U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590

Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

This report is an archived publication and may contain dated technical, contact, and link information
Back to Publication List        
Publication Number:  FHWA-HRT-15-043    Date:  June 2015
Publication Number: FHWA-HRT-15-043
Date: June 2015


Investigating Improvements to Pedestrian Crossings With An Emphasis on The Rectangular Rapid-Flashing Beacon


A variety of engineering (e.g., geometric design, traffic control device) treatments are available with the potential of improving safety at pedestrian crossings. Research studies have been conducted across the United States and in a number of other countries to understand better the effects of these treatments. This appendix contains summaries of a selection of treatments, along with reported results on their effectiveness. The list of treatments considered for this report is provided in table 183.

Table 183. Pedestrian treatments for unsignalized locations.

Included in this appendix?
CMF available?
Advance stop or yield line and sign
Barrier—roadside/sidewalk (railing or fencing)
Bus stop location
Circular beacons
Crosswalk marking patterns
Curb extensions
Flags (pedestrian crossing)
In-roadway warning lights
In-street pedestrian crossing signs
Marked crosswalk
Motorist warning signs
Overpasses and underpasses
Pedestrian hybrid beacon (PHB) (also known as HAWK)
Puffin crossing
Raised crosswalks
Rectangular rapid flashing beacon (RRFB)
No—see chapter 2
Refuge island
Road diet
Zigzag lines
Leading pedestrian interval
No—signal treatment
No right turn on red
No—signal treatment
Pedestrian countdown
No—signal treatment
Pedestrian scramble
No—signal treatment
No—signal treatment

aHighway Safety Manual.(46)
bFHWA CMF Clearinghouse.(47)
Puffin = Pedestrian User-Friendly Intelligent.
CMF = Crash Modification Factor.


Advance yield lines (i.e., pavement markings) place the traditional stop or yield line 30 to 50 ft upstream of the crosswalk and are often accompanied by YIELD HERE TO PEDESTRIAN signs. Advance yield lines address the issue of multiple-threat crashes on multilane roadways, where one vehicle stops for a pedestrian in the crosswalk but inadvertently screens the pedestrian from the view of vehicles in other lanes. Several studies have documented that advance yield lines decrease pedestrian–vehicle conflicts and increase driver yielding at greater distances from the crosswalk. (See references 48 through 51.)

Studies by Van Houten and others have demonstrated the effectiveness of advance yield lines and YIELD HERE TO PEDESTRIAN signs.(49,50,51) This research found a marked reduction in motor vehicle–pedestrian conflicts and an increase in motorists yielding to pedestrians at multilane crosswalks with an uncontrolled approach. These results have been documented at crosswalks with and without amber flashing beacons. Van Houten and Malenfant also demonstrated that the markings and sign together were more effective than the sign alone.(50) In a 2001 study by Van Houten et al., advance yield lines and YIELD HERE TO PEDESTRIAN signs were shown to reduce vehicle–pedestrian conflicts by 67 to 87 percent.(51) The study also found a large increase in the distance at which motorists yielded to pedestrians. These evaluation results were further replicated at 24 additional study sites located in Canada.(52) Results showed that the advance yield sign and advance yield markings reduced the percentage of motor vehicle–pedestrian conflicts involving evasive action and increased the percentage of motorists yielding to pedestrians and yielding further back from the crosswalk line. Treatments were applied only to streets posted at 30 mi/h.

A 2011 paper reported on the installation of advance yield markings with a YIELD HERE TO PEDESTRIAN sign at two midblock locations in Las Vegas.(53) Results indicated that there was an increase in the proportion of drivers yielding to pedestrians at the location with a five-lane cross section, an ADT of 17,100 vehicles/day, and a posted speed limit of 35 mi/h. The increase in driver yielding was not statistically significant at the location with seven-lane cross section, an ADT of 43,000 vehicles/day, and a posted speed limit of 30 mi/h.


Placing a barrier in a median is a pedestrian crossing treatment discussed in a review of pedestrian safety research by Campbell et al.(54) The purpose of barriers in the median is to discourage pedestrians from crossing at undesirable locations and encourage them to cross at a crosswalk. As part of a larger test of various pedestrian countermeasures, median fence barriers were installed at two sites: one in Washington, DC, with a 4-ft fence, and one in New York City, with a 6-ft fence.(55) The median fence barrier at one site consisted of two gaps, each located at an intersecting minor street. After installation of the barrier, researchers interviewed pedestrians to gauge their reactions to the treatment. Regarding crosswalk use, a reported 61 percent of the pedestrians identified the barrier as the reason for using the crosswalk. When asked whether the barrier affected the manner in which they crossed the street, 52 percent stated it had no effect, while 48 percent indicated the only effect was to force them to cross at the intersection. Of those who were crossing midblock before the installation, 61 percent did so out of convenience, and about one-third indicated they would use the crosswalk only if midblock traffic volumes were “very heavy.” After the fence was installed, 32 percent of the 22 pedestrians interviewed who previously made midblock crossings stated inconvenience as the major factor, with high turning volume at the intersection as a close second (23 percent). In particular, older pedestrians were generally concerned with turning traffic at intersections, and many cited recent crash experience as a concern. Almost one-quarter of those interviewed indicated they had walked along the median to the end of the barrier, or an opening, before completing the crossing. While merchants at a control site indicated they did not anticipate much effect from a median barrier, 58 percent of those at the experimental sites indicated that its major effect was to discourage customers from shopping both sides of the street. Most residents accepted the barrier, only 7 percent wanted it removed, and a few complained about inconvenience and unsightly appearance.


A recent FHWA International Scan found that pedestrian railings were common in the United Kingdom, where they were used to direct pedestrian movements to preferred crossing locations at intersections and in median islands.(56) They also offered a useful guide to pedestrians with visual disabilities. The railings appeared to be most common in areas with high pedestrian traffic.

Campbell et al. discuss several studies in which chains, fences, guardrails, and other similar devices were proposed as a means of channelizing and protecting pedestrians. (See references 54 and 57 through 60.) Parking meter post barriers were tested at three urban areas sites.(55) All of the tests used chains that connected parking meter posts. The barrier was 3 ft high and incorporated as many as three chains. In Washington, DC, six parking meter post barriers were created on one side of a street, resulting in a series of 12-ft single chain sections. In New York City, 19 posts were used, 9 on one side of the street and 10 on the other. These were 12-ft sections with two chains. The third site was a section of one-way street along which three-chain sections were installed on eight posts. Results of the study were mixed, in part because of vandalism (i.e., stolen chains) that interfered with the experiment. Twenty-six percent of those interviewed who crossed at the intersections after the installation mentioned that a factor in their choice of crossing location was that it was illegal to cross elsewhere. Because only 12 percent of those interviewed had mentioned this before the change, the authors surmised that the barriers may have reminded pedestrians that it was illegal to jaywalk. While 65 percent of merchants perceived no negative effects from the countermeasure, 15 percent noted that the chains interfered with street crossings, and 18 percent cited a problem when loading or unloading goods.

In London, research was conducted on an 1,800-ft road segment with pedestrian barriers on both sides. Access openings on each side of the road were not located directly across from each other.(54,61) Researchers mapped pedestrian crossing movements and compiled crash data from the site. Crashes during the previous 8 years were shown as a ratio to a 4-hr pedestrian volume, which was fewer than 20,000 people. The resulting risk ratio was compared with that for 11other sites in London that did not have pedestrian barriers. The only significant difference in risk ratio occurred at midblock crossings located within 150 ft of a signalized intersection (these locations had more than twice the risk ratio with the pedestrian barrier) and at other midblock locations within 60 ft of an intersection (where controlled crossings were not present and had approximately 10 times the risk ratio). The overall risk ratio was lower at the test site but was not found to be statistically significant. Researchers also studied the longitudinal path taken by each pedestrian; this path was the distance between barrier openings used to get on and off the roadway, measured parallel to the curb. The results indicated most pedestrians would cross outside of the crosswalk when the longitudinal distance between barrier openings on either side of the street was less than 30 ft. The author suggested that longitudinal distances between the openings on opposite sides of a street should be greater than 30 ft.

Pedestrian barrier fences were installed along 18 sections of roadway in Tokyo.(62) Analysis of crashes before and after the installation revealed that crashes related to crossing pedestrians declined by nearly 20 percent. Researchers observed an overall reduction of 4 percent, including non-pedestrian crashes. It had been thought that even though crashes related to pedestrians crossing out of crosswalks might decrease, crashes related to pedestrians crossing in the crosswalks might increase. However, the results indicated that both types of pedestrian collisions were reduced by an equal percentage.


TCRP Report 125: Guidebook for Mitigating Fixed-Route Bus-and-Pedestrian Collisions provides information on pedestrian–bus crashes and countermeasures and strategies for reducing these crashes.(63) Lack of pedestrian-friendly environments was noted as one of the factors. This includes sidewalk conditions such as broken and uneven sidewalks, narrow sidewalks, sidewalk obstacles, and lack of sidewalks or other positive separation. Lack of lighting was another concern noted.

According to Campbell et al., 2 percent of all pedestrian collisions in urban areas can be classified as pedestrian collisions at bus stops.(54) Most do not involve a pedestrian being struck by a bus, but the bus creates a visual screen between approaching drivers and pedestrians crossing in front of the bus. In rural areas, pedestrian crashes related to school bus stops were identified in 3 percent of all pedestrian crashes. A countermeasure proposed for urban crashes involved relocating bus stops to the far side of intersections to encourage pedestrians to cross behind rather than in front of the bus. This allows the pedestrian to be seen and to see oncoming traffic closest to the bus. To determine the effect of such relocation on pedestrian crossing behavior, two before–after studies evaluated bus stop relocations. One site in Miami, FL, was located on a two-way, four-lane street intersecting with a two-way, two-lane street at an unsignalized location. The other site was located in San Diego, CA, on a two-way, four-lane street intersecting with a one-way, three-lane street at a signalized intersection with pedestrian signals.(54,55) The relocation of the bus stops to the far side eliminated the undesired crossing behavior; previously, half of the riders crossing the street after disembarking crossed in front of the bus.

An analysis of pedestrian crashes in Sweden found school bus stops should be located with greater consideration of pedestrian safety factors.(54,64) Campbell reported that Swedish researchers drew the following conclusions about the location of bus stops:

  • Ensure they are not hidden by vegetation or other obstacles.
  • Place them away from roadway curves or superelevated locations.
  • Provide adequate standing and playing area for the waiting passengers.
  • Provide maximum sight distance to all critical elements.

Additional guidance for the location and design of bus stops is provided in TCPR Report 19.(65)


The use of circular flashing beacons for pedestrian crossings is prevalent in the United States. In some instances, there are concerns that the overuse of flashing beacons or the continuous flashing at specific locations has diluted their effectiveness in warning motorists of a pedestrian crossing. Flashing beacons have been installed in a variety of ways, including the following:

  • At the pedestrian crossing, both overhead and side mounted.
  • In advance of the pedestrian crossing, both overhead and side mounted.
  • In conjunction with or integral within other warning signs.
  • In the roadway pavement itself (see section on in-roadway warning lights).

The operations for flashing amber beacons may also vary, including the following:

  • Continuous flash mode.
  • Pedestrian-activated using manual pushbuttons.
  • Passive pedestrian detection using automated sensors (e.g., microwave, video).
  • Different flash rates, sequences, or strobe effects.

The experience with flashing beacons has been mixed, as would be expected when they have been installed in numerous ways. Several studies have shown that intermittent (typically activated using a manual pushbutton or automated sensor) flashing beacons provide a more effective response from motorists than continuously flashing beacons.(66,67) These beacons do not flash constantly; thus, when they are flashing, motorists can be reasonably assured that a pedestrian is crossing the street. With pedestrian activation, special signing may be necessary to ensure that pedestrians consistently use the pushbutton activation. Alternatively, automated pedestrian detection has been used with some success but typically requires extra effort in installation and maintenance.

Overhead flashing beacons appear to have the best visibility to motorists, particularly when used both at and in advance of the pedestrian crossing. Many installations have used both overhead and side-mounted beacons. The effectiveness of the flashing beacons in general, however, may be limited on high-speed or high-volume arterial streets. For example, overhead flashing beacons have produced driver yielding behavior that ranges from 30 to 76 percent, with the median values falling in the mid-50 percent range; however, the evaluations did not contain enough information to attribute high or low driver yielding values to specific road characteristics. (See references 48, 66, 67, and 68.) The field studies conducted in a TCRP/NCHRP project (documented in TCRP Report 112/NCHRP Report 562) found a similarly wide range of motorist yielding values (25 to 73 percent), with the average value for all flashing beacons at 58percent.(44) The analysis of site conditions and traffic variables also found that traffic speeds, traffic volumes, and number of lanes have a statistically significant effect on driver yielding behavior on arterial streets.

Little and Saak evaluated two installations of pedestrian-activated overhead yellow flashing beacons.(69) Both sites consisted of a five-lane cross section and ADTs of either 7,500 or 18,400vehicles/day. The motorist compliance at the sites was 64 to 65 percent during the day and 68 to 78 percent at night.

Van Winkle and Neal evaluated the use of pedestrian-actuated advance and crosswalk flashers in Chattanooga, TN.(66) The installation of the crosswalk flashers was a compromise solution for a group of senior citizens that demanded a traffic signal so that they could cross a minor arterial street with speed limit of 40 mi/h. City staff conducted a before-and-after study in 1987, with follow-up data collection in 2000. Data were collected on the percentage of drivers yielding or slowing at the pedestrian crosswalk. The original 1987 data showed that driver yielding improved from 11 to 52 percent in the eastbound direction and 6 to 32 percent in the westbound direction. The percentage of drivers yielding at this location has been sustained as a long-term improvement; driver yielding in 2000 was measured to be 55 percent in the eastbound direction and 45 percent in the westbound direction. The authors attribute the success of the flashers to pedestrian actuation. The city of Chattanooga has installed similar flashing crosswalk warning devices at three other locations with what it characterizes as similar results, although no formal studies of their effectiveness have been conducted.

Sparks and Cynecki reported in 1990 on the use of flashing beacons for warning of pedestrian crosswalks in Phoenix, AZ.(70) The city evaluated the application of advance warning flashing beacons at four pedestrian crossing locations. The authors describe the use of several experiments in their evaluation, including before-and-after speed and crash data collection as well as treatment-and-control experiments for traffic speeds. The authors found that the advance warning flashing beacons did not decrease speeds or crashes, and in some cases, the traffic speeds or crashes increased after installation of the flashing beacons. These findings led the authors to conclude the following:

  • …[F]lashers offer no benefit for intermittent pedestrian crossings in an urban environment. In addition, the longer the flashers operate the more it becomes part of the scenery and loses any effectiveness.(70) (p. 35)

The authors do concede that actuated warning flashers may be beneficial in a high-speed rural environment with unusual geometrics, high pedestrian crossings, and unfamiliar drivers; however, these conditions were not tested in their study.


In a 2009 FHWA study of crosswalk markings, researchers investigated the relative daytime and nighttime visibility of three crosswalk marking patterns: bar pairs, continental, and transverse lines.(31) In the study, conducted on the campus of Texas A&M University in College Station, TX, information was collected on the distance from the crosswalk at which 78 participant motorists verbally indicated visual recognition of the crosswalk with the different patterns. The participants were about evenly divided in gender between males and females and in age between younger than 55 years old and older than 55.

The researchers used instrumented vehicles on a route along open roads on the campus. The research team collected data during two periods: daytime (sunny and clear or partly cloudy) and nighttime (street lighting on). The tests used existing markings (six intersection and two midblock locations) and new markings installed for the study (nine midblock locations).

For the study sites, the findings indicate that the marking type (bar pair, continental, or transverse) was statistically significant. The detection distances to bar pairs and continental markings were statistically similar, and they were statistically longer than the detection distance to the transverse markings, both during the day and at night.

For the existing midblock locations, the drivers detected the continental markings at about twice the distance upstream as the transverse markings during daytime conditions. This increase in distance translates to 8s of increased awareness of the presence of the crossing at 30-mi/h operating speeds.

The participants also rated the appearance of markings on a letter-grade scale of A to F. The researchers compared those subjective ratings of visibility for all the groups and variables identified in the preceding analysis. The ratings for bar pairs and continental were consistent over various comparison groups, with better ratings for bar pairs and continental markings than for transverse markings. These results mirrored the findings from the evaluation of detection distances. Overall, participants preferred the continental and bar pairs markings over the transverse markings.

The research team worked with the NCUTCD to develop recommendations for incorporating the findings from the study into the MUTCD. The recommendations were endorsed on June 23, 2011. Figure 103 shows the proposed figure for inclusion in the next edition of the MUTCD.

Figure 103. Diagram. Examples of crosswalk markings (figure proposed to replace existing MUTCD figure 3B-19). Illustration of the following crosswalk markings: two transverse lines, bar pairs, continental, and ladder, including providing the minimum length of 6 ft and dimensions for unique stripes of 8 to 24 inches. The bar pairs, continental, and ladder markings are considered high-visbility markings while the two transverse lines are known as basic crosswalk markings. The drawing includes a note that says “at a non-intersection uncontrolled pedestrian crossing where the speed limit is greater than 35 mph, the high visibility crosswalk marking, if used, should not be less than 8 feet wide.”

Figure 103. Diagram. Examples of crosswalk markings (figure proposed to replace existing MUTCD figure 3B-19).(71)

A 2010 paper presented the findings from an empirical Bayes evaluation of high-visibility school (yellow, continental-style) crosswalks in the city of San Francisco, CA.(72) The analysis used data for 54 treated intersections with high-visibility crosswalks and 54 control intersections, each chosen for its geographical proximity to a treated intersection. The study found a statistically significant reduction in crashes of 37 percent for intersections with high-visibility school crosswalks.


The purpose of a curb extension, also known as a choker, curb bulb, or bulbout, is to reduce the width of the vehicle travel way at either an intersection or a midblock pedestrian crossing location. It shortens the street crossing distance for pedestrians, may slow vehicle speeds, and provides pedestrians and motorists with an improved view of one another, thereby reducing the risk of a motor vehicle-pedestrian collision. Campbell et al. identify multiple studies of variations of this treatment in Australia, the Netherlands, and Canada.(54)

In two Australian cities (Keilor, Queensland, and Eltham, Victoria), researchers indicated that “curb blisters” had little effect on reducing vehicle speeds.(54,73) However, in Concord, New South Wales, researchers compared a subarterial street treated with both curb blisters and marked parking lanes to an untreated street; the comparison showed that the crash rate on the treated street was only one-third that of the untreated street. The number of these crashes involving pedestrians was not stated, nor is it known how the streets compared before treatment.

Australia’s “wombat” crossings usually consist of a raised platform with a marked crosswalk on top, and a refuge and curb blisters where space permits. Thus, they combine features of both speed tables and bulbouts. They are designed to slow motorists, shorten pedestrian exposure to motor vehicles, and increase pedestrian visibility to motorists. Reports of those studies indicate that wombat crossings have generally reduced 85th percentile vehicle speeds by 40 percent.(54,73)

The Dutch towns of Oosterhout and De Meern installed variations of street-narrowing treatments. The Oosterhout project consisted of installing two bulbouts so as to require motorists to deviate from a straight path. Both the 85th percentile vehicle speed and the degree of pedestrian–motor vehicle conflict fell after the deviation was installed. De Meern’s path deviation was created by placing two bulbouts opposite one another to narrow the width of the traveled way. Researchers did not observe a significant reduction in the 85th percentile vehicle speed, and opinions of the treatment were mixed. Residents did not express a strong sense of neighborhood improvement, swerving cars were thought to endanger bicyclists, school teachers thought that children would be confused by the deviation, retailers were concerned about accessibility and parking, and there was some concern about emergency vehicle access.(54,74)

Macbeth reported favorable speed changes seen on five raised and narrowed intersections and seven midblock bulbouts (two raised) in Canada.(54,75) The speed limit was also lowered to 19mi/h. The results of the speed changes are presented in table 184.

Table 184. Speed changes due to bulbouts.(54)


Percent exceeding

19 mi/h
25 mi/h
31 mi/h

Huang and Cynecki reported in 2001 the effects of bulbouts at various locations to determine their effects on selected pedestrian and motorist behaviors.(76) At four intersections in Cambridge, MA, and Seattle, WA, they found no significant effect on motorists yielding to pedestrians in crosswalks, as shown in table 185.

Table 185. Percentage of motorists yielding to pedestrians at bulbout crosswalks.(76)

Cambridge, MA
20.0 (5)
66.7 (6)
Seattle, WA
57.9 (342)
52.2 (471)

aSample size in parentheses.
bSmall sample size.
cNot significant at 0.10 level.

Huang and Cynecki used a treatment-and-control study approach to evaluate four additional bulbouts in Greensboro, NC, and Richmond, VA. Because of low pedestrian activity in both Greensboro and Richmond, it was necessary to stage pedestrian crossings, using a two-person data collection team. Motorists stopped for fewer than 10 percent of the staged pedestrians in both cities. The differences between the treatment and control sites were not statistically significant at the 0.10 level, as shown in table 186.

Table 186. Percentage of motorists stopping for staged pedestrians at bulbout crosswalks.(76)

Greensboro, NC
5.2 (211)
7.6 (185)
Richmond, VA
0.0 (66)
0.0 (66)

aSample size in parentheses.
bNot significant at 0.10 level.


Several cities (e.g., Salt Lake City, UT; Kirkland, WA; Berkeley, CA) use fluorescent orange flags that are carried by crossing pedestrians. The research team found no formal studies in the literature on the effectiveness of crossing flags; however, anecdotal information has indicated that these crossing flags are effective in improving driver yielding behavior. The flags in Salt Lake City are used mostly on streets near the downtown area that have speed limits of 30 mi/h or less. Several of these streets, however, are multilane, high-volume arterials. Field studies conducted in a TCRP/NCHRP project (documented in TCRP Report 112/NCHRP Report 562) found pedestrian crossing flags in Salt Lake City and Kirkland were moderately effective.(44) The study sites with crossing flags had motorist yielding rates that ranged from 46 to 79 percent, with an average of 65 percent compliance. Several of the study sites had four or more lanes with speed limits of 30mi/h or 35 mi/h.


At certain locations, site characteristics can make a crosswalk difficult for the driver to see at night or in dusk/dawn settings. Trees, shadows, or glare from nearby buildings, and roadway alignment can all affect the ability of approaching drivers to see a crosswalk or pedestrians who use it. Adding illumination can improve the visibility and the safety of such crosswalks. Campbell et al. discuss three studies on illumination in Australia, Israel, and the United States, which are summarized in the following paragraphs.(54)

Pegrum conducted a two-stage study of lighting of pedestrian crossings in Perth, Australia.(54,77) A pilot study showed sufficient success to initiate a broader scale lighting program, in which 63sites were studied. The illumination consisted of two luminaires (100-watt sodium lamps), one on each side of the roadway at either side of the crosswalk, mounted approximately 12 ft from the crosswalk at a height of 17 ft and aimed at a point 3 ft above the pavement. Campbell states that Pegrum reported the sodium floodlighting resulted in a significant decrease in nighttime pedestrian crashes; a summary of crashes is shown in table 187.

Table 187. Crash effects of providing sodium floodlights at pedestrian crossings in
Perth, Australia.(54)


Study Period

Pedestrian crashes (fatalities)

Vehicle-only crashes (fatalities)


Pilot Test: 6 crossings

5 years before
19 (1)
7 (1)
26 (2)
5 (0)
1 (0)
6 (0)
5 years after
21 (1)
2 (0)
23 (1)
9 (0)
0 (0)
9 (0)

Follow-On Test: 57 additional crossings

2 years before
57 (2)
32 (1)
89 (3)
19 (0)
2 (0)
21 (0)
2 years after
58 (2)
13 (1)
71 (3)
18 (1)
1 (0)
19 (1)

Polus and Katz developed and tested a combined illumination and signing system for pedestrian crosswalks in Israel.(54,78) Reported changes in nighttime crashes at the 99 illuminated study sites and 39 unilluminated control sites are summarized in table 188, which shows a noticeable decrease in nighttime crashes at the study sites compared with increased crashes at the control sites. The authors concluded that crash reductions were primarily the result of the illumination, because daytime crashes were largely unchanged. Campbell also states that the authors studied other possible influences, including changes in pedestrian and vehicle flow, weather differences, and national crash trends, but none showed any effect on the results.

Freedman et al. conducted a study in Philadelphia to assess the impacts of installing improved lighting at seven sites.(54,79) The impacts were evaluated on the basis of behavior as measured for 728 pedestrians and 191 drivers at the 7 study sites and 7 control sites. The researchers reported that the study sites were high-crash locations, while the control sites were low-crash locations. The illumination improvement consisted of 90-watt low-pressure sodium lamps. Each system was controlled by a photocell that energized the circuit at sundown and turned it off at sunrise, with a provision for experimenter override.

Table 188. Effects of crosswalk illumination on nighttime pedestrian crashes in Israel.(54)


Nighttime crashes

Illuminated Study Sites (99)
Unilluminated Control Sites (39)

The Philadelphia evaluation contained a comparison of changes in five pedestrian attributes—search behavior, crossing path, concentration, erratic behavior, and clothing brightness—before and after lighting improvements. According to Campbell’s summary of the researchers’ report, the comparison showed that “perceived clothing brightness” increased significantly after installing the special illumination.(54) Observers who searched the street in a manner similar to drivers perceived the general appearance of pedestrians as brighter. The researchers also reported significant improvement in the apparent concentration of pedestrians to the crossing task at all signalized locations, and search behavior was found to improve significantly under all conditions. Drivers appeared more aware of approaching crosswalks when the illumination was present. Campbell added a note that changes in the number of crashes at both groups of sites moved toward the mean, consistent with what would be expected because one group consisted of high-crash sites and the other consisted of low-crash sites; however, he reported that the behavioral measures should not have been influenced by regression to the mean.

A recent study in Las Vegas by Nambisan et al, evaluated a midblock crosswalk illumination system with automatic pedestrian detection devices.(80) The “smart lighting” system detected the presence of pedestrians that were using the crosswalk and activated additional lighting during their time within the crosswalk. This strategy was used to address problems related to motorists’ failure to yield and the high proportion of nighttime crashes, and it was thought to be more effective in capturing the attention of approaching drivers than the use of continuous high-intensity lighting in the crosswalk.

Using a before-and-after methodology, the researchers studied the results of the “smart lighting” test based on two MOE categories: safety MOEs, including pedestrian and motorist behaviors, and mobility MOEs, consisting of pedestrian and vehicle delay. Results indicated that safety MOEs improved, as shown in table 189 and table 190. The percentage of increase in the diverted pedestrians from the before to the after condition was reported as statistically significant, as was the decrease in the proportion of pedestrians trapped in the roadway and the improvement in motorist yielding behavior.

Table 189. Results for “smart lighting” pedestrian safety MOEs.(80)

Measure of effectiveness

Before (n = 44)

After (n = 84)

Pedestrians who look for vehicles before beginning to cross
Pedestrians who look for vehicles before crossing second half of street
Diverted pedestrians
Pedestrians trapped in roadway

Table 190. Results for “smart lighting” motorist safety MOEs.(80)

Measure of effectiveness

Before (n = 91)

After (n = 116)


Motorists yielding to pedestrians


Distance motorist stops/yields before crosswalk (ft)

> 20


As a specific design case of flashing beacons, in-roadway warning light installations have proliferated since the 1990s. Their use originated in California and Washington State but has spread to numerous other cities in the United States. In-roadway warning lights are mounted in the pavement near the crosswalk markings such that they typically protrude above the pavement less than 0.5 inch. As with flashing beacons, the experience with in-roadway warning lights has been mostly positive but with a few negative results.

Many early and some current equipment designs for the in-roadway warning lights have been problematic. Some of the problems encountered are as follows:

  • Snow plows damage the flashing light enclosures.
  • Light lenses become dirty from road grit and require regular cleaning.
  • Automated pedestrian detection does not operate effectively.

Many of the early problems have been resolved through experience, but some cities continue to be cautious in specifying more in-roadway warning lights until they have long-term experience. Some cities have noted their preference for overhead flashing beacons instead of in-roadway lights because of poor visibility issues when traffic is queued in front of the in-roadway lights.(67,81) Another concern is that in very bright sunlight, the flashing lights are difficult for drivers to see.

In-roadway warning lights have been evaluated in numerous studies with varying results. It appears that the effectiveness of this treatment varies widely depending on the characteristics of the site and existing motorist and pedestrian behavior.

For most of the installations, in-roadway warning lights have increased driver yielding into the 50- to 90-percent range. (See references 68 and 82 through 86.) In addition, the in-roadway warning lights typically increase the distance that motorists first brake for a pedestrian crossing, indicating that motorists recognize the pedestrian crossing and the need to yield sooner. (See reference 82 through 85.) These results have been even more dramatic at night when the in-roadway warning lights are highly visible. For a few installations, driver yielding decreased or did not increase above 35 percent.(68,69,87) The research team did not include in-roadway warning lights in the early 2000 TCRP/NCHRP project’s field studies because of the abundance of evaluation results in the literature.(44)

On the walkinginfo.org website, Thomas provided a review of an in-roadway warning light (IRWL) system.(88) Nine studies were identified that provided some evaluation of potential safety effects, all using behavioral MOEs. The following results were noted from the review performed by Thomas(88):

  • Short-term improvements in motorist yielding to pedestrians were reported from most sites studied. No improvement or improvement only to low levels was reported for a number of locations, approaches, or study conditions. (See references, 86, 87, 89, and 90.)

  • Trends (from two studies) indicated greater improvements at nighttime; however, effects under other sub-optimal visibility conditions, such as rain or fog, have not been clearly studied.(86,91)

  • There were inconsistent results (between two studies) on whether IRWL improves yielding to pedestrians in the middle of their crossing.(68,89) This MOE may have a greater bearing on safety than yielding for pedestrians waiting or just beginning to cross, but not yet in the path of vehicles. The effect of IRWL on those in the middle of their crossing, particularly for multilane roads should be further studied. In the meantime, caution should be exercised, and perhaps additional treatments implemented, if IRWL is considered for uncontrolled crosswalks at multilane locations.

  • Reported effects on motorist speeds were also mixed, with studies finding the following:

    • Improvements or slight improvement in speeds.(86,92,93)
    • No improvement.(85,89)
    • Mixed results for some locations and study conditions.(86,89)

  • Effects on conflicts between motorists and pedestrians using the crosswalk also varied, along with the definitions of conflicts used in the studies. Authors reported the following:

    • A non-significant increase in conflicts in one study.(92)
    • Reduced conflicts at all four locations in a study from Israel.(90)
    • Reduced conflicts following installation of high-visibility crosswalks and sidewalk improvements, but no improvement related to the IRWL in one study.(94)
    • Fewer conflicts were observed among those using the IRWL crosswalk compared with those crossing at other locations (after period only).(89)

  • Longer-term data are generally lacking. When data were available, improvements in yielding and other measures were typically greatest at the shortest after-interval measured, with worsening trends seen at later time intervals.(86,89,94) Thus, the potential for a degradation of initial improvements is suggested, and the treatment should be monitored at repeated intervals over a year or more. Certainly any available crash data and characteristics should be considered.

  • Most of the studies included one treatment site, and none included comparison sites to control for time-related trends or other unknown factors. Confounding treatments and other conditions were also noted in several of the studies. Most of the studies determined only short-term effects of the treatment, having examined the effects for intervals from a few weeks to several months post-implementation.

The following paragraphs provide additional details for a sample of the evaluations available for in-roadway warning lights.

Whitlock and Weinberger Transportation, Inc., in 1998, summarized the evaluation results of in-roadway warning lights at numerous locations in California.(82) In these installations, the in-roadway warning lights were supplemented with a pedestrian crosswalk sign with warning amber LED lights, as well as a pedestrian-activated pushbutton with flashing LEDs and a CROSS WITH CAUTION sign. Two different MOEs were used to report evaluation results: 1)percentage of motorists yielding to pedestrians and 2) advance vehicle braking distance. For all six study sites, the percentage of motorists yielding to pedestrians increased after treatments were installed; daytime yielding improved from 28 to 53 percent, and nighttime yielding increased from 13 to 65 percent. The improvements in motorist yielding behavior and the actual percentage of yielding motorists were typically much greater for nighttime conditions. The changes in advance vehicle braking distance showed similar results, with increases in braking distance being greater during nighttime conditions.

The city of Kirkland, WA, installed in-roadway warning lights at two midblock locations in the fall of 1997.(84) Whitlock and Weinberger Transportation, Inc. evaluated the crossing treatments at these locations and reported the results using the same two MOEs as the California study. The evaluation team found improvements to both MOEs after installation, with more dramatic improvements evident during nighttime tests. Before installation, nighttime driver yielding ranged from 16 to 65 percent. After installation of the in-roadway warning lights, yielding increased to a range of 93 to 100 percent. Daytime yielding improved from 46 to 64 percent before treatment to 85 to 94 percent after. The study found the following:

  • “The concept of amber flashing lights embedded in the pavement at uncontrolled crosswalks clearly has a positive effect in enhancing a driver’s awareness of crosswalks and modifying driving habits to be more favorable to pedestrians.”(82) (p.1)

Boyce and Van Derlofske compared the effectiveness of in-roadway warning lights to basic crosswalk markings at a single location with two crosswalks in Denville, NJ.(94) The authors found that the in-roadway warning lights decreased the speed at which vehicles approached the crosswalk, but that this speed reduction diminished over time. In addition, vehicle–pedestrian conflicts with the in-roadway warning lights also increased over time. The authors also reported several problems with this specific implementation of in-roadway warning lights.

Katz, Okitsu, and Associates in 2000 prepared a study of in-roadway warning lights for Fountain Valley, CA.(85) Their study analyzed the reported safety record of approximately 30 treatment locations that were in place for more than 1 year and compared it with the expected safety record for traditional crosswalk treatments. The system appears to have reduced the crash expectancy by 80 percent; however, it is not known whether this is a novelty effect or will continue over time. The study also found that marked crosswalks with in-roadway flashers had a lower crash rate than comparable marked crosswalks.

Huang et al. documented in 2000 the evaluation of in-roadway warning lights at a single location in Orlando, FL.(87) The evaluation, which was conducted to determine the effects of the in-roadway warning lights on pedestrian and motorist behavior, collected both before-and-after and treatment-and-control data. The authors reported the following results:

  • Average vehicle speeds decreased by 1.9 mi/h when a pedestrian was present and 0.8mi/h when no pedestrians were present, but the decreases were not significant.

  • Vehicle yielding improved from 13 percent before to 34 percent (when flashers were activated) and 47 percent (when flashers were not activated) after installation. The authors could not explain why more drivers yielded when the flashers were not activated.

  • About 28 percent of the pedestrians crossed in the flashing crosswalk when police officers were not present. The remaining 72 percent of pedestrians crossed elsewhere, depending on what was the most convenient path between their origins and destinations.

  • Of the pedestrians who crossed in the flashing crosswalk, 40 percent did not experience any conflicts, compared with 22 percent of those who crossed within 30 ft and only 13percent of those who crossed elsewhere. The researchers concluded that motorists were more likely to stop or slow for pedestrians who crossed in or near the flashing crosswalk than for those who crossed elsewhere.

In a subsequent study, Huang evaluated in-roadway warning lights at two uncontrolled pedestrian crossings—one in Gainesville, FL, and one in Lakeland, FL.(68) The evaluation used traditional before-and-after data collection and used the following MOEs: 1) motorists yielding to pedestrians, 2) pedestrians who had the benefit of motorists yielding to them, 3) pedestrians who crossed at a normal walking speed, and 4) pedestrians who crossed in the crosswalk. The results for these MOEs were quite different between the two study sites. At the study site in Gainesville, driver yielding actually decreased from 81 to 75 percent. Although the decrease was significant, it was considered practically negligible because of site characteristics. At the Lakeland site, driver yielding improved from 18 to 30 percent; this result was not statistically significant because of low sample sizes. The results from the other MOEs were not that informative because major changes were not observed.

Prevedouros in 2001 reported on the evaluation of in-roadway warning lights installed on a six-lane arterial street in Honolulu, HI.(86) The evaluation consisted of a traditional before-and-after study of traffic volumes, vehicle spot speeds, pedestrian crossing observations, and pedestrians’ and motorists’ perceptions of change in the situation. The author reported the following results:

  • A 16- to 27-percent reduction in vehicle speeds was measured when the flashing lights were activated.

  • The average pedestrian wait time at the curb decreased from 26 to 13 seconds, and the average crossing time decreased from 34 to 27 seconds. The crossing time decreased because pedestrians did not have to wait as long in the refuge island before crossing the second direction.

  • The proportion of pedestrians who were observed to run during the crossing decreased from 22 to 12 percent after the flashing lights were installed. The proportion of pedestrians crossing outside the marked crosswalk also decreased from 16 to 8 percent after installation.


In-street pedestrian crossing signs (2003 MUTCD R1-6 and R1-6a signs) are intended for use at uncontrolled (unsignalized) crosswalks. The signs can be installed on the centerline or in the median with either a portable or fixed base. Because the signs are located between the lanes, they can have a traffic-calming effect from the narrowing of the lanes.

A 2009 report documented the findings from three area-wide countermeasure programs implemented in Las Vegas, NV; Miami-Dade, FL; and San Francisco, CA.(4) The three field teams used different applications of the in-street pedestrian signs in terms of location and number of signs used. The signs proved to be very effective in increasing driver yielding (see table 191). Driver yielding increased between 13 and 46 percent depending on the location. There were no significant changes in the percentage of pedestrian–vehicle conflicts at the Miami sites or at two of the three sites in San Francisco. Only one location (Mission & Admiral) in San Francisco experienced a significant decrease in pedestrian–vehicle conflicts. Conflicts were reduced from 17.1 percent in the baseline to 2.1 percent after installation of the signs.

Table 191. Driver yielding at in-street installations.(4)

Before number
After number
Before—percent of drivers yielding to pedestrians
After—percent of drivers yielding to pedestrians
Percent change
Miami: Collins & 6th
Miami: Collins & 9th
Miami: Collins & 13th
San Francisco: 16th & Capp (marked crosswalk)
< 0.01
San Francisco: 16th & Capp (unmarked crosswalk)
< 0.01
San Francisco: Mission & France
< 0.01
San Francisco: Mission & Admiral
< 0.01
Las Vegas: Bonanza between D and F
> 0.05
Las Vegas: Twain between Cambridge and Swenson
< 0.01

aCounterintuitive result—results are not significant because this is a one-tailed test.

A 2007 study compared the effect on driver yielding behavior resulting from the installation of in-street pedestrian crossing signs. The signs were placed at three positions relative to the crosswalk—at 0 ft, 20 ft, and 40 ft in advance of the crosswalk—and three study sites were evaluated in the study.(4) The data showed that the sign produced a marked increase in yielding behavior at all three study sites and that installation of the sign at the crosswalk line was as effective as or more effective than installation of the sign 20 or 40 ft in advance of the crosswalk. The data also indicated that placement of the sign at all three locations at once was no more effective than placement of the sign at the crosswalk line. These data suggest that the in-roadway sign is likely effective because the in-roadway placement is particularly salient to drivers. Because drivers frequently struck the signs at one of the sites, the authors recommended that these signs be placed on median islands whenever possible to extend their useful lives.

In-street pedestrian crossing signs were examined in the TCRP/NCHRP project.(44) The field studies indicated that in-street signs had relatively high motorist yielding (ranged from 82 to 91percent, for an average of 87 percent); all three study sites were on two-lane streets with posted speed limits of 25 or 30 mi/h.

A 2011 paper reported on installations of in-street pedestrian crossing signs at three midblock locations in Las Vegas.(53) The results either 1) showed a decrease in motorist yielding or 2) were not statistically significant. The signs were installed on roads with 35-mi/h speed limits, five- or seven-lane cross-sections, and ADTs between 17,100 and 21,400 vehicles/day. The wide crossing may have contributed to the decrease in motorist yielding.

To improve pedestrian safety at a relatively low cost, the Pennsylvania Department of Transportation has a program to provide Yield-to-Pedestrian Channelizing Devices (YTPCD) to municipalities. YTPCDs are placed on the centerline of a roadway in advance of marked crosswalks to remind motorists of the need to yield to pedestrians. A research report by Strong and Kumar and a paper by Strong and Bachman summarized an evaluation of these devices.(95,96) Behavioral data were collected in 2006 in four different community types (urban, suburban, small city, and college town) before and after installation. Sites included crosswalks at unsignalized intersections (eight sites) and midblock locations (four sites). Speed limits at all sites were either 25 or 35 mi/h. Data were analyzed with respect to whether motorists were more likely to yield to pedestrians. The analysis showed a statistically significant increase in motorist yielding. Table 192 provides the results from the study along with findings reported from other studies.

A series of treatments were installed at a bike trail crossing site in Michigan in a study that examined the effectiveness of a “gateway” in-street sign configuration with the RRFB used alone and in combination.(10) Because of a sharp curve, the posted speed was 25 mi/h, and there were two through lanes (one in each direction) and a center turn lane. When the signs were absent and the RRFB not activated, yielding averaged 20 percent. The RRFB alone produced an average yielding level of 69 percent. The gateway in-street sign treatment, which consisted of in-street signs on the lane line on both sides of the turn lane and on each side of the road, produced 80percent yielding. The combination of the gateway in-street sign configuration and RRFB produced 85 percent yielding. The authors concluded that the data showed that the gateway in-street signs produced effects that were similar to the RRFBs and that the combination of gateway in-street signs and RRFB may produce effects similar to the gateway in-street signs alone, which suggests that the gateway in-street signs can be more cost effective than the more expensive RRFBs.

Table 192. Evaluation results on in-street pedestrian crossing signs.

Measure of effectiveness
Miami, San Francisco, Las Vegas Motorist Yielding Before range of 7 to 74percent After range of 35 to 78percent Between 13 and 46percent increase Pécheux, K., Bauer, J., and McLeod, P. Pedestrian Safety and ITS-Based Countermeasures Program for Reducing Pedestrian Fatalities, Injury Conflicts, and Other Surrogate Measures Draft Zone/Area-Wide Evaluation Technical Memorandum. Contract #DTFH61-96-C-00098; Task 9842. 2009.
TCRP/ NCHRP Motorist Yielding With signs = 82 to 91percent Fitzpatrick, K., Turner, S., Brewer, M., Carlson, P., Ullman, B., Trout, N., Park, E.S., Whitacre, J., Lalani, N., and Lord, D. Improving Pedestrian Safety at Unsignalized Crossings. TCRP Report 112/NCHRP Report 562. 2006.

Pennsylvania (Philadelphia Haverford Township, Pottstown, and West Chester)

Motorists Yielding—Intersection Locations Before = 27 percent After = 59 percent Increase = 30 to 34percent

Strong, C., and Kumar, M. Safety Evaluation of Yield-to-Pedestrian Channelizing Devices. Western Transportation Institute. Montana State University. 2006.

Strong, C., and Bachman, D. Safety Evaluation of Yield-to-Pedestrian Channelizing Devices in TRB 87th Annual Meeting Compendium of papers DVD. 2008.

Motorists Yielding—Midblock Locations Before = 10 percent After = 30 percent Increase = 17 to 24percent

Previous studies as reported by Strong and Kumar or Strong and Bachmann(95,96)

New York State and Portland, OR

Pedestrians for Whom Motorists Yielded +12 percent

Huang, H., Zegeer, C., Nassi, R., and Fairfax, B. The Effects of Innovative Pedestrian Signs at Unsignalized Locations: A Tale of Three Treatments, Report No. FHWA-RD-00-098, Federal Highway Administration, Washington, DC. 2000.

Pedestrians Who Ran, Aborted, or Hesitated -2 percent
Pedestrians Crossing in Crosswalk No change
Cedar Rapids, IA Motorists Yielding +3 to 15 percent Kannel, E.J., Souleyrette, R.R., and Tenges, R. In-Street Yield to Pedestrian Sign Application in Cedar Rapids, Iowa, Center for Transportation Research and Education, Iowa State University, Ames, IA. 2003.
Minnesota Speed Compliance +20 percent Kamyab, A., Andrle, S., and Kroeger, D., Methods to Reduce Traffic Speed in High Pedestrian Areas, Report 2002-18, Prepared for the Minnesota Department of Transportation, St. Paul, MN. 2002.
Madison, WI Motorists Yielding +5 to 15 percent City of Madison Traffic Engineering Division, Year 2 Field Evaluation of Experimental “In-Street” Yield to Pedestrian Signs. City of Madison Department of Transportation, Madison, WI. 1999.

ITS = Intelligent transportation system.
TCRP = Transit Cooperative Research Program.
NCHRP = National Cooperative Highway Research Program.


Zegeer et al. have performed the most authoritative study to date on the effectiveness of crosswalk pavement markings alone as a pedestrian crossing treatment at uncontrolled locations.(45,97) Five years of pedestrian collisions at 1,000 marked crosswalks and 1,000 matched unmarked comparison sites in 30 U.S. cities were analyzed. The study concluded that no meaningful differences in crash risk exist between marked and unmarked crosswalks on two-lane roads or on low-volume multilane roads. The study indicated that as traffic volumes, speeds, and street widths increase, crosswalk markings alone are associated with a greater crash frequency than no crosswalk markings. The study recommendations indicate that the issue should not be whether to provide crosswalk markings on these high-volume, high-speed streets. Instead, the recommendations point to the necessity of using other treatments in addition to crosswalk markings that will provide a safer street crossing for pedestrians.

Koepsell et al. in 2002 published a study of the effects of crosswalk markings on the risk of vehicle–pedestrian crashes involving older pedestrians.(98) The study gathered crash data and other site characteristics (e.g., traffic and pedestrian volumes, traffic speed, signalization characteristics) from six cities in Washington State and California from 1995 to 1999. The study used a case-control design and compared 282 case sites with 564 control sites. After adjusting for the various traffic and pedestrian characteristics, the researchers found that the risk of a pedestrian–vehicle crash was 3.6 times greater at uncontrolled intersections with a marked crosswalk. At intersections with a stop sign or traffic signal, there was “virtually no association between presence of markings and pedestrian-motor vehicle collision risk.”

Knoblauch, Nitzburg, and Siefert reported on a study of the effects of pedestrian crosswalk markings on pedestrian and driver behavior.(99) The study included 11 unsignalized intersections in four cities: Sacramento, CA; Richmond, VA; Buffalo, NY; and Stillwater, MN. The researchers considered the following behavior in the crosswalk markings evaluation:

  • Pedestrian compliance with crossing location.
  • Vehicle speeds.
  • Vehicle yielding compliance.
  • Pedestrian behavior as related to level of caution.

The authors presented the following conclusions:

  • Drivers appeared to drive slower when approaching a marked crosswalk. The speed reductions are modest but evident nonetheless. This finding implies that most motorists are aware of the pedestrian crossing.

  • No changes in driver yielding behavior were observed after the installation of marked crosswalks. This result implies that motorists may be slowing down just in case they are forced to stop by a pedestrian stepping into the roadway.

  • There were no changes in blatantly aggressive pedestrian behavior after installations of marked crosswalks, indicating that pedestrians do not feel overly protected by marked crosswalks.

  • Overall, crosswalk usage increased after marked crosswalks were installed. The authors found that single pedestrians are more likely to use marked crosswalks than a group of pedestrians traveling together.

Gibby et al. analyzed pedestrian-vehicle crash data at 380 intersections on California State highways.(100) The study found that crash rates at marked crosswalks were 3.2 to 3.7 percent higher than crash rates at unmarked crosswalks (after accounting for pedestrian exposure). This result corresponded to earlier work by Herms in San Diego, and also correlates to Zegeer’s study in the late 1990s. The implication is that marked crosswalks alone are not sufficient on multilane streets with high traffic volumes and speeds.

In the late 1960s, Herms examined 5 years of crash experience at 400 unsignalized intersections in San Diego, CA.(101,102) The study found that nearly six times as many crashes occurred in marked crosswalks as in unmarked crosswalks. After accounting for crosswalk usage, the crash ratio was reduced to about three times as many crashes in marked crosswalks. Many have criticized this study as leading to the removal of pedestrian accommodation on city streets. Many now think that crosswalk markings should not be removed in these cases, but rather supplemented with various other types of safety treatments that enable pedestrians to cross busy roadways.


Pedestrian crossing signs were installed at several locations in the Miami metropolitan area.(4) The signs were tested at a midblock section of Collins Avenue in Miami. Collins Avenue has a two-lane cross section with on-street parking, an ADT of 29,500 vehicles/day, and a speed limit of 30 mi/h. Following installation of pedestrian crossing signs, there were no significant changes in average vehicle speed or the percentage of drivers braking when a pedestrian was present. No conflicts were observed in the before or after conditions. The operating speed at the site in the before condition was 10 mi/h below the posted speed limit of 30 mi/h, which was suggested as a reason that a speed change was not observed.

Huang et al. evaluated three innovative pedestrian signing treatments at locations in Seattle, WA; six sites in New York State; Portland, OR; and three sites in Tucson, AZ.(103) The three treatments evaluated were an overhead crosswalk sign, a pedestrian safety cone typically placed in the roadway, and an overhead flashing regulatory sign prompting motorists to stop for pedestrians in the crosswalk. The evaluation used traditional before-and-after data collection for three MOEs: 1) percentage of pedestrians for whom motorists yielded; 2) percentage of pedestrians who ran, aborted, or hesitated; and 3) percentage of pedestrians crossing in the crosswalk. The results of the study are shown in table 193. All treatments except the overhead flashing sign in Tucson resulted in improvements in motorist yielding. The authors indicated that the effectiveness of the flashing regulatory sign may have been limited because it was installed on four- and six-lane arterial streets with speed limits of 40 mi/h. (The other study locations were primarily two-lane streets with speed limits of 25 or 30 mi/h.)

High-visibility signs and markings were examined in the TCRP/NCHRP project (documented in TCRP Report 112/NCHRP Report 562) in 2004.(44) The results demonstrated the effect of higher posted speed limits. One site with high-visibility signs and markings and a posted speed limit of 25 mi/h had a motorist yielding value of 61percent. However, the other two study sites with high-visibility signs and markings and a posted speed limit of 35 mi/h had motorist yielding values of 10 and 24 percent, for an average of 17 percent.

Table 193. Effectiveness of pedestrian treatments at unsignalized locations.(103)

Study Location
Percent of pedestrians for whom motorists yielded
Percent of pedestrians who ran, aborted, or hesitated
Percent of pedestrians crossing in the crosswalk
Overhead crosswalk sign, (1 site in Seattle)
In-roadway pedestrian safety cone (6 sites in New York, 1 site in Portland)
Overhead flashing crosswalk regulatory sign (3 sites in Tucson)


Pedestrian overpasses (bridges) and underpasses (tunnels) allow pedestrians and bicyclists to cross streets while avoiding potential conflicts with vehicles.(104) Because they are expensive to construct, grade separated crossings should be reserved for locations where there is high demand for crossings by pedestrians, bicycles, and individuals with physical disabilities, and the risks of crossing the roadway are high. Ideally, overpasses and underpasses should take advantage of the topography of a site—grade separations are less expensive to construct and more likely to be used if they can help pedestrians avoid going up and down slopes, ramps, and steps.

Zegeer et al. discussed several grade separation treatment studies.(54) An analysis was made of reported pedestrian crashes for 6 months before and 6 months after the installation of pedestrian overpasses at 31 locations in Tokyo, Japan.(54,62) The overall results are shown in table 194. The table shows data for 656-ft sections and 328-ft sections on either side of each site. Crashes determined to be “related” to the treatment (assumed to be pedestrian crossing crashes) decreased substantially after overpass installation, but non-related crashes increased by 23percent in the 656-ft sections. There was also a greater reduction in daylight pedestrian collisions than nighttime collisions.

The effectiveness of pedestrian overpasses and underpasses depends a great deal on their level of use by pedestrians. A 1965 study by Moore and Older found that use of overpasses and underpasses depended on walking distances and convenience of the facility.(54,105) They defined a convenience measure (R) as the ratio of the time to cross the street on an overpass divided by the time to cross at street level. The researchers found that approximately 95 percent of pedestrians will use an overpass if the walking time in using the overpass is the same as crossing at street level (i.e., R = l). However, if crossing using the overpass takes 50 percent longer than crossing at street level (R = l.5), almost no one will use the overpass. Usage of pedestrian underpasses was not as high as overpasses for similar values of R.

Table 194. Comparison of crashes before and after installation of pedestrian overpasses in Tokyo.(54)

Type of Crash

656-ft sections

328-ft sections

Reduction (percent)
Reduction (percent)
Related crashes
Non-related crashes

Accessibility must also be considered when designing grade-separated crossings. A panel of people with disabilities was asked to comment on accessibility issues after using three pedestrian overpasses in San Francisco, CA.(54,106) They identified the following major elements as creating a barrier or hazard to the user with disabilities:

  • Lack of adequate railings to protect pedestrians from drop-offs on overpass approaches.

  • Greater than acceptable cross slopes.

  • No level area at the terminals of the ramps on which to stop wheelchairs before entering the street.

  • Lack of level resting areas on spiral bridge ramps.

  • Railings difficult to grasp for wheelchair users.

  • Lack of sight distance to opposing pedestrian flow on spiral ramps.

  • Use of maze-like barriers to slow bicyclists on bridge approaches that create a barrier to those who use wheelchairs or who are visually impaired.

  • Lack of sound screening on the bridge to permit people with visual impairments to hear oncoming pedestrian traffic and otherwise more easily detect direction and avoid potential conflicts.

A 1980 study by Templer et al. investigated the feasibility of accommodating pedestrians with physical disabilities on existing overpass and underpass structures.(54,107) A review of 124crossing structures revealed that 86 percent presented at least one major barrier to the physically handicapped; the most common barriers were the following:

  • Stairs only (i.e., no ramps for wheelchair users) leading to an overpass or underpass.
  • Ramp or pathway to ramp that is too long and steep.
  • Physical barriers along the access paths on structure.
  • Sidewalk on the structure that is too narrow.
  • Cross slope on the ramp that is too steep.

Various solutions to these access problems were developed and assessed based on cost effectiveness. The Americans with Disabilities Act has since required the barriers to wheelchair users to be removed, requiring more gentle slopes and periodic level areas for wheelchair users to rest. While use of these gentle slopes also makes it easier for bicyclists and other users, it has also greatly increased the length of ramps, which may discourage usage. Methods such as carefully planned fencing have been used to channel pedestrians to the overpasses and underpasses to increase usage and discourage potentially risky at-grade crossings.


The pedestrian hybrid beacon (PHB) is located both on the roadside and on mast arms over the major approaches to an intersection. The head of the PHB consists of two red lenses above a single yellow lens. It is normally “dark,” but when activated by a pedestrian, it first displays a few seconds of flashing yellow followed by a steady yellow change interval, and then displays a steady red indication to drivers, which creates a gap for pedestrians to use to cross the major roadway. During the flashing pedestrian clearance interval, the PHB changes to a wig-wag flashing red to allow drivers to proceed after stopping if the pedestrian has cleared the roadway, thereby reducing vehicle delays.

A recent study conducted a before-and-after evaluation of the safety performance of the PHB.(108) Using an empirical Bayes method, the study evaluations compared the crash prediction for the before period without the treatment to the observed crash frequency after installation of the treatment. To develop the datasets used in the evaluation, the researchers counted the crashes that occurred during the study period, typically 3years before and 3 years after the installation of the PHB.

The researchers created two crash datasets. The first dataset included crashes coded as occurring at the intersecting streets (identified by using street names). The second dataset was a subset of the first dataset and only included those crashes that had “yes” for the intersection-related code in the police report.

The crash categories examined in the study included total, severe, and pedestrian crashes. From the evaluation that considered data for 21 pedestrian hybrid beacon treatment sites and 102unsignalized intersections (reference group), the researchers found the following changes in crashes after installation of the PHBs:

  • A 29-percent reduction in total crashes (statistically significant).
  • A 15-percent reduction in severe crashes (not statistically significant).
  • A 69-percent reduction in pedestrian crashes (statistically significant).

FHWA added the PHB to the MUTCD in the 2009 edition (see chapter 4F).(2) However, the PHBs included in the FHWA safety study differ from the material in the 2009 MUTCD in the following ways because the installations included in the FHWA study preceded the MUTCD guidance:

  • Section 4F.02 of the MUTCD states the following(2):

    When an engineering study finds that installation of a pedestrian hybrid beacon is justified, then … the pedestrian hybrid beacon should be installed at least 100 feet [31 meters] from side streets or driveways that are controlled by STOP or YIELD signs.

    All 21 pedestrian hybrid beacons included in this study are located either at a minor intersection (where the minor street is controlled by a stop sign) or at a major driveway (where the driveway is controlled by a stop sign).

  • The 2009 MUTCD depicts an R10-23 sign with the symbolic red circle and a white background for the word “crosswalk” on the sign.(2) The signs typically used at the PHB locations do not have the symbolic red circle, and the crosswalk background is yellow.

The MUTCD includes guidelines for the installation of the pedestrian hybrid beacons for low-speed roadways where speeds are 35 mi/h or less, and high-speed roadways where speeds are more than 35 mi/h.(2) Changes proposed for the next edition of the MUTCD (i.e., the version that will follow the 2009 edition) is to remove the 100-ft guidance statement and to add text stating that if the PHB is installed at or immediately adjacent to an intersection with a minor street, a stop sign shall be installed for each minor-street approach.

A study in Oregon investigated the public’s understanding of the PHB display.(109) A survey was conducted in Corvallis, which was selected because there were no PHBs installed there, and users would likely be seeing images of the device for the first time. Images of the PHB display were shown to respondents, who were asked questions about what the various indications meant. Survey questions showed only a replication of the PHB display and consistently labeled the device a “signal.” Results of the survey indicated that the PHB was not widely recognized, especially when it was presented out of context, and there was confusion about the sequence of the six indications. The vast majority of respondents answered that they had not seen a PHB before or were not sure whether they had. Of the respondents that said they had seen a PHB before, a large majority (85 percent) responded that it was installed at a rail crossing. Many respondents did not understand the meaning of the various indications of the PHB. Both younger (67 percent) and older (49 percent) drivers responded correctly to the dark indication, stating that they knew to continue through the signal. Most (71 percent) responded that it was necessary to stop for the solid red indication, but there appeared to be confusion with the alternating flashing red indication. A low percentage of drivers correctly responded that they must stop but could proceed through if the crossing was clear. Researchers concluded that a public education campaign on the different indications of the device should precede the deployment of a new installation.


A recent British report documented safety at pedestrian user-friendly intelligent (Puffin) crossings.(110) Puffins have the following characteristics(56,110):

  • Nearside pedestrian signals that encourage pedestrians to view oncoming traffic.

  • No flashing pedestrian green period as at Pelican crossings (i.e., an acronym for pedestrian light control; a British pedestrian crossing with traffic signals that are controlled by pedestrians) or pedestrian signal blackout period at junctions (simplifies pedestrian signal phasing to “green man” for walk and “red man” for don’t walk and eliminates a flashing “don’t walk” for don’t start phase).

  • On-crossing pedestrian detectors that provide an extension to the pedestrian clearance period while pedestrians are still within the crossing.

  • No flashing amber traffic period as at Pelican crossings.

  • An indicator light that confirms when the pedestrian signal has been activated.

  • Pedestrian curbside detectors to cancel the pedestrian demand if there are no pedestrians in the wait areas.

Puffins were developed to replace Pelican crossings at midblock sites and far-side pedestrian signals at junctions. As reported by Maxwell et al., previous research has shown that, compared with existing pedestrian signal facilities, Puffin facilities can reduce both driver and pedestrian delay at junctions and improve pedestrian comfort (particularly for older pedestrians and those with impaired mobility).(110) The aim of the Maxwell et al. study was to quantify the safety benefit. Accident data was analyzed from 50 sites (40 midblock crossings and 10 junctions) that had been converted to Puffin facilities from Pelican crossings and far-side pedestrian signals at junctions. The sites had no other significant changes in layout or operation, and were in general conformance with current Puffin guidance. Statistical analysis was undertaken by using a generalized linear model, which included time trends and seasonal factors. Midblock Puffin crossings were shown to be safer than Pelican crossings as follows:

  • 17 percent lower at the midblock sites (statistically significant at the 5-percent level).

  • 19 percent lower over all the sites (statistically significant at the 5-percent level).

  • 24 percent lower for all pedestrian accidents (statistically significant at the 10-percent level).

  • 16 percent lower for all vehicle accidents (statistically significant at the 10-percent level).


Huang and Cynecki evaluated three raised crosswalks in Durham, NC, and Montgomery County, MD, using a treatment-and-control study approach.(76) All three sites were on two-lane, two-way roadways. One site in Durham also had a continuously operating overhead flashing beacon in addition to the raised crosswalk treatment, and staged pedestrians were used at the Maryland site. The researchers found that speeds at the treatment sites were lower than at nearby control sites (table 195), but motorist yielding behavior was mixed (table 196).

Table 195. Comparison of vehicle speeds at raised crosswalks.(76)


50th percentile speed (mi/h)

Significance at 0.05 level or bettera

Treatment site
Control site
Difference in speeds
Durham, NC—Research Drive
Durham, NC—Towerview Drb
Montgomery County, MDc

aSignificance based on two-tailed test.
bTowerview site had an overhead flashing beacon in addition to the raised crosswalk.
cSpeeds at Montgomery County site were measured only when the staged pedestrian was present.

Table 196. Pedestrians at raised crosswalks for whom motorists stopped.(76)

Treatment sitea (percent (sample size))
Control sitea (percent (sample size))
Durham, NC—Towerview Drive
79.2 (159)
31.4 (35)
Y (0.000)
Montgomery County, MD
1.2 (169)
1.0 (198)

aSample size in parentheses.
bY = Significant at the 0.10 level or better (p-value in parentheses); N = Not significant at the 0.10 level or better.


See discussion in chapter 2 of this report.


Crossing the street can be a complex task for pedestrians. Pedestrians must estimate vehicle speeds, adjust their walking speeds, determine adequacy of gaps, predict vehicle paths, and time their crossings appropriately. Drivers must see pedestrians, estimate vehicle and pedestrian speeds, determine the need for action, and react accordingly. At night, darkness and headlamp glare make the crossing task even more complex for both pedestrians and drivers.(111) Some midblock crossings may be too wide to be crossed during available gaps without the protection of a signal. Median refuge islands simplify the street crossing task by permitting pedestrians to make vehicle gap judgments for one direction of traffic at a time. Recent refuge island designs can incorporate an angled or staggered pedestrian opening, which better aligns pedestrians to face the second direction of oncoming traffic. