts
0700 - 0730 3 1330- 1400 5 per day of
0730 - 0800 6 1400- 1430 6 Day of Week Conflicts Observation
0800 - 0830 3 1430- 1500 9
0830 - 0900 5 1500- 1530 8 Monday 19 3.17
0930 - 1000 2 1600- 1630 8 Tuesday 33 3.67
1000 - 1030 4 1630- 1700 15 Wednesday 21 2.33
1030 - 1100 2 1700- 1730 13 Thursday 22 2.44
1100 - 1130 3 -- 1730 - 1800 12 Friday 9 3.00

from-left conflicts. The analysis showed that the distribu- Yl, observer differences account for only about 5 percent
tions were greatly different. Whereas about 83 percent of of the total variance, and the (lack of) repeatability ac-
all conflicts were of the same-direction variety, only 55 per- counts for most (84 percent) of the total. All other mea-
cent of the severe ones were of this type. Instead, the se- sured effects yield only 11 percent of the total variance.
vere conflicts were more likely to be of the cross-traffic or The pattern described for Yl holds true, in general, for
opposing left-turn variety—especially the latter, which the other conflict types—that is, most of the variance is due
comprised 18 percent of the severe conflicts as opposed to a lack of repeatability rather than to observer errors or
to 5 percent of the regular conflicts. other measured effects. All of the observer variances are
Finally, the distribution of the 96 severe conflicts among statistically significant except Y6 and Y26 (significance de-
the 24 basic sites (an average of four per site) was ex- termined by (0.2 + Na-,2 /n)/a-2 F09 (11, 00) where
amined. There is no pervasive evidence that the location is n = 12 (observers) and N = number of 15-min counts).
associated with severe conflict counts. The number per site It is emphasized that these numbers pertain only to the
ranged from 1 to 8 with a standard deviation of 2.02 (vari- 15-min counting periods used for convenience in the field
ance of severe conflicts per site of 4.09). If severe con- experiments. In actual practice, much longer periods would
flicts per site were Poisson, the collection of 24 site results be used, at least in the aggregate. The use of longer pe-
would be distributed: (1) Poisson, if the severe conflict riods would, basically, leave all variances unchanged ex-
rates were the same at all sites; (2) negative binomial, if cept 0-2, which would decrease as 1/n, where n is the
the severe conflict rate varied between sites. Since the length of the sampling period.
variance is equal to the mean, the collection appears more For example, the standard deviation due to both repeat-
like (1). More formally, the Poisson distribution is not ability and reliability (again, using Yl for illustration) is
rejected by a goodness-of-fit test (x2 (6) = 3.88).
= Vo. 2 + cTO2 = 1.7163 (G14)
PART 7—RELIABILITY AND REPEATABILITY for a 15-min observation period. However, if data are ag-
The measures of reliability and repeatability are the gregated into 4-hr observation periods (16 times as long as
variances, and o52, or quantities derived therefrom. the samples), then
Table G-7 displays these measures, along with the mean = V 0 2 / 16 a-2 (G-15)
+ = 0.5789
values obtained, for the first 3-week phase of the experi-
ment for each of the 45 conflict categories analyzed. (As Thus, the total variance (which is ultimately the quantity
discussed in Part 5 of this appendix, some categories re- of practical importance) can clearly be reduced substan-
sulted in so few counts that they were not subjected to for- tially by using longer observation periods, as would be the
mal analysis, and are therefore not included in this or the case in practice.
following tables.) The second phase results were similar, Another way of viewing the repeatability and reliability
and are discussed subsequently. The interpretation of these is given in Table 0-8. Here the coefficients of variation
numbers is as follows, using the left-turn same-direction (CV's) are given, where a CV is the standard deviation
conflict as an example. divided by the mean. They are rank ordered in Table G-8,
The average number of Yl conflicts in a 15-min period such that the lowest CV's appear first. (A low CV is pre-
of observation was 1.1191. The variance, o,2, was 3.3044 ferred, because it implies more precision in the measured
(standard deviation, a-,,, was 1.8178). It is recalled that conflict count.) Thus, the "best" conflict count, based on
conflict counts are from a skewed distribution—not a a low repeatability CV, is Y28 cross-traffic-from-right op-
normal distribution—so standard deviations as large as or portunities), although many other counts are nearly as
larger than the means are quite acceptable if interpreted good. The outstandingly bad count is Y34 (right-turn-on-
correctly. This variance, as noted earlier, can be ascribed red opportunities), which had very poor repeatability (high
to many causes. After accounting for all known causes CV).
(0.2 through o
in the list of Part 3 of this appendix), a Similarly, the best counts from a reliability standpoint
residue, o-,, of 2.7847 remains. This is the repeatability were Y57 (sum of all conflict opportunities) and Y13 (op-
measure. The variance attributed to the observer-to- posing left-turn opportunities). The low CV's (less than
observer differences (reliability) is = 0.1611. Thus, for 10 percent) indicate that all observers counted them nearly
 75

TABLE G-7
CONFLICT COUNT VARIANCES, FIRST 3 WEEKS

e 2S'
Mean Conflict Count Residual Observer
Description Value Variance Variance Variance_

Yl LT same direction conflict 1.1191 3.3044 2.7847 0.1611

Y2 LT same direction PV conflict 0.9104 3.0821 2.3369 0.3802
Y3 LT same direction SC 0.6534 1.8366 1.4595 0.0290
Y4 Ri same direction conflict 0.3232 1.0687 0.9023 0.2838
Y5 Ri same direction PV conflict 0.9482 2.9501 2.3496 0.1659
Y6 Ri same direction SC 0.3327 0.6300 0.5482 0.0012
Y7 Slow vehicle conflict 0.2913 0.5319 0.5143 0.0099
YS Slow vehicle PV conflict 0.7344 1.9520 1.6627 0.1387
Y9 Slow vehicle SC 0.1613 0.2880 0.2658 0.0054
YlO Lane change opportunity 1.2828 4.4775 3.0926 0.1718
Y13 Opp. LT opportunity 10.3918 132.8464 95.7491 0.9474
Y14 Opp. LT conflict 0.2435 0.4497 0.4041 0.0035
Y16 RTR opportunity 8.6450 118.0569 76.9607 1.7269
Y17 RTR conflict 0.1664 0.2569 0.2369 0.0069
Y19 LTL opportunity 8.4621 105.0318 58.5601 1.3107
Y20 LTL conflict 0.2249 0.4003 0.3580 0.0085
Y22 LTR opportunity 8.4444 102.6595 57.3613 1.0843
Y23 LTR conflict 0.1770 0.3407 0.3108 0.0060
't25 Cii. opportunity 6.9057 31.7780 26.3947 2.0502,
Y26 Cii. conflict 0.0984 0.1104 0.1067 0.0009-
Y28 CTR opportunity 6.8313 30.7127 25.0241 2.3113
Y29 CTR conflict 0.1226 0.1374 0.1281 0.0017
Y33 RTOR opportunity 0.2035 2.2340 1.9440 0.0488
Y38 LT same direction C+PV ronflict 2.0319 7.8322 6.1322 0.1854
Y39 Ri same direction C+PV conflict 1.4711 4.6665 3.6183 0.0822
Y40 Slow vehicle C+PV conflict 1.0240 2.7593 2.3803 0.1780
Y31 Rear end conflict 1.5251 4.6539 3.8086 0.3268
Y42 Rear end PV conflict 2. 1253 9.0294 6.7666 1.1291
Y43 Rear end SC 0.9001 2.3944 2.2618 0.0523
Y44 Rear end C+PV conflict 3.6504 16.3531 12.3249 0.3252
Y45 Rear end C+PV+SC conflict 4.5504 30.1752 22.3128 1.2601
Y46 Same direction C+PV conflict 3.6897 16.6815 12.3810 0.3541
'(47 Same direction SC 0.9041 2.9166 2.2809 0.0539
'(43 Same direction C+PV+SC conflict 3.5939 30.6617 22.3934 1.3049
Y49 All CTR opportunities 18.2739 349.3395 189.7430 3.9411
'(50 All CTR conflicts 0.3590 0.7088 0.6108 0.0245
'(52 All TR C+SC conflicts 0.4030 0.9195 0.7882 3.0380
'(53 All Cii. opportunities 9.7131 129.4270 69.9333 1.9124
Y54 All Cii. conflicts 0.2235 0.3906 0.3462 0.0070
Y56 All Cii. C+SC conflicts 0.2494 0.5393 0.3816 0.0088
Y57 All opportunities 30.1809 717.7148 504.2101 7.9394
'(58 All C+PV conflicts 4.3121 20.4286 14.8279 1.2096
Y59 All SC 0.9706 3.1841 2.4609 0.0702
Y60 All C+PV+SC conflicts 5.2827 36.3247 26.2166 1.8287
Y61 Slow LT and opp. LT C+PV conflicts 1.7051 6.8749 5.7132 0.1306
'(62 Slow LT and opp. LT SC 0.5034 1.4872 1.2493 0.3186
Y63 Slow LT and opp. LT C+PV+SC 2.2085 13.3971 11.0050 0.2416
conflicts

aI Y - conflicts/15 mirt
o/ Total variance in the conflict counts.
c/ Repeatability neasure; variance not attributable to observers, time of day,
site, day of week, or ocher measured parameter.
d/ Reliability neasure
e/ Not statistically significant at 0.95 confidence level.
76

TABLE G-8 alike. At the other extreme, Y4 (right-turn, same-direction

COEFFICIENTS OF VARIATION OF REPEATABILITY conflicts) and Y34 (right-turn-on-red opportunities) were
AND RELIABILITY MEASURES not counted very uniformly by different observers. The
CV's of more than 100 percent suggest that these conflict
Reoeatabilitv - Reliability - counts are not too useful. However, these data also show
Conflict
ae
Conflict that Y5 (right-turn, same-direction paired-vehicle conflicts)
Rank Type (Percent) Type (Percent) also gave the observers difficulty (CV = 42.96 percent).
But the combination of Y4 and Y5 (= Y39) had a very
1 28 73.25 57 9.34
2 25 74.40 13 9.37 low CV of 19.5 percent. This suggests that different ob-
3 57 74.40 6 10.61 servers all saw right-turn conflicts about the same, but had
4 49 75.37 39 12.16 difficulty distinguishing the paired-vehicle concept, which
5 53 86.06 22 12.33 splits them into the Y4 and Y5 categories.
6 38 39.30 53 14.23 The second block was analyzed in the same way as block
7 22 89.69 16 15.20
1, with very similar results. Means, total variances, and
8 19 90.43 19 15.90
residual variances were obtained for the same categories.
9 13 94.16 39 19.50
10 46 96.13 25 20.74 The observer variances were manually calculated for the
11 4' 96.17 61 21.19 categories with the greatest percentage changes in
12 60 96.92 38 21.19 Then, the variances between the two blocks were tested
13 16 101.48 53 22.26 for significance using F-tests, and the means were tested
14 48 104.15 14 24.27 using t-tests. The results are given in Table 0-9, and
15 45 104.27 28 24.54
discussed in the following.
16 38 121.87 45 24.67
122.40 38 24.87
The mean values of most (but not all) conflict categories
17 42
13 41 127.97 44 24.88 declined slightly in the second block. Of the seven cate-
19 39 129.30 46 25.05 gories for which changes in mean counts were significant,
20 10 137.09 43 25.42 all were declines. It is also noteworthy that all the signifi-
21 61 140.18 33 25.50 cant declines were in same-direction conflicts, or in com-
22 1 149.11 47 25.68 binatorial categories the counts of which are predominantly
23 63 150.21 3 26.05 same-direction conflicts. This has the possible implications
24 40 150.66 52 27.12
that the observers, on the average, became less sensitive to
25 59 161.62 59 27.29
this general kind of conflict. (Alternatively, seasonal or
25 3 161.65 50 29.60
27 47 167.05 26 .30.49 other unknown factors resulted in fewer traffic conflicts of
28 43 167.08 10 32.31 this kind during the second 3-week phase.)
29 2 167.92 29 33.45 Similarly, the residual variances, o,showed significant
30 8 175.59 7 34.16 changes in 5 of the 1 5 categories tested, and all were de-
31 4 181.56 1 35.37 creases. These changes, however, involved various traffic
32 3 195.91 54 37.36
movements (same direction, opposing, and cross). By re-
33 30 217.69 41 37.39
calling the factors contributing to a,, it appears that either
34 42 220.30 36 37.33
35 62 222.03 20 40.95 some phenomenon caused traffic to be more repeatable dur-
36 6 222.54 40 41.20 ing the second phase or the individual observers become
37 7 46.19 5 .42.96 more repeatable with the additional experience. The latter
38 13 261.07 30 33.55 is intuitively a more satisfactory explanation.
39 54 263.27 23 43.57 The interobserver variances were not significantly dif-
30 20 266.03 9 43.57
ferent except in 1 of the 15 categories tested. Because a
31 56 278.27 52 38.39
confidence level of 0.95 was used, one could expect 1 out
42 29 291.92 42 30.00
43 17 292.49 50.00
of 20 such tests, on the average, to yield a (false) indica-
17
44 23 314.97 8 34.25 tion of significance. As no physical explanation can be
43 9 319.63 2 57.63 advanced for this isolated finding of significance, it appears
45 26 331.96 4 101.32 best to tentatively attribute it to the laws of probability.
34 635.16 34 108.55
PART 8-EXTRA OBSERVERS
a! These data based on 15-minute counts from first
th:ree-week period. The data obtained by the extra observers were analyzed
omer.ciature: separately, and the results were compared with those of the
Conflict type--See Table 0-1 regular observers. The comparison was facilitated by limit-
Y--oean count ing the analysis to the data obtained only at the basic 24
;--residual variance(reoeatabilitv reasure) sites. (The extra observers collected data at other sites, as
0 --observer variance(re1iabi1ic: neasure)
well.) Moreover, to maximize the sample size, extra ob-
server data from all three phases (9 weeks) were pooled.
Generally speaking, the means and variances for the
various conflict categories obtained by the extra observers
were comparable to those of the regular observers. How-
 77

ever, some significant differences were noted. Although the TABLE G-9
reliability measures, Cr02, for the regular observers were all TESTS OF SIGNIFICANCE BETWEEN FIRST AND
small, they were (with two exceptions as noted) all statis- SECOND 3-WEEK BLOCK
tically significant. For the extra observers, however, 11 of
Residual Observer
the 35 variances tested were not significantly greater than Variance Variance
zero (at 95 percent confidence level). However, these find- Conflict Category t-StatiaticJ RatiokJ RatioS!

