Total Transportation System Solutions (T2S2)

Presented at the 77^th Annual Meeting of the Transportation Research Board

Washington, DC

January 1998 

 

 

ABSTRACT

 

The purpose of this study was to test the hypothesis that turn signal usage reduces stopped delay; and, specifically, the hypothesis that turn signal usage by a right turning vehicles on the through roadway of a three-legged intersection reduces stopped delay to a waiting vehicle on the side street.   A total of seven intersections, four in theSaint Louis,Missouri, area, and three in theLittle Rock,Arkansasarea, were studied.

The data were evaluated using three different statistical analyses.  A scatterplot of the average stopped delay per vehicle versus the percentage of turn signal usage by right-turning vehicles was inconclusive.  Both the binomial test and the Wilcoxin Rank Sum Test indicate that the null hypothesis, turn signal usage does not affect the stopped delay, should be rejected.

 

 

INTRODUCTION

 

Why Study Driver Behavior?

 

As the amount and severity of traffic congestion increases, highway officials are devoting more attention to solving traffic congestion problems.  The most common suggestion for improving traffic flow is to modify the roadway environment, be it widening, signalization, turn restrictions, limiting access, or installation of medians.  However, changes to the roadway environment may not necessarily be effective in reducing traffic congestion.

The physical characteristics of the roadway facility are not the only cause of traffic congestion.  The manner in which a vehicle is operated on the roadway also causes congestion.  One common example of such a driver behavior is the queuing of vehicles through intersections.  For example, Driver Smith is at an unsignalized intersection in a queue of stopped vehicles waiting for a nearby traffic signal to change.  Driver Smith, instead of waiting behind the intersection, waits in the intersection.  Because of this action, drivers on the other approaches to the intersection cannot proceed through the intersection due to the blockage, resulting in gridlock.

This is not the only driver behavior that can cause congestion.  Other examples of driver behavior which can and do create traffic problems include weaving between lanes (in the same direction), misuse of a two-way left-turn lane, and making turns where they are restricted (e.g. a left-turn where left-turns have been prohibited).

Thus, changing the physical roadway is not the only possible solution to traffic congestion problems.  Changing driver behavior may be an inexpensive and viable method to reduce congestion.  By changing driver behavior, the roadway not only may operate much more efficiently, but also may have an increase in capacity.  In short, capacity and efficiency may be increased without a major capital investment in infrastructure.

 

Purpose Of This Study

This study was both a traffic flow study and driver behavior study.  Its primary premise was to determine if a relationship exists between a particular driver behavior and a particular traffic flow parameter.  Specifically, the purpose was to test the hypothesis that not using turn signals increases delay, and, in particular, the hypothesis that a right turning vehicle on the through street of a three-legged intersection that does not use turn signals increases the stopped delay to a waiting vehicle on the side street.

 

LITERATURE REVIEW

After an extensive literature search, which included a computer search of the Transportation Research Information Service (TRIS) database, no mention of any previous study of the effect of turn signal usage on delay could be found.  Therefore, due to the lack of literature found on this subject, it was concluded that no studies on this topic had been published.  However, a limited amount of literature on topics closely related to the study topic exists.  These related topics are in the areas of human factors in engineering and design, driver behavior patterns, and laws and regulations of driving.

 

Human Factors

Extensive research has been conducted in the area of human factors.  Much of this research pertains to physical elements of the transportation network, most notably the physical elements of the automobile (e.g. seat design, dashboard design, etc.) and the roadway environment.  Since this study was not concerned with physical elements, most of this research was not applicable to this study.

However, a study by Alexander and Lunenfeld does mention, albeit briefly, a relationship between the driver and efficiency.

“As a principal controlling element, drivers are primary determining factors in the system's successful operation.  Skillful driving task performance, maintenance of vehicle control, safe and efficient guidance through roads and traffic, and proper navigation using an optimum mix of routes, represent ways in which driver performance enhances operations and safety.”(1)

In short, the manner in which a driver operates a vehicle will affect the efficiency of the traffic network, that is, how the network operates.

 

Driver Behavior

As with the area of human factors, much research has been done in the area of driver behavior.  Some of this research pertains to the topic of turn signal usage.

 

                        Lansing Study

One of the earliest studies of turn signal usage was a 1958 study in Lansing, Michigan, by Barch, Nangle, and Trumbo, of the turn-signalling behavior of 10,467 drivers at seven intersections.  (2)  The conclusions drawn from this study were:  women use turn signals more than men; left turns are signalled more frequently than right turns; the type of intersection affects turn signal usage; signal usage is independent of the time of day, the presence of preceding or following traffic, and the use of signals by preceding vehicles.  No mention was made in this study of the effect of turn signal usage on delay.

 

                        Hawaii Study

The focus of Papacostas’ 1984 Hawaiistudy was to determine the effect of not using turn signals on the lane preference of following drivers proceeding through an intersection. (3)  Specifically, this study determined the percentage of turn signal usage and the effect of non-use at signalized intersections with a lane drop at the far side of the intersection.  Researchers found that a sizable percentage of left turning drivers failed to properly indicate their intentions to turn left, which had a significant effect on the lane choice of following through vehicles.  Like theLansing study, this study also did not mention delays.

 

                        Other Studies

Several other studies were conducted which relate to turn signal usage.  One of these concerned the effect of education on driver behavior, particularly turn signal usage.  This study, by G.W. Blomgren, T.W. Scheuneman, and J.L. Wilkens, reported an increase in turn signal usage due to an educational message on a strategically placed sign.  (4)  The Blomgren, Scheuneman, and Wilkens study also showed that women use signals more than men, and that left turners use signals more than right turners, a verification of the results of theLansing study.

 

STUDY DESIGN

Scope Of This Study

As previously stated, the purpose of this study was to test the hypothesis that the non-usage of turn signals increases delay, and, in particular, the hypothesis that a right turning vehicle on the through roadway of a three-legged intersection that does not use turn signals increases the stopped delay to a waiting vehicle on the side street.

As signals could be used by vehicles on the through roadway at any point prior to the intersection with varying degrees of influence on the decisions of the motorists on the side street, it was deemed to be essential to select one distance at which it could be concluded that signals were either on or off.  After much consideration, a distance of 30.5 m (100’) was chosen.

There were several reasons why a distance of 30.5 m was selected.  First, this is the distance cited in the Uniform Vehicle Code.  Section 11-604-(b) of the Uniform Vehicle Code states “a signal of intention to turn . . . when required shall be given continuously during not less than the last 100 feet traveled by the vehicle before turning.”  (5)  Second, this value, 30.5 m, is the value derived from the graph on page 65 of Greenshield’s, Schapiro’s, and Ericksen’s book, Traffic Performance At Urban Street Intersections, for vehicles traveling approximately 45 km/hr (28 mph), which is representative of the approach speeds of many urban intersections.  (6)  Third, the distance required to stop specified in the Manual On Uniform Traffic Control Devices (Condition B) for a 48 km/m (30 mph) approach speed is given as 30.5 m. (7)   Finally, 30.5 is approximately the minimum brake reaction distance on wet pavement for a roadway with a design speed of 48 km/h. (8)

For the aforementioned reasons, it was decided that 30.5 m (100’) would be a good value.  This value, 30.5 m, is the value at which a turning vehicle would begin braking but would not telescope its intentions to the stopped vehicle on the side street.  If the value was less than 30.5 m, a turning vehicle would be noticeably breaking; thus, the stopped vehicle would already know to proceed into the intersection.  Under these circumstances, the point of the study, to analyze the delay to the stopped vehicle caused by not knowing the intentions of the through motorist, would be moot.  The study objective could only be achieved by setting the distance at a point at which the through vehicle’s intentions were not known.

The 30.5 m distance, then, represents the probable closest distance to the intersection at which the motorist on the side street cannot determine the intentions of the motorist on the through street.  A distance in excess of 30.5 m was not chosen because it was felt that the actions of a vehicle more than 30.5 m from an intersection would probably not have as much of an influence, if any at all, on the decisions made by the operator of the stopped vehicle.  Although several of the intersections had approach speeds in excess of 48 km/hr, it was felt that since vehicles on the through road would not have to come to a complete stop to make a right turn, 30.5 m was still a reasonable value to use for the point at which braking would commence.

 

Requirements For Inclusion In The Study

Three-Legged Intersection  The first requirement for inclusion in the study was that all intersections be unsignalized, three-legged intersections.  This restriction was established to simplify data gathering, for it is much easier to gather data on turning movements at a three-legged intersection than at a four-legged intersection (due to the fact that there are twelve possible movements at a four-legged intersection, and only six at a three-legged intersection).  Thus, due to the simplified geometrics of the intersections being studied, data were required for only three movements, which permitted data gathering by a single individual.

 

Turning Movements  The second requirement concerned the operational characteristics of the intersection.  Because the overall objective of this research was to determine delay caused by right turning vehicles, the intersections selected had a high percentage of right-turning vehicles from the through roadway onto the side street.  However, it should be noted that although a high percentage of right turning traffic from the through to the side street was highly desirable, forced right turns, such as when the through street becomes a one-way at the intersection with the one-way flowing into the intersection, was not.  Such a forced diversion does not allow a choice by vehicles on the through street; all vehicles must turn right.  Thus, turn signal usage by drivers on the through street is irrelevant, as motorists on the side street know that the motorists to their left will always be turning right.  (Figure 1A)

 

   

 

Traffic Volumes (Road Classification) All intersections studied met a minimum road classification requirement.  This third requirement was that the side street approach and the “left through” approach (left from the perspective of a waiting vehicle on the side street approach) of the intersection must be classified as a collector or higher (Figure 1B).  This requirement was established to prevent the inclusion of low volume intersections, which would yield statistically suspect data due to the small sample size.

While the “left through” and side street approaches should not be classified lower than a collector, the “right through” approach (from the perspective of vehicles on the side street) should not be classified higher than a collector.  This restriction was incorporated to minimize, if not eliminate, delays to waiting vehicles on the side street approach caused by vehicles approaching from the “right through” approach, as these delays were not within the scope of the study and were expunged from the data sets.

Presence of Driveways  For all intersections included in the study, there were no driveways along the “left through” approach within 30.5 m of the intersection.  The reason for this exclusion is that driveways along the “left through” approach within close proximity of the intersection may affect delay to vehicles on the side street because a waiting vehicle on the side street may not know the intentions of the approaching vehicle, and, if that approaching vehicle does not enter the intersection but instead turns into a driveway, that approaching vehicle has nonetheless delayed the waiting vehicle.  In other words, although the approaching vehicle did not enter the intersection, that vehicle still delayed the waiting vehicle on the side street approach because of the uncertainty by the driver of the waiting vehicle of the intentions of the driver of the approaching vehicle.

This restriction does not include low volume driveways and additional street approaches at the intersection itself, for in these cases, approaching vehicles still must enter the intersection.  In short, the intent of this restriction was to attempt to eliminate delays to vehicles on the intercepted approach by vehicles not entering the intersection.

Weather Conditions  Because weather can greatly affect driver and vehicular performance (for example, a vehicle on wet pavement requires a longer stopping distance for a given speed than that required for the same vehicle on dry pavement), all data were collected under similar weather conditions.  For all data in this study, the weather conditions at the time of data collection were dry and sunny.

Pavement Conditions  Just as weather can affect driver and vehicular performance, so too can pavement type and condition.  Drivers and vehicles tend to respond differently on high quality pavements (Portland cement concrete, asphaltic concrete) than on lower quality pavements (gravel, bituminous seal-coat).  Thus, to effectively eliminate bias due to pavement type, all approaches to the intersections included in this study were either Portland cement concrete or a high quality asphaltic concrete.  Furthermore, all approaches to the intersections included in this study were in good condition.  More specifically, the approaches were free of potholes and ruts, as potholes and ruts also affect driver and vehicular response, and thus can also significantly affect any data collected.

Other Considerations  Many other factors may affect driver and vehicular performance, and ultimately all data collected at that intersection.  Examples of some of these factors are the approach speed limit and the percentage of trucks.  However, these factors were not considered for two reasons.  First, further stratification would have created a small sample, which could greatly bias the data.  Second, these other factors were considered to have a negligible impact on delay when compared to the impact of the five factors previously mentioned.

Intersections Studied

A total of seven intersections that met all of the requirements for inclusion were selected to be studied.  Although more than seven intersections met the requirements, only a small number could be studied due to time and financial restraints.  However, the seven intersections chosen represented a wide spectrum of intersections, as they differed in geometrics (e.g., Y-intersections v. T-intersections), traffic volumes and characteristics, and location (e.g., suburban v. urban).  Thus, while all seven of the intersections studied were similar, they were by no means identical.  In fact, the similarity between the intersections studied ended with the fact that they all fulfilled the requirements necessary to be included in the study.

Of the seven intersections studied, four were located in the Saint Louis, Missouri, area, and three are located in the Knoxville, Tennessee, area.  This splitting of the study between two different states was done to reduce bias caused by a homogeneous sample, in this case drivers from a particular locality being a homogeneous group.  The seven intersections studied and the characteristics of each intersection are listed in Table 1.  It should be noted that at intersections M-1 and T-7, the numbered routes, Missouri Route 231 and Tennessee Route 131, respectively, turned onto the side street from the through street.  These were also the two intersections with arterial approaches.

Table 1.  Characteristics of the intersections studied.

General Considerations for Data Collection

Bias Reduction

Because variables other than the six previously enumerated can affect traffic flow, a sample may be biased due to these other variables.  However, the manner in which a data group is collected can greatly reduce, if not eliminate, the bias caused by those other variables.  The most important of the additional variables which may cause biasing are the day of the week and the time of day.  For example, if a data group is taken on Monday, the sample may or may not be representative of the actual conditions at that intersection, since traffic patterns on any one day may be different than patterns on another day of the week.  The same holds true for the time of day.  Traffic conditions at two in the afternoon are probably not the same as conditions at eight in the morning.  Thus, data were taken on different days and at various times among the sites to attempt to eliminate the effect of these biases.

Although theoretically data could be collected at any time on any day of the week, there are some practical limitations to when data should be collected.  Thus, every effort was made to collect the data when the intersection was operating under conditions which were the best representation of its typical operating characteristics.

Study Conditions

Day of the Week  All data were collected on weekdays for two primary reasons.  First, traffic volumes at the intersection studies were higher on weekdays than on weekends; thus, data sets collected on weekdays were larger and more representative of the typical operating characteristics of the selected intersections.  Second, the percentage of non-local traffic, traffic not familiar with the intersection, was lower during the weekdays at the study locations.  Thus, by evaluating the intersections on weekdays, truer representations of both the typical traffic flow and typical driver behavior characteristics were obtained.

Time of Day  All data were collected in the afternoons and evenings.  There were several reasons for this. First, commuter traffic volumes tend to be more concentrated in the afternoon than in the morning.  One explanation for this higher concentration of commuter traffic is that morning commuter traffic tends to be distributed over a wider time frame than afternoon commuter traffic.  In the mornings, commuters do not all arrive (at work) at the same time; on the other hand, afternoon commuters tend to leave at the same time (quitting time), causing a higher concentration of commuter traffic.

Second, afternoon peaks generally involve higher overall traffic volumes than their morning counterparts.  These higher volumes are not only the result of more commuter trips (as a result of the higher concentration), but also more non-commuter trips as well.  In the mornings, most vehicular trips are home-based work or home-based school, as many retail establishments do not have business hours during the morning peak period.  In the afternoons, the number of home based non-work trips is higher, due to the fact that most retail establishments are open during the afternoon peak period.  In addition, the presence of recreational drivers in the afternoon traffic stream is another reason for higher afternoon traffic volumes, as these types of drivers generally travel more in the afternoon than in the morning.

Because the number of non-commuter trips is higher in the afternoon than in the morning, a sample taken in the afternoon is less homogeneous than a sample taken in the morning, and thus less biased.  For example, if a sample is taken when the overwhelming majority of drivers are commuters, such as during a morning peak period, it cannot be conclusively determined whether the results are applicable in general or just to commuters.  On the other hand, the results of a sample taken when no one type of trip is the dominant majority probably cannot be attributed to one particular group, and instead can be considered to be applicable to the intersection in general.

Although all data were collected in the afternoon and evening hours, care was taken to avoid collecting two data sets from any particular intersection at the same time interval during the day, as that might tend to bias the data.  Four distinctive time intervals were established to help prevent biasing of the data: afternoon, pre-peak, peak hour, and pre-twilight.  These intervals were defined as follows:

Afternoon, from12 noon until2 p.m., local time;

Pre-peak, from2 p.m. until4 p.m., local time;

Peak hour, from4 p.m. until6 p.m., local time;

Pre-twilight, from6 p.m. until dusk, local time.

Local time was used because of the time zone difference between theTennesseeandMissouristudy locations.  Despite the time zone difference, peak hour occurred at or near5 p.m.at all sites; thus, the intervals were valid for all sites if local time was used.

The starting time for a particular data set, that is, the time at which data collection began for that data set, was the parameter used to determine the time interval.  For example, if a technician were to collect data from5:30 p.m.to6:30 p.m., that data set would be classified as peak hour, because that would be the time interval when data collection began.

Parameters and Guidelines Used In Data Collection

In order to maintain consistent data collection techniques, several guidelines were established to define the actions being studied.  These guidelines established when the vehicle on the through approach was considered to have its signal on, and the beginning and ending of delay for the vehicle waiting on the side street.

Turn Signal Activated

As stated earlier, an approaching vehicle on mainline was considered to have its turn signal on if the signal was on 30.5 m from the intersection.  This 30.5 m was measured along the edge of pavement or curb along the mainline approach on which signalling vehicles were travelling, on the same side of the roadway as the intercepted approach.  In order to clearly identify this point for the observer, it was then marked with either a lathe adjacent to the roadway or a paint mark on the roadway itself.

Determination of Delay

A vehicle was considered to be delayed when:  (1) the vehicle on the intercepted roadway was stopped and ready to proceed, and (2) a vehicle on the through roadway approaching the ready vehicle (on the side street) from the left (from the perspective of the ready vehicle) was within 30.5 m of the intersection.    It should be noted a vehicle was not considered to be delayed if it was not “ready” at the time that the approaching vehicle crossed the marked 30.5 m threshold.  All delays were stopped delays to individual vehicles and were not total intersection delays.  In addition, any delay to a ready vehicle caused by vehicles on the through roadway approaching from the right was not considered, regardless of whether or not the delay was initially caused by a vehicle approaching from the left.  Finally, one and only one vehicle on the intercepted approach was considered to be delayed at any given time.  Thus, if a queue existed on the side street, only the vehicle at the stop bar was considered to be delayed.

Delay ended when the driver showed an intention to proceed into the intersection.  For automobiles, motorcycles, and other small vehicles, this intention was considered to be shown when approximately one-half the vehicle crossed the stop bar.  For trucks, this intention was considered to be shown when the tractor crossed the stop bar.  Intent was used as the parameter because it minimized the amount of delay caused by the acceleration of the vehicle from a stop.

The vehicle considered to be causing the delay was the first vehicle to cause the delay.  In other words, if delay began because of the first vehicle in an approaching queue, and other vehicles in the approaching queue further contributed to that delay, then the entire delay was said to have been caused by the first vehicle.  In all cases, delay was considered to be caused by only one approaching vehicle.  For each vehicle delayed, there was one and only one vehicle responsible for that delay.  Thus, in the case of a queue of approaching vehicles, the lead vehicle was considered responsible for the delay to a "ready" vehicle on the intercepted roadway.

Methods

Data were collected only for delay situations, that is, situations where a vehicle on the intercepted roadway was "ready" and an approaching vehicle on the through roadway was within 30.5 m of the intersection.  Two different methods were used in the gathering of the data:  the “stop-watch method”, and the “event-recorder method.”

With the “stop-watch method”, a stop watch was used to measure the delay.  This method provided flexibility in the selection of intersections, as a vehicle was not needed at the intersection in order to collect data.  However, this method also required a very alert and attentive research team, as the chance of human errors such as forgetting to reset the stop watch or missing a delayed vehicle entirely, was high.  All data for Intersections M-2 and M-3, and some data for Intersections M-4 and T-5 were collected using this method.

With the “event-recorder method”, all data were collected with an Esterline-Angus Event Recorder.  Information on vehicles causing the delay was encoded in such a manner that the researcher was able to extract that information at a later time, rather than extract the information at the intersection as is required by the “stop watch method”.  Delay data were also able to be determined at a later date using this method.  The use of the “event-recorder method”, in comparison to the “stop-watch method”, reduced the responsibilities of the researcher at the site, and allowed for verification and rechecking of data that appeared to be erroneous.

However, the use of the event recorder also had several disadvantages.  First, as the instrument must be connected to a vehicle's battery, the number of intersections that could be studied was restricted to those that provided a safe and legal parking area for a vehicle.  Second, the instrument was cumbersome, and required at least 10 minutes to set up, as well as 10 minutes to put away, and had to be cleaned after every use.  Third, the visible presence of the researcher at the intersection altered driver behavior in several instances, thus resulting in the expunging of some data.  All data for Intersections M-1, T-6, and T-7, as well as some data for Intersections M-4 and T-5, were collected using this method.

Data Collected

The vehicular data collected consisted of the stopped delay to "ready" vehicles and information on whether or not the approaching vehicle had its turn signal on 30.5 m from the intersection.  All delay data were recorded in seconds.  No differentiation was made for the type of vehicle (i.e. auto, tractor-trailer, motorcycle, etc.) either causing the delay or being delayed.

DATA ANALYSIS

A total of fourteen data sets were gathered at the seven intersections.  The first nine of these data sets were gathered at the intersections in Missouri, and the remaining five were gathered at the intersections in Tennessee.  These data sets are shown in Table 2.

Percent Signal Usage vs. Delay

The first step in the analysis of the data was determining whether or not a correlation existed between the percentage of turn signal usage by all right-turning vehicles causing delay and the average delay to all "ready" vehicles delayed.  For each data set, the percentage of all vehicles causing delay that used turn signals was determined, as well as the average delay to all vehicles delayed, whether delayed by vehicles using turns signals or vehicles not using turn signals.  The results for an individual data set were then analyzed with the individual results for all other data sets.

The percentage of right-turning vehicles on the through roadway that used signals ranged from 25 percent to 75 percent.  A scatterplot was made to show if any correlation existed between the percentage of turn signal use by right-turning vehicles on the through roadway and the average stopped delay to "ready" vehicles on the side street.  This scatterplot was widely scattered, and suggests only a weak correlation, if any, between the percentage of turn signal usage and delay.

Table 2.  Delay data.

Several regression equations were plotted for the scatterplot, and a statistical correlation factor (R value) was computed for each regression curve.  When an exponential regression equation was used, the resulting R value was 0.12, which means little correlation.   Linear and logarithmic regression equations yielded an even lower correlation factor, 0.06.

Signal Usage and Non-Usage vs. Delay

The data sets were stratified to take into account whether or not the approaching vehicle on the through roadway was signalling a right turn.  In other words, delays caused by vehicles not using turn signals were separated from delays caused by vehicles using turn signals.  It is important to note that although the data sets were encoded to differentiate between the delays to left-turning and right-turning “ready” vehicles, this differentiation was not considered for this study.

Several different statistical tests were employed to determine if the difference in delays associated with turn signal usage were significantly lower than those delays associated with signal non-usage.

Binomial Test

The binomial test, one of the simplest statistical tests, determines the probability of a particular combination of outcomes for a variable for which only two outcomes are possible.  In the case of this study, the outcomes were: signal usage reduced delay, and signal usage did not reduce delay.  If a relationship between signal usage and delay does not exist, then the probability for either outcome is 0.50; that is, for a particular data set, there is a 50-50 chance that delay was reduced when signals were used.

Using this hypothesis, a binomial test was performed.  Results of three of the fourteen data sets indicate no decrease in delay when signals are used.  The binomial probability of this occurrence is 0.029.  Thus, the null hypothesis, that turn signal usage does not reduce stopped delay, can be rejected at the 95% level of confidence.

Wilcoxon Rank Sum Test

One major disadvantage of the binomial test is that it does not take into account magnitudes of the differences in the stopped delays.  Thus, a second statistical test was performed to determine if not only the frequency of delay reductions was significant, as shown by the results of the binomial test, but also if the magnitudes of these reductions were significant.  Thus, a Wilcoxon Rank Sum Test was performed to determine if the differences in magnitudes between the delays caused by vehicles using signals and the delays caused by vehicles not using signals were significant.

The Wilcoxon Rank Sum Test was chosen as the device for this evaluation instead of the “T” test because the Wilcoxon Test is valid for non-parametric samples such as these.  The results of the Wilcoxon Test showed that the probability of the distributions of stopped delays caused by vehicles using turn signals and not using turn signals being identical, which is the probability that turn signal usage by a right-turning vehicle on the through roadway did not have an effect on the stopped delay of a “ready” vehicle on the side street, is 0.005.  Thus, the null hypothesis, that turn signal usage does not reduce stopped delay, can be rejected at the 95% level of confidence.

CONCLUSIONS

The results of this study support the hypothesis, “turn signal usage reduces delay.”  A scatterplot of “the percentage of turn signal usage versus delay” shows, albeit weakly, that as turn signal usage increases, delay decreases.  Furthermore, two statistical analyses, the binomial test and the Wilcoxon Rank Sum Test, clearly indicate that the null hypothesis, turn signal usage does not reduce delay, can be rejected at the 95% level of confidence.  In short, the statistical analysis supports the hypothesis of this research, which is that turn signal usage by right turning vehicles on the through roadway at an unsignallized three-legged intersection reduces delay to waiting vehicles on the intercepted approach.

However, these results should be used with caution.  This study was limited in scope, and the results should only be extrapolated to similar intersections.  In addition, because of the small sample sizes, in terms of the number of intersections studied, the number of times each intersection was studied, and the number of vehicles within each individual study, definitive conclusions should not be drawn without a much larger and more detailed study.  However, this study can provide the rationale and justification for embarking on a more in-depth study.

Three factors may have contributed to variance of the data sets:  (1) the physical limitations of the researcher pertaining to reaction time, (2) altered driver behavior due to the driver's detection of the researcher at the intersection, and (3) distractions to the researcher from inquiring minds.

Because of the limited scope of this study, it was not intended to be the definitive study in this area.  One suggestion for future study is a general analysis in which a large heterogeneous sample is used.  The results could then be compared to the results of this study, which used a small heterogeneous sample, for verification of the results.  In addition, a larger sample may provide a stronger correlation between the percentage of signal usage and delay when the regression analysis of those two variables is performed, and a stronger correlation equation could be useful as a nomograph for traffic analysis.

Another suggestion for future study is a comparison of the delay reductions due to turn signal usage for different intersection configurations.  This type of information could be useful in the design of intersections, especially if a particular intersection configuration has consistently lower delays than other configurations.  Many other possibilities for future research exist that have not been enumerated here.  This vast number of possible research hypotheses suggest that the need exists for more study in this area, as this study has covered only a fraction of the vast potential of this field.

REFERENCES

1.         Greson J. Alexander and  Harold Lunenfeld.  "The Role of Driver Expectancy in Highway Design and Traffic Control."  Civil Engineering Practice, Volume 4, 1988, p.429.

2.         A. M. Barch, J. Nangle, and D. Trumbo.  "Situational Characteristics and Turn Signalling."  Highway Research Board Bulletin, 1958, Number 172, pp. 95-103.

3.         C. S. Papacostas.  "Influence of Leading Vehicle Turn Signal Use on Following Vehicle Lane Choice at Signalized Intersections."  Transportation Research Record, Number 996, 1984, pp. 37-44.

4.         G. W. Blomgren, Jr., T. W. Scheuneman, and J. L. Wilkens.  "Effect of Exposure To a Safety Poster on the Frequency of Turn Signalling".  Traffic Safety and Research Review, Volume 7, Number 1, March 1963, pp. 15-22.

5.         National Committee on Uniform Traffic Laws and Ordinances.  Uniform Vehicle Code and Model Traffic Ordinance, Revised 1987.

6.         Greenshields, Bruce D.; Shapiro, Donald; and Ericksen, Elroy L  Traffic Performance At urban Street Intersections, 1947.

7.         U.S. Department of Transportation, Federal Highway Administration.  Manual On Uniform Traffic Control Devices, 1988 Edition.

8.         American Association of State Highway and Transportation Officials.  A Policy on Geometric Design of Highways and Streets, 1984.

This was originally published as a Master's thesis.  The official citation of this version is-

Schrader, M.H. (1998) The effect of turn signal usage on delay at three legged intersections.  In the proceedings of the 77th Annual Meeting of the Transportation Research Board, Washington, DC.

This is the condensed version as presented at the 77th Annual Meeting of the Transportation Research Board.  The unabridged version is available as a .pdf on the "Schrader, Michael H." home page.