ings were based on a sample of only about one-fourth the RT same direction Sc 1.58 1.1099 1.3876
size available for the regular observers, and many of the
RT same direction PV conflict 2.29* 1.5129° 1.0693
11 insignificant variances were "close" to being significant.
The implication is that a bigger sample may have resulted Opp. LT conflict 1.69 1.3689* 2.3143

in fewer insignificant findings. RTR conflict 0.37 1.1646 1.2319

Mean conflict counts were examined, and some differ- LTR conflict 1.27 1.6104* 1.4833
ences were found. Generally, the extra observers noted
RT same direction C+PV conflict 2.69° 1.2834 2.0550
slightly more conflicts in the various rear-end categories,
and substantially fewer in the cross-traffic categories. Also, Slow vehicle C+PV conflict 3.41° 1.7001* 1.3244
lane-change opportunities were recorded more than twice Same direction C+PV+SC conflict 3.06° 1.2271 2.4223
as frequently by the extra observers.
All CTR C+SC conflicts 0.90 1.1089 1.0211
The major interest in examining the extra observer data
was to further establish (or discount) the reliability find- All CTh C+SC conflicts 1.48 1.5708* 1.4194

ings for the 12 regular observers. The ratios of the vari- All Opportunities 0.26 1.0209 2.3554
ances, Cr02 , were examined using the F-test. Only three of
All C+PV conflicts 3.49* 1.2326 1.8572
the conflict categories for which both groups had statis-
All SC 2.05* 1.1996 1.9554
tically significant (hence, testable) variances were different
at the 95 percent confidence level. In each of these three- All C+PV+SC conflicts 3.28* 1.2491 2.1000
left-turn, same-direction conflict (not paired vehicles), all Slow LT and opp. LT C+PV+SC conflicts 1.67 1.1392 3•7457*
rear-end conflicts (not paired vehicles), and all rear-end
aI To be significant, t-statistic must exceed 1.96
secondary conflicts-the extra observers exhibited higher
variances (lower reliability). The first two of these cate- b/ To be significant, ratio must exceed F095 (500,500) 1.33

gory differences can be attributed to the exceptionally high c/ To be significant, ratio must exceed F095 (11,11) - 2.82
counts of just one of the extra observers who had difficulty
in distinguishing between paired vehicles and unpaired ve- *Statistically significant at 95% confidence level

hicles. The third category exhibited a just barely significant

difference, which has no obvious interpretation. significance of these differences. Thus, for example, there
The residual variances (repeatabilities) were also ex- was an average of 0.26 more Yl conflicts per 15 min at
amined, in two ways. F-tests determined that for about high-speed intersections than at low-speed intersections, but
half the categories, the extra observers showed smaller this difference was not significant. However, the difference
variances (higher repeatability), although they had higher in Y2 conflicts at such intersections (0.47 per 15-mm) was
variances for four categories (three of which may, again, highly significant (0.999 confidence level). Other entries in
be attributed to the extra observer previously cited). The Table G-10 are interpreted in an analogous fashion.
second analysis, involving the Spearman Rank Correlation, Over-all, there were no clear cut, uniform differences in
was applied to determine if the repeatabilities among the conflict counts by day of week; a few conflict types were
many conflict categories were ranked in about the same more prevalent on some days than on others, but most
order between regular and extra observers, despite differ- differences were not significant or consistent. The only
ences in magnitudes. The result was a correlation co- possibly important tendency is that Mondays may have
efficient, r8 = 0.984, indicating a very high correlation experienced a few more conflicts of some types than did
between the rankings of the two groups. the other week days (Yl, Y53, Y57). Fridays (not in-
cluded in Table G-10 because they were not used in the
PART 9-OTHER FACTOR EFFECTS - first phase, but they were analyzed in the same way with
second phase data) experienced about the same conflict
Several factors, in addition to reliability and repeat- rates as Tuesday through Thursday, but with somewhat
ability, were included in the statistical model (Part 3 of this more conflict opportunities.
appendix) and analyzed. The results are described here Other differences notable in Table G-10 are that high-
with the exception of the time-of-day effects discussed sub- speed intersections experienced more rear-end conflicts and
sequently in a succeeding section. conflict opportunities than low-speed intersections: 3-way
The impacts of these major sources of variation are given intersections experienced more turning conflict oppórtuni-
in Table G-10 for the first 3-week phase. The second-phase ties but fewer conflicts than 4-way intersections; signalized
results were similar, except for the day-of-week effects de- intersections generally experienced more rear-end and op-
scribed later. Rather than stating the results as variances, posing left-turn conflicts, and fewer cross-traffic conflicts;
it is more informative (and appropriate) to examine the and there were generally more conflicts on 2-lane roads
differences in appropriate mean values, and the levels of than on 4-lane roads.
78

'C 0'U'In015.'. C 07 51In7flIn1'lN- O'In..,.??ll .0.0-Ct-c

..705.C5-'054.000'00 5'5400.0 5'''.7'7 .0.07.-P-In _tCIn.TtCP.

.0.7.0.0 C'r'.C.07,.0.SInr.'CtU' In

tt0000 Cost 000C©CC0C00 0C© 0C©C

C.1,
70_ 77.
2.
0'.0.0.7P4
Cr-Inr'.tN-PlO-tl
CC 05''05'0'

0.0.7.0 0' 9t#--54eCrSC 70.0.05.5,54.00.0.1

000.0.0 - C-
0'InU'C5'5-U'InP.I.r17.17.IU' In115'In7.0C C7.I7.J1"l'

00000 C54C _54CCCC00CCCCC0 .0CC 0 00 ©OC -

.0000.00.0.0 <5.0r'OlflF'.Ifl
.O5...7U'CC*0In1l1).07..7.J0' 0o
U'00'C..P7.,54.00'0'InIn 5''
5'l.011)r'. -.7 S.07.l..yeflc-r'Ir--r--r .05-I0'7..N.In-In
1
'0'
7'.00' -
7.4 CCC - CCC
55.0.05.7.40 .05,.07.Jr1).0U1).0 Q1tIltlO0P'IO 5'.J5.547.'.tl5.0-InN.
0 5' C 7.4 C C 0 -0
7.1 75 0 0 -0 0
5' C C C C CU' C 0 .75
70 -17.1 CC 0 C C 7.1 --
00 0000000C.C,.00C0000000000 54 00C 05 00 0000

• .7 .7 4 .7 4:

0'.0In5..00'5.InIOICNOO.0U10< 015'..TU'IUSC'I.. .000.5' 0'0'0' 5' 5'"1 "' C

.0C.00C .0.0.0.0 .0 -0O7.00'4fl.07.l5'IOC
.OIn0'In0
7.
I 000 0 05'CSSCSS0000 0054CC 'C

1'

U'NU'545'5'7.I.05'05'U5.00r.7.. 7.47... O' .c

75 0.
.0.0.0.0CC-c
7.1 .7 "5 C CU' 755. 5' 5' 5' .0
C
a-
C.0CCC CCCPCC7..0C_ C.54CCC0CC0C0.0CC 54 - 05- CC 5

14 ...0 •- .

14 .
>+
.5 0
.5>114 0>0'
110.141147.11CC
C
041.
++ 0
011414
14
U
-0
0>
11144
-*
'4 41
UI1'4
4.
-
-
CC>14' C +0'
----- C
0.5000014 1...
---
000000
0'- 0014
.5 0 0. U 11.
14_I _114 1414.5.010. l.I*IJl4U14I.I
70 00.11.1..
11.
l4140
---
0
000010041001.10011

0 C0I.I014 +
07.5.0. - - 5'.,
01)0 .1.0 >>O 0014.., 14... 000
0414400
014.01401404100140.50710 11. 7.0.'lO'.- '..0 0 Q0
-------
0110000)111
00.000.4000 OUUI.1140>0+ +1JIJI.00.0010.Z4h 00 000041
U0007. 0.11414.41J1..'
0U141.1.
0'- 014.. 0L
14-
0 04.10.
+
UU7.C1)00U
•
'407.0+
Ot0
11000000140
.Z1)000'01. 1414
5.0+U
0>
750001..

0'
1)1)714
-1 --
0.4-4.-I-_
000-000
0111111.5 E
1)777.1)
>11
C
041_I
5.00.00.00.0.00001100011
000140 U 000 fl
+
Z0.00.0.00.000 .5000 C'.C InInX_1_1415.U.
5.,-
0L00140 00111141 .5
0 11 .9

00C5.0.X_1.41XO
------------------------------

P1en-7C4•0.0 -------5-17.15.5454545,5,5, .0.1.7.0'? ................0.0.0.0 41

PART 10-SITE ANALYSES The general approach was to compare mean counts for
these categories among the 24 basic sites (see App. F), in
Certain summary categories of conflicts were examined relation to the traffic volume counts at the sites.
on a site-by-site basis to supplement the factor effect analy- First, however, the rank order of corresponding mean
ses previously described. The categories examined were:
category values of the first and second phases were com-
Y57 - all opportunities pared using the Spearman Rank Correlation test. The cor-
Y63 - opposing left-turn plus slow-for-left-turn relation coefficients, r3, for the five categories and the vol-
conflicts ume counts ranged from 0.830 to 0.893, indicating that the.
Y48 - all same-direction conflicts sites ranked in essentially the same order in each phase.
Y52 + Y56 - all cross-traffic conflicts Thus, just the first phase data were used in the remainder
Y60 - all conflicts of the tests.
 79

Spearman rank correlation coefficients were then deter- relatively few turning movements from the mainline
mined between each of the five conflict categories and the (18 percent) of the total intersection traffic, and had good
traffic volumes. The resulting correlation coefficients were sight distance. Site 19, however, featured a little more
r8 = 0.157, 0.252, 0.286, -0.140, and 0.257, respectively. cross traffic, a significant left-turning movement from one
Thus, the numbers of conflicts observed at the sites were leg (36 percent of its volume), and the sight distance was
essentially independent of the traffic volumes. However, impaired as the intersection was on a crest vertical curve.
this result should not be unexpected because the set of sites, Site 22 was in a residential area, and had no obviously
rather than being homogeneous, covered a variety of geo- unusual geometric or traffic characteristics but, as discussed
metric configurations and traffic controls, usually reflecting later, experienced many accidents. Other locations with
the traffic volumes. For example, with higher volumes abnormally high conflict counts in some categories were
there tend to be more lanes and more sophisticated traffic sites 1 and 20 (high conflict opportunities and cross-traffic
control, which, in turn, tend to reduce traffic conflicts. conflicts), site 21 (high conflict opportunities), and site 23
Some observations can be made about some of the par- (high cross-traffic and total conflicts).
ticular locations with unusually high or low conflict counts. One site exhibited especially interesting conflict patterns.
First, the eight sites with signalization (sites 9-16) all Site 3 had an "average" volume compared with the 24 in
tended to have fewer than average conflicts in relation to the experiment (1,000 vph), but had the highest number of
the traffic volumes. Particularly noteworthy in this respect conflict opportunities and nearly the lowest number of con-
are sites 9, 14, and 16. Site 14 had conventional signaliza- flicts of all the sites. This unsignalized, rural, high-speed
tion and pavement markings (e.g., no special provisions for site experienced significant cross-traffic and left-turning
left turns), carried moderately heavy volumes (average movements, but the mainline had a narrow median strip
1,620 vph fairly evenly split between the 4 legs), of which near the intersection which "protected" these movements.
23 percent represented turning movements, but had ex- Another way of looking at these conflicts is to construct
cellent sight distance. Site 9 had a little less volume (1,120 conflict rates using appropriate traffic volumes. Table G-1 1
vph), split 70/30 between mainline and cross traffic. There gives these rates for the 24 sites, together with the means
were separate left-turn lanes, but no special left-turn phases. and standard deviations. On the basis of the traffic move-
Site 16, a 3-way intersection, also carried moderate vol- ments affected, the following sites had the highest conflict
umes (1,430 vph), mostly on the main road. However, rates:
50 percent of that mainline traffic turned at the intersection.
Left-turn conflicts-sites 17, 18, 21, 22 (unsignalized,
The signalization included a left-turn phase, although no 4-way, 2-lane sites).
left-turn bay was available. Same-direction conflicts-sites 17-24, especially 17
Locations with unusually high conflict counts in most and 18 (unsignalized, 2-lane sites).
categories included sites 18, 19, and 22. The first two were Cross-traffic conflicts-sites 2, 5, 23 (unsignalized).
2-lane rural roadways, carrying 55-mph traffic at fairly Total conflicts-sites 17, 18, 19, 22 (unsignalized,
modest volumes (700 to 800 vph). Site 18 experienced 2-lane sites).

TABLE G-11
SELECTED CONFLICT RATES BY SITE
Left Turn Same Direction Cronn Traffic Total Conflicts
Conflicts Per Conflicts Per 100 Conflicts Per 100 Per 100 Total
100 Left Turning Main-Line Cross-Road Intersection
Site Vehicles Vehicles Vehicles Vehicles

2.6917 1.5071 1.1364 1.4140

2 12.1000 1.6064 2.6250 1.7785
3 2.0133 0.7006 0.5710 0.7186
4 20.2267 1.8536 2.3061 1.9698
5 18.9583 1.7220 3.1333 1.8833
6 . 17.6250 1.6074 2.2394 1.7246
7 2.6667 0.8767 1.8708 1.0123
8 1.9889 1.6124 . 2.0895 1.7726
9 1.1125 0.5636 0.0560 . 0.4989
10 17.1000 1.8637 0.4833 1.8432
11 15.5833 1.4957 0.4267 1.4622
12 19.4167 1.3516 0.0573 1.2705
13 15.2481 2.6769 0.0549 1.4343
14 9.0722 1.5902 0.0617 0.9877
15 12.9611 1.8358 . 0.0483. 1.5029
16 1.8667 0.9480 0.0924 0.8062
17 34.8143 4.9867 2.2238 4.6502
18 35.6733 5.7850 1.6556 5.0492
19 13.6000 3.9029 1.5051 3.5763
20 7.3600 3.2207 1.1963 2.7965
21 22.5667 2.8490 . 0.9152 2.3844
22 25.1133 3.8751 2.0039 3.5361
23 12.9667 3.0785 3.1250 3.0987
24 13.1333 2.6253 0.9667 2.4423
flean 13.9943 2.2555 1.4102 2.0672
Standard DeviatIon 9.6729 1.3325 1.3820 1.1889
 High Low PART 11-EXTRA SITES
Speed Speed Traffic conflicts were counted at four sites with charac-
teristics different from those of the basic 24 sites. These
4-Lane .9 13 extra sites were all signalized, 4-way intersections on 2-lane
10 14 roadways. Two were on high-speed routes and two were
on low-speed routes. Data from these sites were analyzed,
using an analysis of variance, in comparison to four sites
2-Lane 25 27 that differed only in that they were on 4-lane roadways
26 28 rather than on 2. The experimental design is shown in
Figure 0-2.
A summary of the extra-site data is given in Table G-12.
(Al-1 sites were Four-Way, Signalized) Four conflict categories are represented-all opportunities,
Figure G-2. Experimental design for extra sites, all left-turn conflicts, all same-direction conflicts, and -all
showing pertinent site numbers. conflicts. The AOV results are displayed in Table G-13.
In no instance was there a significant difference attributable
to the number of lanes. This was due in large part to the
TABLE 0-12 highly significant differences observed between the mem-
EXTRA SITE CONFLICT DATA bers of the "replicated" pairs. That is, because the two sites
in each cell of the design with nominally identical features
No of All All All Same All tended to have substantially different conflict counts (see
15-Min Opportunities Left-Turn Direction Conflicts
(Y48) (Y60)
Tables 0-11 and 0-12), the differences between cells at the
Site Counts (Y57) (Y63)
same speed level were not great enough to be significant.
25 90 11.618 0.344 3.089 3.467 This finding is quite different from that found for unsignal-
(6.208) (0.810) (2.909) (2.988)
ized intersections, where conflicts tend to be higher on
26 112 18.464 0.420 0.741 0.851 2-lane roads than on 4-lane roads.
(16.065) (0.824) (1.718) (1.775)
A speed effect was significant for three of the four cate-
27 80 10.225 3.775 6.913 7.750 gories in Table 0-13. Low-speed signalized intersections
(4.625) (3.721) (4.987) (5.552) tended to have more traffic conflicts than high-speed sig-
28 130 5.131 2.754 7.177 7.462 nalized intersections. This is the reverse of the more gen-
(2.692) (2.366) (5.037) (5.166) eral finding in Table 0-10, which indicated that higher
a/ 15-Minute averages and, in parentheses, standard deviations speed intersections tend to have more conflicts. Howeveç,
that finding was based predominantly on unsignalized inter-
sections. Table 0-13 also. shows one significant statistical
TABLE 0-13 interaction. Whereas 4-lane signalized intersections had
AOV TABLES FOR EXTRA SITES about the same number of left-turn-related conflicts regard-
less of traffic speeds, for 2-lane, signalized intersections the
Conflict
Category Factor df SS MS F.' low-speed intersections had significantly more traffic con-
flicts than the high-speed intersections.
All L 1 1.790 1,790 1.16
Opportunities S 1 1.936 1,936 1.25
(Y57) LS 1 3,360 3,360 2.17 PART 12-TRAFFIC VOLUMES
R 4 6,184 1,546 20.86*
e 760 56,327 74 For completeness, the observed traffic count data are
summarized in Table 0-14. These data are listed by site,
All L 1 16 16<1
Left-Turn S 1 3,622 3,622 26.60*
day of week, and time of day. The volume data, them-
(Y63) LS 1 1,038 1,038 7.62* selves, were not analyzed extensively. Rather, they were
R 4 545 136 21.99* used in a number of contexts to help explain the traffic
e 760 4,706 6
conflicts findings.
All L 1 113 113 Fl These data indicate, for example, relatively large site-to-
Same Direction S 1 2,194 2,194 12.39*
(Y48) LS 1 651 651 3.67 site differences (e.g., from 456 total vph through location
8 4 708 177 11.13* 20, to 2,292 vph through location 12). There were no
e 760 12.093 16
appreciable differences by day of week. There is some indi-
All L 1 46 46<1 cation that volumes increased slightly through the week,
Conflicts 5 1 2,506 2,506 11.41* and that the cross-traffic volumes were greatest on Fridays.
(Y60) LS 1 723 723 3.29
8 4 879 220 11.89* The morning and evening peak flows are evident, with the
e 760 14,042 18 evening peak a little higher than the morning (1,908 vs.
a/ Nomenclature: df = degrees of freedom, SS sum of squares, 1,584 vph), and lasting longer.
MS = mean square. F F-ratio, L lane factor,
S = speed factor, .LS lane/speed interaction,
8 replicate site factor, e = error term, PART 13-TIME-OF-DAY EFFECTS
significant at 0.95 or higher.
A number of count data categories was reviewed for
b/ F-ratio based on replicate site MS except for F-ratio for 8 their time-of-day characteristics. An illustrative, yet corn-
which is based on the error MS.
 81

TABLE G-14
TRAFFIC VOLUME COUNTS

F.ocacion 1 2 3 4 5 6 7 8 9 40 11 12 1-3 4-6 79 10-12 I

1 2 17 6 2 4 S 2 16 5 2 5 .5 25 11 22 11
2 2 35 2 3 1 2 4 42 4 1 1 3 39 6 50 4
3 15 14 --------- 12 14 16 -- 15 29 -- 24 31 86
3 5 41 ---------- 46 2 3 -- 2 46 -- 48 ii
3 1 46 2 3 2 1 3 44 0 1 2 0 39 6 48 3 ,
3 3 38 1 1 1 1 1 36 3 4 2 3 42 3 39 3
7 -- 33 6 2 -- 6 1 33 -- -- -- -- 39 8 34 - 81
8 9 47 . 50 10 11 -- 8 56 -- 60 79
9 4 27 S 5 7 6 6 24 2 3 6 2 36 17 30 11
2 31 2 2 1 2 2 30 2 2 2 2 35 5 33 5
11 4 39 ------ 35 1 3 -- 1 43 -- 35 5
33
13 -- 71 18 6 -- 19 6 71 -- -- -- -- 89 25 77 -- 797
13 4 23 2 7 29 3 5 24 5 4 26 3 28 38 33 33 133
13 3 26 4 2 25 4 3 33 6 4 20 5 33 31 42 29
135
15 6 39 --------- 45 18 8 -- 22 43 -- 63 29 137
If. 25 25 ---------- 17 17 16 -- 19 50 -- 34 35
119
17 2 39 3 4 1 4 5 34 2 2 1 2 34 9 31 5 95
18 1 20 1 5 2 1 4 22 1 2 1 7 23 8 28 3
63
19 -- 15 4 9 -- 4 -10 18 -- -- -- -- 19 13 27
-- 59
20 5 10 ---------- 9 4 6 -- 4 15 -- 74 9 38
21 1 15 2 1 3 2 0 13 1 1 3 1 19 6 15 5
4 26 4 1 4 .3 1 30 2 3 3 2 34 8 33 9
23 3 18 ---------- 18 6 2 -- 6 20 -- 25 3 53
24 -- 26 4 3 -- 6 3 34 -- -- -- -- 31 9 37 -- 47
25 2 38 5 3 7 -4 3 36 5 3 7 4 45 15 33 13 18
26 9 4 6 2 53 6 2 3 1 8 30 - 0 19 60 5 33 112
27 3 39 5 2 1 6 3 39 1 3 1 3 47 9 43 7 106
28 1 42 6 1 8 3 1 44 2 1 9 2 47 14 48 42 121

2s9 of Ileek

5 5 7 5 35 13 37 13
7. 4 3 30
5!onaay 4 9
4
3
3 5 3 6 37 11 36 . 98
29
Thesdav 6
5 3 4 4 6 5 3 31 6 5 5 - 38 13 8 13 102
Wdnesdav 10)
5 6 8 5 3 13 36
Thursday 5 30. 5 .3 9 5 3 30 105
11 4 37 .17 17
6 29 4 4 12 4 3 27 5 6
Friday

Time Period

44 5 5 4 37 8 7 11 6 54 13 46 19 132
1 6 8 5
7 4 2 32 6 5 8 5 37 13 39 15 103
2 5 29 6 3
4 23 3 2 4 3 2 26 5 4 6 4 28 8 32 12 80
3
23 4 4 5 5 25 9 29 11 71
4 3 20 3 3 5 3 3
4 2 5 3 -2 21 4 4 4 4 26 9 26 10 71
5 21 3
4 3 3 22 4 4 4 4 27 9 27 10 73
6 4 22 3 3
4 3 4 3 2 23 4 4 4 4 28 9 28 10 75
7 23 3
3 4 26 4 3 5 3 3 27 - 5 4 6 4 29 9 33 11 82
25 3 3 5 3 2 25 4 4 5 5 30 10 30 13 53
9 3
6 25 3 3 4 3 2 25 3 - 4 5 5 30 10 29 12 81
10
28 4 3 6 4 3 26 4 5 5 5 34 12 32 13 91
11 5
4 30 4 3 6 4 3 28 4 4 5 5 36 12 33 12 93
12
6 34 5 4 8 6 5 37 6 7 6 6 42 16 45 16 119
13
7 44 5 5 12 7 5 47 7 9 9 7 53 21 57 21 152
14
7 49 6 5 15 7 6 47 8 8 7 8 59 23 57 20 159
15
10 6 4 37 7 5 6 5 52 19 46 14 131
16 7 43 6 5

a/ Mean values of S-minute counts.

9/ See sketch,
8 7
- Note, on 3-way intersections some
Observer i
movements are not possible.

L -
12 • 6

11

10

1 2 3 Observer

prehensive, presentation of the results is shown in Figure than traffic volumes. The traffic volumes show morning
G-3. In this figure are shown plots of several categories, and afternoon peaks, with the latter being slightly higher.
Each category is normalized by its daily average. Traffic conflicts and severe conflicts have similar, but
The distributions of the volumes and conflict opportuni- more marked peaks. The conflicts counts drop more dras-
ties are very similar, although the opportunities exhibited tically during midday, and peak at a higher level during the
a slightly more stable (less variable) count during the day late afternoon. Severe conflicts, on the other hand, show
82

almost no peak at all during the morning, but rise con-

tinuously starting in the early afternoon, and remain high
- - Volumes
200
- - -- Opportunihes through the afternoon peak.
-. - Conflicts These findings can also be viewed from a rate standpoint.
Severe Conflicts
Table 0-15 gives the number of conflicts and severe con-
Note: flicts per 10,000 vehicles passing through the intersection.
Data are from 1st 3-week block
except severe conflicts, which
The heavier peak in the . early evening hours is clearly evi-
include all 3 blocks. dent, especially for severe conflicts. It appears that driving
habits, on the average, are deteriorated late in the day, due
perhaps to fatigue, alcohol, a greater preponderance of
younger drivers, or some combination of such factors.

PART 14—RELATION TO ACCIDENTS

A brief examination of accident data comprised the final

element in the analysis process. Accident data were ob-
tained for the years 1976-1978 for a number of the sites
(those sites for which such data were readily available).
These data, together with other details, are given in
Table 0-16.
Correlation coefficients were calculated between the total
accident numbers in Table 0-16 and the counts in several
conflict categories. The results are as follows: for severe
conflicts, r = 0.214; for all conflicts (Phase 1), r = 0.373;
for all conflicts (Phase 2), r = 0.302; for all opportunities,
r = —0.053; and for traffic volumes, r = 0.689. Of these,
only the correlation with traffic volumes was statistically
significant.
These findings are probably not particularly meaningful,
50 I however, because the accidents and the conflict categories
are each agglomerations of heterogeneous events. A more
useful approach is to compare accidents of certain types
with conflict categories of analogous types. Because this.
0700 0900 1100 1300 1500 1700 means dividing the (limited) accident data into subsets, the
Ti me first step was to limit the analysis to only those intersec-
Figure G-3. Time-of-day effects. tions with relatively large numbers of accidents. As can
be seen from Table 0-16, the seven sites with 20 or more
accidents in 3 years were retained. (Site 2 had enough
accidents, but the detailed accident reports for 1976 and
TABLE G-15
CONFLICT RATES BY TIME OF DAY

Conflicts Per 10,000 Vehicles TABLE 0-16

Time Period Regular. Severe - ACCIDENT DATA FOR 1976-1978

0700 - 0730 149 0.4 Conflict- Opposing

0730 - 0800 200 0.9 Site

Total
Accidents
R1oted a/
Accidents7
Left
Turn
Same
Direction
Cross
Traffic
0800 - 0830 146 0.6
0830 - 0900 138 1.0 2
5
30
39 29 4 2 23
0930 - 1000 126 0.4 7 8
1
1000 - 1030 120 0.9
9
10
21
29
4
13
2
10
1
2 1

1030 - 1100 132 0.4 11 6 b/

16 6 7 3
1100 - 1130 133 0.6.
12
13
41
35 10 7 1 2

1330 - 1400 151 0.9 14

15
35
13
4
b/
2 2 0

1400 - 1430 179 1.1 16 11 b/

1430 - 1500 148 1.5 19

21
4
14 bI
1500 - 1530 176 1.3 22 20 6
-0 1 5

1600 - 1630 202 1.0 24 2 b/

1630 - 1700 218 . 1.5 Totals 308 82 31 16 35

1700 - 1730 231 1.3 a/ Accidents involving 2 or more vehicles on dry roads daring daylight

1730 - 1800 229 1.4 hours

b/ Accident details not obtained
 83

1977 were not immediately available, so it could not be conducted between the total conflict-related accidents and
included.) the total conflicts. The results of these analyses are shown
All accidents not satisfying the following criteria were in Figures G-4 through 0-7.
eliminated from further consideration: involved two or The findings are encouraging, but not conclusive. All
more vehicles, occurred during daylight hours, and oc- four analyses resulted in positive correlations, but none
curred on dry pavement. These limitations helped ensure were extremely strong. The two with highest correlation
that the accidents would be comparable in nature to the coefficients—the opposing left-turn and the cross-traffic
traffic conflicts observed. The remaining accidents were situations—each was based on a relatively large number of
then divided into three categories according to the vehicle accidents. They also have relatively large slopes, suggesting
movements—opposing left turn, same direction (rear end), that the accident rate is fairly sensitive to the correspond-
and cross traffic. The results are included in Table G-16. ing conflict count. On the other hand, the rear-end acci-
Regression analyses were performed between these acci- dents are less correlated to rear-end conflicts, and there are
dent numbers and the mean 1.5-min conflict counts in the substantially fewer accidents per conflict. These findings
corresponding categories. Also, a regression analysis was suggest that the rear-end conflicts are not as serious, in

15

20

LCoc

0b1
"Cs

5
01 A I I I I I 1 I I
0 0.2 0.4 0.6 0.8 1.0
Conflicts/15 Minutes
Figure G-4. Opposing left-turn accidents and conflicts. I I, I I I 1 I I I 1 I
0 0.2 0.4 0.6 0.8 1.0
Conflicts/15 Minutes
Figure G-6. Cross-traffic accidents and conflicts.

30

0
) 0
Conflicts/15 Minutes
Conflicts/15 Minutes
Figure G-5. Same-direction (rear-end) accidents and
conflicts. Figure G-7. Total accidents and related conflicts.
 84

general, as cross-traffic and opposing left-turn conflicts, in

10
agreement with conclusions stated by others.
The relationship between total accidents and conflicts is
perhaps misleading. As seen in Table G-16, 66 of the 82
8
accidents (80 percent) were of the opposing left-turn or
cross-traffic variety. However, the over-all mean conflict
count of 5.375 per 15 min contained only 0.650 counts of
these types (12 percent). Thus, most of the conflicts, but
few of the accidents, were of the rear-end types.
Finally, regression analyses were performed between ac-
cidents and corresponding conflict opportunities (or vol-
umes, as appropriate). The correlation coefficients were
negative and near zero for left-turn and cross-traffic acci-
2
dents, and essentially zero (0.060) for total accidents. For
same-direction (rear-end) accidents, however, the correla-
tion coefficient between accidents and (mainline) volumes
0 50 100 150 200 250 was 0.971, indicating a significant relationship (see Fig.
Mainline Volumes/5 Minutes G-8). Data limitations (7 sites, 16 total accidents) suggest
Figure G-8. Same-direction accidents and volumes. that further verification is necessary, however.

APPENDIX H

PROCEDURES MANUAL FOR TRAFFIC CONFLICTS OBSERVERS

INTRODUCTION For these reasons, other traffic indicators or measures are

useful. The traffic conflicts technique (TCT) is one im-
The most direct indicators of traffic safety problems are portant way to measure the accident potential of a high-
accidents. If an unusually high number of accidents hap- way or street location without having to wait for accidents
pen at an intersection, something is probably unsafe about to happen.
its design or operation. Traffic engineers, therefore, study The TCT was originally developed by the General Mo-
accident data to determine: tors Research Laboratories in 1967. It was designed to be
Which intersections are most hazardous. a systematic method of observing and measuring accident
What kinds of hazards are present. potential at intersections. Conflicts were defined as the oc-
Whether a change in the design or operation has been currence of evasive vehicular actions and were recogniz-
effective. able by braking and/or weaving maneuvers.
Figures H-i and 1-1-2 show two traffic conflicts. In
But, there are many problems with accident data. All Figure H-i, a vehicle onthe cross street pulled into the
accidents are not reported, and, when they are, the infor- intersection in front of the other vehicle, which had to
mation may not be accurate or complete. Also, accidents brake to avoid a collision. In Figure H-2, the oncoming
are fairly infrequent and happen sporadically, so a long vehicle made a left turn when the other vehicle was too
time is needed to collect enough accident data to be useful. close, causing it to brake. These are just two examples, but
For example, the numbers of accidents at one intersection they show that traffic conflicts are situations where a driver
from year to year are 10, 24, 12, 20, 18, 14, 20, 10, 14. If brakes or swerves to avoid a collision.
one looked at only one or two years, one would not know This procedures manual gives the basic information for
what is typical. Is it 10 to 12 per year, or is it 20 to 24 per conducting a traffic conflicts study. These studies are usu-
year? This is a statistical problem that most people do not ally under the direction of a traffic engineer who deter-
appreciate. Many times, people overreact to a "rash" of mines that the study is needed, schedules the activities,
accidents that may be just a normal statistical fluctuation. supervises the data collection, and performs or supervises
 85

Figure H-I. A traffic conflict with a crossing ic/tic/c Figure 11-2. A traffic conflict wit/i an oncoming vehicle.

the analysis. He also interprets the findings and makes TRAFFIC CONFLICTS DEFINITONS
decisions about indicated intersection improvements. The
General Definition
data collection is normally the function of traffic techni-
cians or other persons working for the traffic engineer. A traffic conflict is an event involving two or more road
This manual should help one in making traffic conflict users, in which the unusual action of one user, such as a
counts. It contains the definitions that will be needed as change in direction or speed, places the other user in
well as step-by-step instructions for use in the field, and a danger of a collision unless an evasive maneuver is taken.
number of examples. An instructor's and engineer's guide Generally speaking, the road users are drivers, but the
(App. I) is available for the engineer who will make the definition also includes pedestrians and cyclists.
decisions about where and when to use the TCT, and how The action of the first user is unusual because it is not
to interpret the results. what one would expect most drivers to do under the same
A traffic conflicts survey usually takes from several hours circumstances. (But, it does not have to be violent or ex-
to several days of careful observation of traffic at an inter- tremely rare!) An exaniple is when a driver brakes while
section. Specific procedures are used to assure uniform going through an intersection even though there is no cross
data collection so that valid comparisons and judgments traffic. This restriction does, however, rule out actions that
can be made. A survey requires one or more observers. nearly all drivers take under the same conditions, such as
who follow a set schedule and perform a number of sepa- stopping for a stop sign or red traffic signal, or reducing
rate but related tasks. These tasks include recording di- speed before turning. Thus, traffic conflicts do not include
mensions and other details about the intersection and its actions that result from obeying a traffic control device or
traffic control devices, making judgments about the traffic that are normal responses to the roadway.
flow problems and their causes, and (most importantly) ob- For a traffic conflict to occur, an actual impending col-
serving and recording the traffic events in the TCT survey. lision is not necessary. An action or a maneuver that
The next section of this manual provides detailed defini- merely threatens another user with the possibility of a col-
tions of the types of traffic conflicts that are observed and lision is sufficient. Also, some collisions occur without eva-
recorded in a TCT survey. The third section describes how sive maneuvers. They are included as extreme cases under
to prepare for a traflic conflicts survey. including instruc- this broad definition.
tions and suggestions for getting started on the survey. The An intersection traffic conflict is described as an event
fourth section tells how to conduct the survey, and includes involving several stages as follows:
time schedules, forms to be completed, etc. The final sec-
Stage 1. One vehicle makes some sort of unusual or Un-
tion gives many examples of traffic situations and explains
expected maneuver.
how they should be handled using the traffic conflict
definitions. Stage 2. A second (conflicted) vehicle is placed in dan-
ger of a collision.
Finally, it is important to note that this manual is in-
Stage 3. The second vehicle reacts by braking or swerv-
tended to support a formal training program or to be use(]
ing.
as a handy reference after training. One cannot do a traffic
Stage 4. The second vehicle then continues to proceed
conflicts study from this manual alone. A one- to two-week
through the intersection.
training course is very important to assure uniform and
accurate data collection. The last stage is necessary to convince one (the observer)
86

that the second vehicle was actually responding to the then continues through the intersection (see Fig. H-7 for
maneuver of the first vehicle and not, for example, to a the directions of the two vehicles).
traffic control device.
Within this framework, a basic set of operational defini- Left-Turn, Cross-Traffic-From-Right Conflict
tions can be stated, corresponding to the different types of
A left-turn, cross-traffic-from-right conflict occurs when
maneuvers. Over-all, 9 basic intersection conflict situations
a vehicle approaching from the right makes a left turn, thus
are useful in pinpointing operational or safety problems,
placing a second vehicle in danger of a broadside collision.
and several others may be important in special situations.
The second vehicle brakes or swerves, then continues
The following paragraphs describe each one. All are de-
through the intersection (see Fig. H-8).
scribed from the viewpoint (direction of travel) of a driver
that is being conflicted with (the second vehicle) rather
than from that of the road user creating the conflict Thru, Cross-Traffic-From-Right Conflict
situation. A thru, cross-traffic-from-right conflict occurs when a
vehicle approaching from the right crosses in front of a
Operational Definitions second vehicle, thus placing it in danger of a broadside col-
lision. The second vehicle brakes or swerves, then con-
Left-Turn, Same-Direction Conflict tinues through the intersection (see Fig. H-9).
A left-turn, same-direction conflict occurs when the first
vehicle slows to make a left turn, thus placing a second, Left-Turn, Cross- Traffic-From-Left Conflict
following vehicle in danger of a rear-end collision. The
second vehicle brakes or swerves, then continues through A left-turn, cross-traffic-from-left conflict occurs when a
the intersection (see Fig. H-3). vehicle approaching from the left makes a left turn, thus
placing a second vehicle in danger of a broadside or rear-
Right-Turn, Same-Direction Conflict end collision. The second vehicle brakes or swerves, then
A right-turn, same-direction conflict occurs when the first continues through the intersection (see Fig. H-b).
vehicle slows to make a right turn, thus placing a second,
following vehicle in danger of a rear-end collision. The Thru, Cross- Traffic-From-Left Conflict
second vehicle brakes or swerves, then continues through
the intersection (see Fig. H-4). A thru, cross-traffic-from-left conflict occurs when a ve-
hicle approaching from the left crosses in front of a second
Slow-Vehicle, Same-Direction Conflict vehicle, thus placing it in danger of a broadside collision.
A slow-vehicle, same-direction conflict occurs when the The second vehicle brakes or swerves, then continues
first vehicle slows while approaching or passing through an through the intersection (see Fig. H-Il).
intersection, thus placing a second, following vehicle in
danger of a rear-end collision. The second vehicle brakes Secondary Conflicts
or swerves, then continues through the intersection (see
Fig. H-5). In the foregoing nine conflict situations, when the sec-
The reason for the vehicle's slowness may not be evident, ond vehicle makes an evasive maneuver, it may place yet
but it could simply be a precautionary action, or a result of another road user (a third vehicle) in danger of a collision.
congestion or some other cause beyond the intersection. This type of event is called a secondary conflict. Nearly
always, the secondary conflict will look much like a slow-
Opposing Left-Turn Conflict vehicle, same-direction conflict (or a lane-change conflict,
An opposing left-turn conflict occurs when an oncoming which has not yet been described). The difference is that,
vehicle makes a left turn, thus placing a second vehicle, in a secondary conflict, the conflicted vehicle is respond-
going in the other direction, in danger of a head-on or ing to a vehicle that, itself, is in a conflict situation (see
broadside collision. The conflicted vehicle brakes or examples in Figs. H-12 and H-13).
swerves, then continues through the intersection (see By convention, do not count more than one secondary
Fig. H-6). conflict for any initial conflict. Even if a whole line of cars
By convention, in this and the following conflict situa- stops because the first one turns left, count it as just one
tions, the conflicted vehicle is presumed to have the right- secondary conflict.
of-way, and this right-of-way is threatened by some other
road user. Situations such as a "conflicted" vehicle placed Other Types of Traffic Conflicts
in danger of a collision because it is running a red light, for
example, are not treated as traffic conflicts. Under certain special conditions, one may be asked to
watch for and record other types of traffic conflicts. One
of these is the lane-change conflict, which occurs when a
Right-Turn, Cross-Traffic-From-Right Conflict
vehicle changes from one lane to another, thus placing a
A right-turn, cross-traffic-from-right conflict occurs when second, following vehicle in the new lane in danger of
a vehicle approaching from the right makes a right turn, a rear-end or side-swipe collision. The conflicted vehicle
thus placing a second vehicle in jeopardy of a broadside or brakes or swerves, then continues through the intersection
rear-end collision. The second vehicle brakes or swerves, (see Fig. H-14). However, if the lane change is made by
 87

_
Ii iHL_

II Ix I
igure H-3. Left-turn, same-direction conflict. Figure H-4. Right-turn, same-direction conflict.

'igure H-S. Slow-vehicle, same-direction conflict. Figure H-6. Opposing left-turn conflict.

a vehicle because it is in danger, itself, of a rear-end colli- from the left makes a right turn across the center of the
sion with another vehicle, the following vehicle in the new roadway and into an opposing lane, thus placing a vehicle
lane is said to be faced not with a lane-change conflict situa- in that lane in danger of a head-on collision. The conflicted
tion, but with a secondary conflict situation, vehicle brakes or swerves, then continues through the inter-
Another unusual conflict is the right-turn, cross-traffic- section (see Fig. H-is). This conflict is sometimes ob-
from-left conflict. It occurs when a vehicle approaching served when the cross street is narrow, or when large trucks
88

Figure H-7. Right-turn, cross- traffic-f rom-righ t conflict. Figure H-S. Left-turn, cross-traffic-from-right conflict.

Figure H-9. Thru, cross-traffic-from-right conflict. Figure H-JO. Left-turn, cross-traffic-from- left conflict.

or buses make right turns. Note that the first vehicle must turning, conflicted vehicle (which has the right-of-way) in
cross the center line for there to be a conflict! danger of a broadside or rear-end collision. The conflicted
An op posing right-turn-on-red conflict can only occur at vehicle brakes or swerves, then continues through the inter-
a signalized intersection with a protected left-turn phase. It section (see Fig. H-16).
happens when an oncoming vehicle makes a right-turn-on- There can also be pedestrian conflicts. They occur when
red during the protected left-turn phase, thus placing a left a pedestrian (the road user causing the conflict) crosses in
 89

igure H-il. Turn, cross-traffic-from-left conflict. Figure H-I 2. Slow-vehicle, same-direction secondary conflict.

H. H
I Ix I Ix

Figure H-I 3. Right-turn, cross-traffic- from-rig/it secondary Figure H-i 4. Lane-change conflict.
conflict.

front of a vehicle that has the right-of-way, thus creating of the intersection are not considered to create conflict -
a possible collision situation. The vehicle brakes or swerves, situations if the movements have the right-of-way, such as
then continues through the intersection. Any such crossing during a "walk" phase.
on the near. side or far side of the intersection (see Figs.
Traffic Volumes
H-17 and H-18) is liable to be a conflict situation. How-
ever, the pedestrian movements on the right and left sides Along with a traffic conflicts survey, traffic volume
all

I I I

Figure H-15. Rig/it-turn, cross-traffic-from-left conflict. Figure H-16. Opposing, right-turn-on-red conflict.

_JHLL ii _

I Ix I Ix
Figure H-17. Pedestrian, far-side conflict. Figure H-18. Pedestrian, near-side conflict.

counts, turning movement counts, or traffic classification one-way volume. All vehicles moving into the intersection
counts are also made. These cannot be done by a traffic in the same direction as the traffic being observed are
conflict observer when he is making the traffic conflicts counted. More is said about this in the section onsurvey
counts. They are done either by another person or by the procedures.
same person before or after the traffic conflicts observa-
tions. Each agency has its own procedures for doing these STARTING THE TRAFFIC CONFLICTS SURVEY
counts, so they are not described here.
However, one kind of traffic volume is observed and A traffic conflicts survey includes making conflicts counts
recorded along with the conflicts. That is the mainline, along with collecting other data needed to make a corn-
 91

plete study package. The collected data can be used to ture makes observation from the car uncomfortable. Drink-
answer questions about safety and operational problems to ing water is highly desirable, especially on hot days.
recommend improvements and treatments, or to show the Finally, a complete set of data collection forms, includ-
effectiveness of improvements already implemented, as in ing extras, are needed as follows: Physical Inventory, On-
the before/after study. The traffic conflicts survey can also Site Observation Report, Traffic Volume Counts, and Inter-
be a portion of a larger study within a safety upgrading section Conflicts. The purpose and use of these forms are
program. In this case, traffic conflicts data are used along discussed in the next section.
with traffic accident data, signal warrants, capacity analysis,
and other investigations needed to recommend improve- Arriving at the Study Location
ments.
When the observer arrives at the study location, he
Preparing for the Survey should familiarize himself with the location. Is this the
intersection to be studied? Which approach legs are to be
Before leaving for the study location, all of the equip- used for conflicts observation? How many laness are there?
ment and materials needed for the survey should be assem- How are they used, etc.
bled. These include: Observation points are selected next. For conflicts ob-
servation, a location from 100 to 300 ft (30 to 91 m) up-
Observation procedures. stream of the intersection and on the right side of the
Schedule of survey locations. approach is usually best. This depends on vehicle speeds
Map showing location of test sites. and approach geometry. Figure H-19 shows a diagram of
Supervisor's telephone number. an intersection with the observer locations marked by
I.D. card. circled X's. At high-speed locations, a location should be
Mechanical count board. picked farther away from the intersection so that all actions
Tablet. and maneuvers relating to the intersection can be observed.
Pencils. The observer should park his vehicle off the roadway.
Watch. Figure H-20 shows a typical situation. One should not use
Stop watch. a vehicle that could be mistaken for a police or other offi-
Camera with film (at least 20 exposures for a 4-leg cial car. Traffic will surely be affected by this.
intersection).
If an adequate parking place is not available, the ob-
Folding chair (optional).
server will have to perform the study outside of his vehicle.
Water.
For this reason, he should have a folding chair. Without
Data collection forms.
a comfortable seat, fatigue will soon make concentration
The observation procedures are described in other sec- difficult. If possible, he should be hidden or inconspicuous
tions of this manual. The schedule of survey locations from the traffic on the study approach. In Figure H-21, the
should list future test sites, which could also be used in observer is behind the second tree. When a suitable loca-
emergency situations—that is, if for some reason data can- tion is not available along the right side of the approach,
not be collected at the scheduled site. If the observer is not he will have to observe conflicts from the left side. Ob-
familiar with the other locations, they should be noted on server comfort and, of course, safety are also deciding fac-
a current map. tors. When summer temperatures exceed 90 F (32 C), the
The project supervisor should be telephoned whenever heat and sun can create health hazards. For this reason,
the observer is unsure of the procedures or something un- one may wish to find a shady place from which to observe.
usual occurs. Also, the observer should always carry valid Observation positions are not as critical during a stan-
identification in case property owners or the police are con- dard traffic volume count, if required. Such counts usually
cerned by his presence. Questions should be answered include all turning movements, on all approaches at the
courteously but quickly so that one can maintain concen- same time. The observer will then usually need an observa-
tration on the study. tion position much closer to the intersection. He may even
The count board should have enough mechanical count- be able to record volumes from a corner of the intersection.
ers to record traffic volumes on all approaches of a stan- However, if conflicts and volumes are counted at the same
dard 4-leg intersection (i.e., three counters per approach). time by two observers, the volume observer should remain
This will be more than enough to record the most common inconspicuous also.
traffic conflicts. Any additional counts can easily be re- Once the observation positions are determined, the re-
corded by hand. The count board can also be used, of quired number of data collection forms (this depends on
course, to do traffic volume counts, if requested. the amount of data to be collected) should be prepared.
Space pencils and a tablet for additional notes are a All heading information should be completed and double
necessity. A watch is needed to note count start times, and checked before any data are collected. The count board
a stop watch is needed to accurately determine signal tim- has to be "zeroed." If there is more than one observer,
ing. To adequately record the physical features of the study watches will have to be synchronized. (Make sure that
location about five photographs per approach are needed. watches and stop watches are wound.)
A folding chair should be included as part of one's equip- For uniformity in the field study, observations should
ment in case one is unable to use his car or if the tempera- always start at the prescribed times, and count periods
92

LEE5 SUMMIT RD.

5ERVICE
STAT1014

-mv
9T
0 _
[K4NSAs
ciir

Figure H-19. Typical site diagram

should always be of the prescribed length. To start on CONDUCTING THE SURVEY

time, the observer will have to arrive at the test site at least After arriving at the site, the observer should make cer-
30 min before starting to count. This is the minimum
tain that he has proper supplies, checks basic traffic op-
time required to become familiar with the intersection and
erations, etc.; he should then be ready to conduct the sur-
prepare for data collection. If there is much auxiliary data
vey. The following presents the time schedule for conflicts
to be collected, the observer might have to arrive 1 hr
counting, how to use the count board, the forms to use and
before the count start time.
how to use them, how to deal with special problems, and
A very important preliminary activity is to watch the safety precautions.
traffic, and become familiar with the major traffic move-
ments, the signalization characteristics, and any unusual ac-
Time Scheduling
tivities. Also, locations of nearby driveways, parked ve-
hicles, or other features that may cause traffic problems Whether the survey lasts for several hours or. several
should be noted. days, the observation process is conveniently thought of as
 93

1 •, ,- , ; - 'a

-
Figure 11-20. Conflict obsener in vehicle.

Figure 11-2 / C un//ut ui's yr rev in c/in r.

being in 1-hr blocks. The traffic engineer in charge will two observers, one will be alternating approaches every
determine how many, and which, hours are to be used for half-hour.) Then, the process should be repeated during
succeeding hours, as required. Usually, after every 2 or
this data collection.
3 hr of a survey, a break will he scheduled.
For illustration, suppose a 1-hr block begins at 0800
For consistency. it is best to start counting exactly at the
(8:00 am.). Table H-I shows how the 1 hr is split up into
hour and half-hour marks. An exception can be made for
several activities. During the first 20 mm, the observer signalized intersections, where the signal cycle may not he
should observe and count traffic conflicts from one of the in phase with one's watch. In such cases, one should start
designated approach legs of the intersection. After 20 mm, observing after the hour or half-hour marks the first time
he should stop and record his counts on the special forms the signal turns red for his approach. Then, a stop watch
and then move to the opposite approach and prepare for should he used in order to observe for just 20 min after the
a second count starting on the half-hour. The same pro- start. This should be coordinated with one's partner (if
cedure should be followed on this approach. (If there are any), so both are counting at the same time.
 94

TABLE H-I counts). The one that is most frequent, clearly, is the
A ONE-HR TIME SCHEDULE traffic volume count on the approach leg. The counter used
for this event should be positioned in the most convenient
0800 Start observing conflicts place—maybe the lower right corner.
The other kinds of events, listed in order of decreasing
0820 Stop observing conflicts, and record
counts on data form
frequency. are as follows:

0825 Move to Opposite approach leg I. Most frequent:

Left-turn, same-direction conflict
0830 Start observing conflicts Slow-vehicle, same-direction conflict
0850 Stop observing conflicts, and record
Right-turn, same-direction conflict
counts on data form Less frequent:
Opposing left-turn conflict
0855 Move to opposite approach leg
Right-turn, cross-traffic-from-right conflict
Left-turn, cross-traffic-from-right conflict
Thru, cross-traffic-from-right conflict
Using the Count Board Left-turn, cross-traffic-from-left conflict
Thru, cross-traffic-from-left conflict
At nearly all intersections a mechanical count board is Least frequent:
necessary to "record" conflict counts. Some traffic events All secondary conflicts
happen very rapidly, so one's attention must be focused on All special conflicts
the road and vehicles rather than on pencil and paper. Af-
ter some practice, the count board allows one to record the It is recommended that the count board be used for the
events 'by touch." without looking down. most frequent events. The least frequent events can be
The type of count board is not important. One designed written directly on the forms when they occur.
for making traffic volume counts is very acceptable: an- Before a 20-min count is begun, the observer should
other is shown in Figure H-22. niake sure that all counters are reset to zero. After the
The mechanical counters should be used to record the count is completed, all figures should be recorded from the
most common occurrences (the ones with the highest counters to the form, and double checked. A common
error is to reset the counters before recording the results,
which obviously "erases" all the hard work. Do not make
that mistake!

The Conflict Form

A recommended, general-purpose form for recording

traffic conflict counts is shown in Figure H-23. The head-
ing information, which is self-explanatory, should be filled
out in advance. The diagram in the upper right corner indi-
cates the approach-leg numbering system. That is, traffic
approaching from the north is on leg 1, etc. A separate
traffic conflicts form should be used for each approach leg.
The start time should be recorded for each 20-min count
in the first column. If, for any reason, the count was for
oilier than twenty ninutes, record actual time in the left
margin. The results should be copied from the count board
into the proper columns for the form, making sure all
marks are legible.
The common types of conflicts each have separate col-
umns for recording. (Note, however, that at signalized
intersections the cross-traffic conflicts may not be very
common.) If any special types of conflicts are observed
very often, or if the traffic engineer requests any extra kind
of counts, additional columns are provided for their re-
cording. It should be macic certain that they are clearly
labeled.
Sometimes, conflicts of a severe nature will be observed,
such as obvious "diving" of the front end of a vehicle,
squealing of brakes, rubber skid marks, violent swerves,
honking of horns, shaking of fists, etc., and even collisions.
Figure 11-22. Ira//ic conflicts count board. Special note should be made of such conflicts. These notes
 INTERSECTION CONFUCTS

Location Lg Number

Day Date Observer

Conflict - C, Secondary Conflict - SC

COUNT

TIME APPROACH,00MEMENEEMEN

_______•u•u••••uuu•••uuu•••u

andt.]u1.it4sWt

Figure H-23. Traffic conflicts recording forms.

96

are very important, especially if such severe conflicts occur red times for all approaches. Any separate phases, such as
very often. a left-turn phase, should also be noted. A good place to
Try to determine the causes for same-direction conflicts. record this information is on the intersection diagram.
Is the problem just past the intersection (a driveway, shop- If the signal is traffic-actuated, signal timing and phas-
ping center, traffic back-up etc.), a "blind" spot, unclear ing may be more difficult to determine. The traffic signal
or missing pavement markings, erratic signal operation or plan should then be obtained. Traffic-actuated signals will
what? Also, by using one's best judgment, comments have loops cut into the roadway, metal pressure pads near
should be added about what is thought to be wrong with the stop line, or sensors mounted on utility poles.
the traffic operations at the intersection and how they can Photographs should also be taken of the intersection.
be improved. Usually this should include five pictures per approach,
On completion of a survey, or portion thereof, the counts taken from the driver's viewpoint. Three pictures would
should be added in each column. If one is working with a be taken from positions 200 ft (61 m), 600 ft (183 m), and
partner, data sheets should be exchanged and each other's 1,000 ft (305 m) upstream of the intersection. Two cross-
forms checked for completeness and accuracy. Otherwise, corner photographs should be made from the stop line as
one's own forms should be double checked, making sure though a driver were looking left and right at the approach-
all the heading information is correct, all blanks are filled ing cross traffic. Photographs are the only means to ensure
out, and all entries are clearly readable. that no physical devices or their details are overlooked.
They also help decision-makers, who may never see the
Collecting Auxiliary Data intersection, gain a better understanding of the physical
and operational characteristics.
Traffic conflicts counts are not meaningful unless they During the course of the day, the observer should note
can be related to the existing site conditions. The site data any apparent operational and safety deficiencies of the in-
needed may include: physical inventory, intersection dia- tersection. This includes both causes of problems as well as
gram, signal timings, photographs, on-site observation re- possible solutions. A formalized procedure for this is shown
port, and traffic volumes. Each agency has standardized in Figure H-25. An experienced conflict observer will be-
procedures for collecting most of these data, but general come fairly good at recognizing these deficiencies. By for-
guidelines are given here, using example forms. malizing the procedure, full advantage can be taken of his
The signs, signals, pavement markings, type of inter- expertise.
sections, and other useful information are noted in the If a total traffic volume count is to be used with the traffic
physical inventory. One also needs to include speed lim- conflict count, it is best performed at the same time. Gen-
its on all approaches. This information creates a perma- erally one person can collect the volume data. If volumes
nent record and is especially useful for before/after stud- are high or pedestrian data are needed, two people may be
ies. A typical "Counters Inventory of Existing Highway required. A typical traffic volume count form is shown in
Features" is shown in Figure H-24. This form was estab- Figure H-26. It is important that the correct approach leg
lished as part of the General Motors TCT and is used by names and numbers are recorded.
the States of Ohio and Washington. The physical inventory, intersection diagram, signal tim-
A sketch (intersection diagram) should be made of the ing, and photographs can all be completed before conflicts
test site. This should be drawn on ruled paper in order to observation starts. This information takes at least an hour
include the intersection geometrics along with all signs, sig- to collect, however. If all of this information cannot be
nals, pavement markings, channelization, driveways, houses, collected beforehand, it can be obtained during a break
businesses, utility poles, trees, shrubbery, and any other from counting or after completion. The On-Site Observa-
physical objects of interest. The diagram should be made tion Report should be completed after the conflicts count
to scale and include a north arrow, street names, and ap- is finished. At this time, the observer should have a good
proach numbers. A sample intersection diagram is shown understanding of how the intersection operates and some of
in Figure H-19. the problems associated with it.
If the intersection diagram might be used to redesign the
intersection, an accurate drawing is needed. The best Special Problems

method is to first have the intersection surveyed. This, of Changes in the weather may interrupt or postpone the
course, is costly and takes at least a day' to do. A good conflict study. Normally, observations are not performed
scale drawing can also be made from data collected by during inclement weather, such as rain, snow, or fog. If
using a device to make accurate measurements. For a the roadways are completely wet or visibility is reduced,
standard intersection this may take as much as 2 to 3 hr observation should be stopped. The observer should do
using a measuring device such as the "Walking Wheel," other tasks, such as collecting auxiliary data, until roads are
"Rolatape," or "Measure Master." However, it is prefer- in a near-normal condition or until he or his supervisor
able to record all devices and changes on an existing draw- decides to postpone data collection for the day. The traffic
ing such as a construction djawing. If the intersection is engineer should be contacted if there is any doubt.
signalized, a good scale drawing can usually be obtained Other problems may also occur. Before a site is sched-
from the local traffic department. uled for study, it should be determined if any construction
At signalized intersections, the signal timing should be is planned that could alter the normal traffic-flow patterns.
recorded. This includes cycle length and green, amber, and Unscheduled emergency repairs by the street department
 Cheek Reel REGIILATOAY AARIIII1C IMlIDE
Fill ml TYPE INIERSECTIPY -
TORN BESTIr)-
lIO.I TEE T1 510 HO LONE OR
-
STOP SIGUOL JIIIICVIVH 011011
Check Ael 'V' —< 7 LEFT 010111 AIITROL OIlIER OTIIRH OSSEATLY' I
1IREE- ArID All - VTIIF.O
HEllS MIEAD VTPII
TARN TOhil ARNOlD TITIYAI, MIl,E301
PEfflOD
BEFORE 2
h
RIGIIT ANGLE
-4-. 3 -
SIGNS .... - - - _,,,_. - - - - - - . - -
- ASSEYNLY ASIFYYLYI - -

TANGLE EIFFTET
AFTER
IIJ SEEPED
- T
_71_ 5
-

P01 InI SKECEA OFFSET H ,-

INTERSECTION
IXULTILEG -p--- , -
APPROACH.
ROTARY __ N
____________________ I A 3 3 5 6 I 6 1 A 4 C 0 E F A II I I H L IT N 0 P Q R S I A A P

OTHER ec 31-35
2 1 Fill Im Ch bAT 2 _ - - -

CONFLICTS
1 COUNT DATE Tn
Check One,
MONTH PAY TEAR
Li TYPE AERBOAGO 5130-50
FLAREOIONE LAIIE ADDED' I

In [y] ~-
11 11M FLARERITWA OR MARE ADDED' 2
210 FLARE 3 —
ij - ------- ---- -

j Check Onel
ENVIRONMENT
flfl ICheck One'
22
IYRAL ROSINESS DISTRICT I MEDIAN
INDUSTRIAL 2 HOME Ilocluding Mounlable- 2" I - SIGNALS a
RETAIL STRIP DEVELOP.' 3 PAINT OR rVMNT TEXTURE 2 -

MRIINTAOLEI --2" 3
RETIOEUTIAL I
3ORRIER TYPE 4 -
RURAL IS
A B C II E F G H I I N L II j II A P
I Ism-s
LBOE
LEFT-TORII THRA -

(Fill Inl
CHANNELIZATION 23
lChechUp
3 Ii . i I
NONE llncluding Nm,eIabIe 21 I PeLae 1-66 ...---- -
.. .
NO. INTERSECTIOTIv -
5
WITHIN 0.5 MILL PAINT STRIPED I
7
MOUNTABLE
14
IFiII A l BARRIER TYPE 4 -

DISTANCE 0.1 MA TO
NRARESTIN1ERIUN'I I
iJ
L IChe
ck And
RIGHT-TORN THRU
CHANNELIZATION 24 TRAFFIC CONFLICTS UAF,A SHEET A"
ICheck A',el MARKINGS COIINTERS IFIYET1TORV
TDAPFIC CONTOOL OF
NEAREST INTERSECTION
r9 NONE (Including Menelakle . 2') I -
OF
PAINT STRIPED 2 EAISTIAEIIIGHWAV FEATIIT1ES
SIGNAL I1
INTERSECTION TIAAIE:
MOUNTABLE 3
SlAP SIGN 2
BARRIER TYPE 4 REPORTED BY: -

YIELD SIGN
31 11-12
J
AB C I' E F G H
PAl.T.FEII TA:
UTIlE L5BL
Pill lel 7/3-Il
NUMBER OF LANES ICheckIJ
375-10
4- OUTLYING HIlT 01ST. (Include Th,u N Twniog 4 Il-TA
4: FARM. ASIDE VELOPED. ETC
r LanesI Pe, Lane'
19-VA
- - - - -- - -

* UPSTREAM ON THIS APPROACH

Figure H-24. Counters inventory of existing highway features.

 Form #1 Form #2
ON-SITE OBSERVATION. REPORT ON-SITE OBSERVATION REPORT 00

LOCATION CONTROL No ! (See Comments)

DATE TIME Is channelization (islands or paint markings) inade-

quate for reducing conflict areas, separating
traffic flows, and defining movements? -
OPERATIONAL CHECKLIST: Does the legol parking layout affect sight distance,
through or turning vehicle paths, or traffic flow? -
No Yes (See Comments) Do speed limits appear to be unsafe or unreasonable? -
Is the number of lanes insufficient? -
Do obstructions block the drivers view of opposing Is street lighting inadequate? -
vehicles? Are driveways inadequately designed or located? -
Do drivers respond incorrectly to signals, signs, or Does the pavement condition (potholes, washboard,
other traffic control devices? or slick surface) contribute to accidents? -
Do drivers have trouble finding the correct path
through the location? COMMENTS:
Are vehicle speeds too high? Too low?
Are there violations of parking or other traffic Operotionol--'O' and Item Number
regulations? Physical--P and Item Number
Are drivers confused about routes, street names, or
other guidance information?
Can vehicle delay be reduced?
Are there traffic flow deficiencies or traffic con-
flict patterns associated with turning movements?
Would one-way operation make the location safer?
Is this volume of traffic causing problems?
Do pedestrian movements through the location cause
conflicts?
Are there other traffic flow deficiencies or traffic
conflict patterns?

PHYSICAL CHECKLIST:

Can sight obstructions be removed or lessen?

Are the street alignment or widths inadequate?
3.. Are curb radii too small?
Should pedestrian crosswalks be relocated?
Repainted?
Are signs inadequate as to usefulness, message,, size,
conformity and placement? (see MUTCD)
Are signals inadequate as to placement, conformity,
number of signal heads, or timing? (see MUTCD)
Are pavement markings inadequate as to their
clearness or location?

Figure H-25. On-site observation report.

 TRAFFIC VOLUME COUNTS

*N
Location 2

Day Date Time Period

Observer

Leg No. Leg No. Leg No. Leg No.

COUNT
START
TIME
(MILITARY)

TOTAL
NOTES AND COMMENTS:

Figure H-26. Traffic volume count recording form. 0

100

or utility companies will also disrupt flow. If this occurs, abruptly to avoid a southbound vehicle he did not see until
the-observer should speak with the person in charge to the last minute. No conflict. This common situation often
learn the extent of the work and how long it will take. leads to accidents, however. Especially on 4-lane roads, the
Other disruptive events such as accidents, stalled vehicles, oncoming center-lane vehicles may be stopped waiting to
police arrests, etc., will also occur from time to time. The turn left, hiding through-vehicles in the outside lane. But,
observer should always have a contingency plan that can unless there is a left-turn phase, the through-vehicles have
be adopted when problems occur. If in doubt, telephone the right-of-way. If the conflicted vehicle does not have the
the supervisor. right-of-way, it is not classified as a conflict. However, if
this situation is observed often at an intersection, record it
Safety Considerations as a note.
On a 4-lane street, an oncoming vehicle makes a left
An important item that should not be overlooked is
turn, causing drivers in both northbound lanes to brake.
safety. Any time a person is working near moving traffic
Although this could be debated, it appears most logical to
there will be some drivers that will not see him. Of course,
count this as two opposing left-turn conflicts. Although
when collecting volume and conflict data, the observer
there was only one instigating vehicle, accidents could have
should be hidden. But when walking along the roadway
occurred with either northbound vehicle had they not
or taking pictures from the middle of a lane, the observer
reacted. Also, this is not a secondary conflict situation,
must be seen. He should always wear clothing that will
because the two northbound vehicles reacted independently
attract attention. All street and highway agencies main-
to the left turner, not to each other.
tain a supply of florescent orange vests for this purpose.
If the observer is to enter the roadway, l should do it dur- A car is stopped with a flat tire on the other side of
ing a gap in the flow of traffic. He should not try to stop the intersection, blocking the lane for a half-hour. Mean-
or direct traffic. while, northbound traffic is slowed considerably because it
is forced to maneuver around the disabled vehicle. Fre-
CASE STUDIES
quent slow-vehicle and secondary conflicts are noted. Go
ahead and record the conflicts, unless traffic becomes to-
It is not possible to describe every possible traffic event tally congested. But, make prominent notes about the situa-
and classify it into one of the conflict categories. One will tion and, if possible, explain it personally to your super-
have to judge most situations using experience, and by ap- visor. He may decide not to use the data, but it is better to
plying the basic principles presented earlier. Studying the record the data, even if they will not be used, than to miss
selected case studies given here may help the job become important insights about the traffic operations.
easier. Same as 6, except traffic becomes totally congested,
In all of these examples, assume the observer is on the and nearly every vehicle brakes one or more times ap-
south approach, viewing northbound traffic from the rear proaching or going through the intersection. The traffic
as it approaches the intersection. In each case, a situation conflicts technique, as such, does not appear suitable dur-
is first described and then interpreted. ing periods of congestion. However, the existence of traffic
The signal turns red for the northbound traffic, but a congestion is possibly indicative of operational deficiencies.
driver apparently does not notice it until the last minute, During such times, cease making formal conflict counts,
then slams on the brakes. The interpretation depends on but carefully note any apparent causes for the congestion
the other traffic. If, as would normally be the case, the (it could be simply heavy traffic) and how long it lasts.
intersection is empty when braking begins, there is no con- Every 10 min or so, a city bus slows and stops just
flict. The driver is just responding to the signal. But if a across the intersection to discharge passengers. Cars behind
westbound vehicle is in the intersection, classify it as a it are forced to brake or swerve. Record as slow-vehicle
thru, cross-traffic-from-right conflict. This would probably conflicts. But, it is extremely important to note the cause.
be rare, and you should make a special note about it on This may or may not be judged a hazardous situation—
your recording form. that is for the traffic engineer to decide—but make sure he
A car on the right (east) approach stops, starts to pull gets the information.
out, then stops abruptly because the driver sees a north- You hear a squeal of brakes behind you. Turning,
bound vehicle that just passed you. This is not a conflict you see a heavy, slow-moving truck and, behind it, the car
from your perspective. Only when a northbound vehicle that had just braked. No conflict. You are counting only
reacts to an impending collision is there a conflict. If, how- the events between you and the intersection. The purpose
ever, the northbound vehicle also braked or swerved and of the study is to learn more about the intersection.
the car from the right had pulled far enough forward to be Chances are that things behind you (such as the slow-
in his path, you would record a conflict. moving truck) have little to do with the intersection itself.
A northbound car slows and turns right. Another car, But, if you think it was due to the intersection (for exam-
right behind it, brakes severely and then it, too, turns right. ple, the truck was moving slowly because the light was
Do not call this a conflict. You do not know if the second going to change shortly), make a special note.
vehicle braked because of the first vehicle or because he There is a fast-food restaurant 200 ft past the inter-
was going to turn. It is a conflict only if the second vehicle section, and many vehicles slow to enter the drive. Often,
proceeds through the intersection. other vehicles are then forced to slow during, or after, the
A northbound driver begins a left turn, then stops time they cross the intersection. In theory, you should re-
 101

cord these incidents as slow-vehicle conflicts if the braking even a right-turn, cross-traffic-from-right conflict. If this
vehicle is in (or on your side of) the intersection, otherwise does not happen very often, the classification probably does
not. In practice, do likewise. But, in either case, if it hap- not matter very much. It is preferred to record it as a
pens frequently, make notes about it. Although there may slow-vehicle conflict, then to note the cause.
not be an intersection problem, you may have located a
12. A southbound cab enters the intersection, then
driveway problem that bears on how the intersection
makes a U-turn and heads north. In so doing, another
operates.
northbound vehicle was caused to brake. If this happens
11. A car parked at a meter ahead of you, pulls in front
very often, make up a separate column, define these as
of another vehicle, causing it to brake. This is a conflict;
the question is, what kind? Arguments could be made for U-turn conflicts, and count them. Otherwise, record them
calling it a slow-vehicle conflict, a lane-change conflict, or as slow-vehicle conflicts and note the cause.

APPENDIX I

INSTRUCTOR'S AND ENGINEER'S GUIDE

INTRODUCTION their different outlook brought about by police training and

experience.
This guide is a supplement to the material in Part I of
NCHRP 219 and the procedures manual in Appendix H
How to Train
(Part II). This guide presents suggestions to persons who
wish to train traffic conflict observers, and/or to use the If an agency is just beginning a traffic conflicts program,
results of the observations. It is assumed that the instruc- it will probably want to train several persons at the same
tor has access to the documents just mentioned. Their de- time. This is the ideal arrangement. With a group, more
tailed contents will not be repeated here to avoid excessive effort can be devoted to planning, acquiring audiovisual
repetition. aids, etc., than is usually possible with just one trainee.
With a group, a combination of class work, group observa-
TRAINING tion, and group discussions can be effectively used along
with individual tutoring. It also enables the use of com-
Whom To Train
parative analysis between observers to determine who needs
Persons who will be conflicts observers must be ex- special attention or what topics need additional emphasis.
tremely conscientious and trustworthy. They will be on If just one person is to be trained, the apprentice con-
their own much of the time, without supervision. They cept is probably best, where the trainee works with an
must be trusted to record what they see, and not to fabri- experienced observer for two weeks or more, under the
cate data. general direction of the traffic engineer or other person in
The job is both demanding and tedious. Once learned, charge.
the observational method is not difficult. Some people will
find it boring and seek greater challenges. The ideal ob- Training Program
server is one who can maintain his alertness and enthusiasm
for the task, and who can find challenge in it on a day-to- Assuming groups of persons are to be trained together,
day basis. it is best to use a formal schedule. A suggested program is
Age and sex present no inherent barriers. The majority given in Table 1-1. In summary, about two weeks of train-
of persons are trainable. There may be some for whom the ing are recommended, with most of this time devoted to
task is too great, but there are just as likely to be some for field practice and review. (The agency may want to bring
whom the task is too easy. Most importantly, some per- in a consulting instructor from outside the agency. In this
sons will have such a fixed opinion about driving and traffic case, the schedule can easily be modified so that the con-
behavior (probably reflecting their own habits) that they sulting instructor need be on site for only the first 5 days,
will be psychologically unable to accept the concepts of with the second 5 days devoted to field practice under the
traffic conflicts that must be used. Such persons should be supervision of the agency traffic engineer or equivalent.)
identified before or during the training and given alterna- As will be noted, the period could be shortened somewhat
tive assignments. if the trainees are already experienced traffic technicians or
Persons presently employed as traffic technicians or para- equivalent.
professionals usually make good observers. Some agencies The first day is devoted to activities of an introductory
report that police officers may not be as good, because of nature, and will not be needed for some groups. It includes
102

TABLE I-i ommended. Also, the use of videotapes of conflicts might

TRAINING SCHEDULE be considered as a classroom exercise and a focus for
discussion.
Day 1. Introductory Remarks; On the fourth day, the trainees should be allowed to
Orientation to the agnecy;
Orientation to the training program;
work in pairs rather than in larger groups, but the partners
General background on traffic safety; should be alternated. This fosters the interchange of ideas
Traffic Counting; between trainees that might not otherwise occur. Also,
Turning movements; plenty of time should be allowed for discussion and con-
Use of mechanical counting boards;
Introductory field work. sideration of case studies. The trainees by this time should
be asking very perceptive questions based on their practice
to date that should be shared with all trainees. If time
Day 2. Same-direction conflicts;.
allows, the more specialized traffic conflicts (pedestrian,
Group field instruction (signalized intersection);
Discussion. lane change, etc.) should be introduced; otherwise, this
should be done early on Day 5.
The fifth day should be devoted to a simulated conflicts
Day 3. Cross-traffic conflicts;
Group field instruction (unsignalized intersection);
count, with some monitoring by the instructor but without
Discussion; full-time supervision. The trainees can again work in pairs,
Use of Video Tape. and should follow normal field practices such as maintain-
ing a certain time schedule, alternating legs of the inter-
Day 4. Small group field practice; section, completing and checking the intersection conflicts
Question and answer session; forms, etc.
Special conflict types; The sixth day is devoted primarily to the other forms
and procedures to be used in the field, including the inter-
Day 5. Simulated, limited conflict count (8-hour day). section inventory, photographs, etc. This may not be neces-
sary if the trainees are traffic technicians already accus-
tomed to these procedures. Also, on the sixth day, some
Day 6. Use of auxiliary forms;
Data analysis. time can be devoted to analysis procedures (completing the
Intersection Conflicts Analysis form).
The seventh day should be a full-scale conflict count,
Day 7. Simulated full conflict count (8-hour day)
using all auxiliary forms. Observers should work inde-
pendently (or with a partner on the opposing leg). No
Day 8. Review; supervision is suggested, but the instructor may want to
Analysis of Day 7 data. stop by the site once or twice to answer questions.
Day 9,10 ... More field practice. Day 8 is set aside for a full review of all activities to
date, with emphasis on problem areas. It would also be
instructive to analyze (summarize) the data collected the
previous day and provide some interpretation. If any of
orientation to the agency, to traffic activities in general, to the trainees is experiencing individual difficulties with any
the program, and to traffic counting. The field work on this of the concepts, this would be a good time to provide some
day should be fully supervised and should emphasize traffic special attention.
counting procedures. It should also involve general obser- From here on, additional practice is recommended—at
vations and discussion of traffic behavior, certain driver least 2 days. Some of the data collected at this time may
actions, potentially unsafe practices, etc. The idea is to get actually be usable, so if there are particular intersections of
the trainees thinking about how people drive, and why. interest, they should be used.
The second day is directed to the simplest and most com-
mon class of traffic conflicts, the rear-end or same-direction Visual Aids
conflicts and the opposing left-turn conflicts. They should
be introduced in the class room using lectures, films, slides, Visual aids are highly recommended, especially early in
sketches, or other audiovisual aids available, as described the training. A film and a set of 35-mm slides were pre-
subsequently. The basic principles should be emphasized. pared as part of NCHRP Project 17-3, and are available
Then, one or more convenient, simple, uncongested sig- through their offices. These materials essentially parallel
nalized intersections should be used for supervised field the definitions and course outline given here, and are self-
practice. Using signalized intersections avoids most cross- explanatory.
traffic conflicts, so focuses attention on the conflicts of Videotape is often quite useful in explaining real-life
interest. Time should be reserved late in the day for dis- situations. With some practice, it is possible to tape con-
cussion and questions, probably back in the classroom. flicts as they happen, at a variety of intersections, for later
On the third day the cross-traffic conflicts are introduced, use in training.
and field practice is conducted at an unsignalized inter- A chalkboard or equivalent is nearly indispensable as a
section. The format for this day is similar to that of Day 2 training aid. It enables sketches illustrating various situa-
except that the use of actual conflict count sheets is rec- tions to be made quickly.
 103

Case Studies conflicts of interest, the traffic volumes, the type of inter-
Case studies are probably the best way to illustrate the section, and the precision required.
subtleties of the TCT. A dozen are given in the procedures Table 1-2 provides some guidance for various types of
manual in Appendix H. Further,' each of these can easily conflicts. The research report contains additional informa-
be modified by changing timings, direction, etc., to create tion on traffic volumes, intersection types, and how to ad-
still other examples. In addition, the trainees are sure to just for different precision and confidence levels.
pose questions in the form of "case studies" based on their
field observations. Full use should be made of these as UNUSUAL GEOMETRICS
teaching aids in the classroom. And inquisitiveness should The basic operational definitions and recording forms
be encouraged among the trainees—stimulate questions of described in the research report refer to relatively "stan-
the "what if" variety. dard" intersection geometrics. Certain modifications will
be required for other geometries. Suggestions are given
WHEN TO USE THE TCT here for some of the more common departures from
Applying the traffic conflicts technique is somewhat time- normality that may be encountered.
consuming, so it should not be used indiscriminately.
Rather, the TCT should be applied only for one of the Left-Turn Bays
several well-defined reasons. If an approach leg contains a left-turn bay, more lane
The TCT is an excellent tool for diagnosing safety/ changing than usual will be observed. It should be made
operational problems of intersections that have previously certain that these are not mistakenly recorded as swerves
been singled out for attention, usually because of an ad- indicative of rear-end conflict situations. However, the ob-
verse accident history. It is not appropriate, however, for servers should be alert for lane-change conflicts, which are
identifying hazardous intersections because of the cost per otherwise rather rare. The same comments apply to right-
intersection required for its application. However, traffic turn lanes.
conflicts are well suited to confirming (or denying) sug-
gestions that a specific site has an accident problem or has Driveways at 3-Leg Intersections
inherent safety problems not yet illuminated by an exten-
Many 3-leg intersections have a driveway where a fourth
sive accident history. Typical sources of such suggestions
leg would normally be. Unusual conflict situations may
are complaints, a prominent serious or fatal accident, or a
occur, especially if the intersection is signalized and there
short-term "rash" of accidents at a particular intersection.
is appreciable driveway traffic (which is not signal con-
Also, in this category, would be sites that have temporarily
trolled). Observers should be alert for such movements,
or recently been modified, so that accident histories are not
and record them as notes or under appropriate column
suitable (e.g., a construction site).
headings.
Traffic conflicts are also applicable to before!after evalua-
tions of intersection improvements, both on a site-specific
One.Way Streets
basis as well as to gather "research" data on counter-
measure effectiveness. One must be cautious, however, to If the street of interest is one-way, observation is simpli-
ensure that the changes in conflict counts are causally re- fied because only the approach leg needs to be monitored.
lated in a logical fashion to the type of improvement Also, there will be no opposing left-turn conflicts. On the
implemented. other hand, if the cross street is one-way, one obviously
needs to watch for cross traffic from only one direction—
WHEN TO COLLECT DATA again, a simplification.
It should be kept in mind that traffic conflicts are indica-
Traffic Circles
tive of hazardous situations involving interactions between
vehicles. Generally speaking, therefore, traffic conflicts will Each approach to a traffic circle is similar to an ap-
be most numerous (and most meaningful) when traffic vol- proach to a one-way street. Likewise, traffic within the
umes are heavy. The most effective time for traffic con- circle is somewhat like traffic on a one-way street with
flict counting is during the morning and afternoon peaks, frequent intersections. It differs, however, in that there is
unless traffic becomes totally congested. Other good count- more frequent lane changing. In this respect, it is like a
ing times at some locations are during the noon "peak," series of weaving sections.' Thus, lane-change conflicts will
and during the late midafternoon. be seen frequently.
Accident data or other information at some intersections
may suggest obtaining conflicts counts at other times, such Five-Way Intersections
as during the early evening, or on Saturdays or Sundays. Intersections with more than four approaches are more
But, traffic conflicts data should not be confused or inter- complicated, but no new concepts are required. Cross-
mixed with single-vehicle accident data, which may be traffic conflicts will have to be clearly labeled according to
unrelated in a causal sense. the approach leg used by the cross traffic. If the inter-
section is one with major merging!divergiñg movements
HOW MUCH DATA
(i.e., where traffic on one approach splits fairly evenly be-
The amount of TCT data needed depends on the types of tween other legs and vice versa), three observers will be
 104

TABLE 1-2 Initial Review

GUIDELINES FOR DATA COLLECTION AMOUNTS As the data are returned from the field, the first step
Mean Hourly should be an immediate scanning for completeness or ob-
Hours of
Conflict Category Count Observation 11 vious errors. This should be done while things are still
fresh in the minds of the observers. Many simple mistakes
Left-Turn, Same Direction 7.14 4.6
can readily be corrected at this time, but not after a few
Right-Turn, Same Direction 4.89 5.1 days when details are forgotten.
Some of the items to check are indicated as follows:
Slow Vehicle 3.21 5.9
Accountability for all forms. Are any missing and, if
Opposing Left Turn 0.77 21.6 so, why?
Right-Turn from Right 0.71 23.9 Proper completion of heading information. Are all
blanks filled? Are leg numbers and observation times
Cross Traffic from Right 0.31 39.3
consistent on all forms? Is observer's name on the
Left Turn from Right 0.59 24.5 form?
Are all data entries completed? Are they legible? Do
Left Turn from Left 0.78 18.1
they "make sense"?
Cross Traffic from Left 0.39 30.0 Are there Comments? Are they clear and understand-
able? Are there any observer questions?
All Same Direction 15.48 3.4

All Cross Traffic from Left 0.82 20.0 Data Summations

All Cross Traffic from Right 1.45 14.8 The raw conflict counts, themselves, are not as useful as
Hours of data required to estimate mean hourly count within ± 50%
certain sums and rates. Figure I-i shows a form that can
with 90% confidence. be used in the office to assist in the manual calculation of
these quantities.
As a first step, combine the conflicts and secondary con-
required. Also, the engineer should define for the observ- flicts in each category for every 20-min period. Assume
ers the "straight-thru" path, as opposed to right- and left- that the data in Figure 1-2 were collected in the field.
turn movements, even though a "straight-thru" movement Figure 1-3 shows the summary form of these same data.
may require a slight turn. For example, at 1730 hr the observer recorded four left-
turn, same-direction conflicts; and three left-turn, same-
Offset Intersections direction secondary conflicts. Enter the sum (7) on the
The major difficulty with offset intersections is whether analysis form. Continue this process for each period and
to consider them as two 3-way intersections separated by a category.
short weaving section or as a single 4-way intersection with Next, determine the sums of the counts in the various
a longer than normal clearance interval. In the latter case, categories to create the new categories shown in the last
observation of opposing left-turn conflicts involving ve- four columns of Figure 1-3. The components of each sum
hicles on the offset legs may be difficult for the observers to should be self-evident. These sums are the most likely to
see from their normal vantage points. If so, rather than be suggestive of the presence or lack of operational or
observing from the right side of the approach leg, using the safety problems. The individual counts, from which the
left side may be advantageous. sums are derived, provide more detailed information that
may help in pinpointing specific areas or suggesting cer-
DATA ANALYSIS tain countermeasures. Note that no grand total (e.g., all
conflicts) is calculated. Such a total is not particularly
The conflict data, as collected in the field, are usually not useful as a diagnostic or evaluative measure.
sufficient for decision-making. They must be compiled and Next, create the totals for each approach leg for the time
analyzed to determine if they indicate favorable or unfavor- covered by the data sheet. This time is typically 4 to 8 hr,
able traffic operations, typical or unusual situations, an with 20 min of observation per hour for an observer on
improvement or a worsening. each leg. That is, 4 to 8 entries may be on a sheet, which
Various levels of sophistication can be used in the statis- would correspond to 80 to 160 min actual observation on
tical analyses. One can automate the process using com- that leg. Figure 1-3 also shows these totals.
puters or analyze the data by hand. This decision depends The final step is to divide each total by the total (one-
on the amount of data and the level of analysis. way) traffic volume counted for the leg and recorded on
This discussion assumes that manual procedures of a the Intersection Conflicts sheet in the field (Fig. I-i). This
fairly straightforward nature will be applied. Users can yields the set of conflict rates. It is convenient, during this
easily automate these steps, if desired. Such automation is step to multiply the answers by 1,000 so that more con-
particularly useful if a data base is to be developed for venient numbers result. In Figure 1-3, with 16 left-turn,
future comparisons. This, in fact, is a current need in the same-direction conflicts and 281 total approach vehicles,
TCT field—to more firmly establish norms that can be used the conflict rate is (16/281) (1,000) = 56.9 conflicts per
for comparative purposes. 1,000 vehicles.
 INTERSECTION CONFUCT SUMMARY

Location Leg Number

Day Date Observer

COUNT
STAR

Figure 1-I. Conflict summary form.

 0
0•

INTERSECTION CONFLICTS

Location •,4t'i- Leg Number 3 2

.
Day Date

Conflict - C,
/ 9g Observer

Secondary Conflict - SC
6f
5

COUNT

VOLUME

4 . BEENE EMMEMENEENEWEEMENE

Possible .[3s]

till TINotes andI.1illu.I1fluI1

!igure 1-2. Illustrative field data.

 INTERSECTION CONFLICT SUMMARY

Locaton i'i..c ,1,-'L Leg Number )

Day. Tfl; Date /~E, / 9 7R Observer

L.h T..n RghI 1,n Oppoong Lois lion C,os, 1,oIIio Loll T,n C,o., 110110 R;9h1 hn
Slow V,,l,icl., Lois Iwo I.e., LoIs Rigi,i l.om RighI l,o.,, Righl All C0 £11 C,ow
COUNT Sc,..,, D;,.c,.on
$
50.00. Di,,,clion I.e.,, Loll 1,0.,.
Ofi 0.00. t.
if ic T.Ok All C.ws

START / —
\_ i,.c,iw. lrn 1.1.

TIME
.-f_ I '-i
(MILITARY)

/3 2 C :-J Z /4

/73 0 7 3 C C I 1 1(0

TOTAL /3
/ ;' 4 37
RATE 4f
3 2.S 3. 3.(- 0 /0.7 7.1 1 14 . Z )31.7 3& 32. 0 3SJ
Comments:

Figure 1-3. Illustrative summary dada.

108

Statistical Techniques The variance is calculated using

To apply these findings, certain statistical properties of
NIN
the counts must be known. Once these are known, before/ NY<2_(Y j )2
after or diagnostic evaluations can be made. In either case,
means and variances must be calculated. They can be done Var (Y) = (1-3)
N(N- 1)
simply on inexpensive hand calculators or by using the
formulas given in the following. Using the same data
The mean value is given by the formula 4(0 + 9 + 36 + 49)- (16)2
Var (Y) = (I-4a)
4(3)
(I-I)
= 376 - 256
Var (Y) (I-4b)
12
For instance, using the left-turn, same-direction counts in
Figure 1-3: or
Y= 1/4 (0+3+6+7) (I-2a) Var (Y) = 10 (I-4c)
=1/4 (16) (I-2b) The standard deviation, o-, is the square root of this
amount, or 3.16.
or
For before!after tests, one can usually assume negligible
y=4 (I-2c) changes in traffic volumes, so the counts themselves can be

1O.LI Vulsoddo >0

'I,
z
o Uell300pfld 2< 2< 2< 4<

0< 4< 5<

Cl) 14LI W01
j

4.0 401
4.140
3Ja1 00d 4< >0 >0

Z 01 40
0 .I qni woiA >0

404.0401
0 01 01
0i. U 2304I W0 0< 0< >0
'4.
40 .I )q:T( W0.I.J >0 2< >0 2< 0<X 2<

04
.01
004.1
01 40
0 401
'1-I
01 4- 01
00 E OH W01 >0 >0
1J01 4< 2<

11
Q
N
5-
U144 J.-3Ja1 u1soddo 0< >0 2< 4<

pU1oa4 jL')Oj >0 >0 2< >0 4-0 4<

01
z
o 0 '
0.04441
01 1.4 II
0U404D OUO'1

0IqA o01
0< 2< 4< 2< 2< 4<

>0 2<
2<

4< >0
0<

4<
uanj'-14)rlm
Cl) 001
4. 01 >4 >4
0 Uiflj-) J 4.Oi 0< 2< >0 >0 2< >2 4< 4< 0<
o
-401
'0 000
Cl) 0
O 0 00 1.44/4
0 4. 40 0 1.4 01
'-4 0 40 - 0 01 .0 0
Z U 0
.0
0.
,-4
00
4. 0 00 0.001
40 'H 'H 0 0
'.4 01 10 0-40 4.4 444 OZ
04.0 U 000)
t 0
4104.,'0
'00000
0 00
01
"44.0
1.10.0
'.-400
5.100
001.4
'44(20
01 0.00 '4 'HO OH 404.1.4
- > 0 004.
c. 0 4.4.01
4) 414,0 01100 0010 000<4.40
EO1.4 00000Z0UO000l'O
0100.0'0<00..'I,4O 00 01
' , 010-4-)
Z 01
i-I
4
10110400000
1-0 4.404141 000.-400N40000
0001.400 O.-5>U
00_4-'.I00l0•0000041
U.0U4
'40,0,0000 SD0.-40000 .4/04'.'H00"400
1-I •1.4 00 00 00 00 _4 ..i C u- o t '0 '0 5.. 40 01. 00 0 > 0/4 0 44.4
('I 00- 00 .0 (0 CE 0.'O '00
- ..4400>0 444 <.H 000.0.<0. <144
 109

analyzed. Two-sample f-tests are appropriate for ascertain- determine significance. Note that the proper rate variance
ing the significance of any change in mean counts. The must account for variabilities in both the conflict counts
value of t is calculated from and in the volume counts. Refer to standard textbooks for
this procedure, as well as to more sophisticated analysis
techniques.
(1-5)
INTERPRETATION
V N B -i NA -i
where the subscripts B and A refer to before and after data. The traffic engineer should treat traffic conflicts data as
The calculated value is then compared with those tabulated supportive of other information rather than to be used ex-
in any standard statistics text, using the number of degrees clusively. The TCT allows field observation on a formal-
of freedom, v, given by ized basis, which can be accomplished by a traffic tech-
nician. The engineer need not spend all of his time in the
v = ( N1 + NA - 2) (1-6) field, but instead can apply his efforts to other activities.
One such activity is making judgments based on a variety
The user can determine his own variances, 0 2, or, if data of data and information.
are insufficient for this, take them from results given in the The traffic conflicts data should be examined and inter-
research report. preted in light of such other information as accident data,
For making diagnostic evaluations, it is necessary to es- traffic volumes, intersection inventory, intersection draw-
tablish the range of normal (or average) conflict rates. ings, intersection photographs, signal timings, and observer
Some agencies have collected sufficient traffic conflict data comments.
at a large enough number of sites that they can provide If deficiencies are detected, the same improvement meth-
approximations. The NCHRP Project 17-3 also obtained odology as is used with high-accident locations can be ap-
some preliminary estimates for conflict rates. These kinds plied. Some countermeasure suggestions are given in Tables
of comparison values can be used in one-sample f-tests to 1-3 through 1-5.

UO111010p>dI 1< 1< 0<01

II0 UO.1j >1 4< >1 >11

44-
.0 11 It
>0 4-1 ue.]lsopo, 101 2< 011<
JO1I 00 'd >1 01

-I fl1TII O0.1. 1< 4< >0

0 44-
1.. O 1:
O 044(0 1< 4< 0< 0< 04 >0 )< 0
.2I-.L>

>1 0< >04< 7<

is 010.141
0011 0'4- 0< 010< >11< IF
-.4 01 4.4 II 1<
4..) J0-I 010.141 4<
1<

U.1fl1-Ijol 13Uisodd0
C
0
pu-.100l1 I1°J. >0 4< 0<01 4<
C 1 pU41-JII.)8 10(14- 4< 4< >01< 4<
Ci
0 ..-' 1124 08ue43 O(101 :i •I
0 4-'
It 0 4-' 0 t 01T 1l0A
MOIS 4< 010< 1<
I .01
11) 0 cml
uin 0< 4<1< 1<4< >00< 0<
I OIZlTqaA (.7<41 1< >04< ><
10
u_In_I. 101 4<4<4< >11< 010< >0
u.lnJ._(j011 >07< >0
C 0
C 00 .<
00
Ut 10

0 0 0.0 0 0 00
11
r
000
0 0
11.. 0
4 000
0
0 0 .041 .440
0. .4 1-410 OOZ C
0 OZ 0004 0 0
.1.1 1.1 0 00014 0
0.0 14 .411 CCI-'" OCt It 4.0 -.4)0010 I-I 11
01)t It It 4.404 .1(0001 0)00 00 4-C C000 0
00)10 0) ICI- 01.400 011 1100 110½-I
0) 011 I.. 00111.4 01CC C 0 0000.11 1.104- V
..I CC C 000010(1 1-IC 01.411 >It 01000 C0UO'01-I 0
10 C 4.1 4.1 >It 1,1000 C 000)01 CO 010.-I 0 CO
C11 0101(0<0) - CO 0 0 001-11-1)1.. 0000Itt-1 01.44-1000 0
00014-1-.)).. COOl) ION 04-41040 I tIlt COO 0.014-40010 •<
.1-11:1 I 0ll(tC>II 0.041-IOU 114.44-00.2< OOt COCCI Cl- >
4-It l,_7-0 0 Ot C 00(0 C 4 .0.0100 .1411<010.011)0010
00 11.<01t.Hl4ItO V 00 00 10 V V 1.4 0. 0 0 0 lIt 1- 0 44.4 0
t II 00 00 10 0 0 11 0. 0 0 0 10 4-' 0 1-1 40 0 01.0 V C 1-
0 0 01 .1< 0 0 0 0 Ct 0 C <1010 (0< <14410 ml
01< <00
THE TRANSPORTATION RESEARCH BOARD is an agency of the National
Research Council, which serves the National Academy of Sciences and the National
Academy of Engineering. The Board's purpose is to stimulate research concerning the
nature and performance of transportation systems, to disseminate information that the
research produces, and to' encourage the application of appropriate research findings.
The Board's program is carried out by more than 150 committees and task forces
composed of more than 1,800 administrators, engineers, social scientists, and educators
who serve without compensation. The program is supported by state transportation and
highway departments, the U.S. Department of Transportation, and other organizations
interested in the development of transportation.
The Transportation Research Board operates within the Commission on Sociotech-
nical Systems of the National Research Council. The Council was organized in 1916
at the request of President Woodrow Wilson as an agency of the National Academy of
Sciences to enable the broad community of scientists and engineers to associate their
efforts with those of the Academy membership. Members of the Council are appointed
by the president of the Academy andare drawn from academic,. industrial, and govern-
mental organizations throughout the United States.
The National Academy of Sciences, was established by a congressional act of incorpo-
ration signed by President Abraham Lincoln on March 3, 1863, to further science and
its use for the general welfare by bringing together the most qualified individuals to deal
with scientific and technological problems of broad significance. It is a private, honorary
organization of more than 1,000 scientists elected on the basis of outstanding contribu-
tions to knowledge and is supported by private and public funds. Under the terms of its
congressional charter, the Academy is called upon to act as an official—yet indepen-
dent—advisor to the federal government in any matter of science and technology,
although it is not a government agency and its activities are not limited to those on
behalf of the government.
To share in the tasks of furthering science and engineering and, of advising the federal
government, the National Academy of Engineering was established on December 5,
1964, under the authority of the act of incorporation of the National Academy of
Sciences. Its advisory activities are closely coordinated with those of the National
Academy of Sciences, but it is independent and autonomous in its organization and
election of members.
TRANSPORTATION RESEARCH BOARD . NON-PROFIT ORG.
National Research Council U.S. POSTAGE
2101 Constitution Avenue, N.W. PA I D
Washington, D.C. 20418 WASHINGTON, D.C.
PERMIT NO. 42970
ADDRESS CORRECTiON REQUESTED

4.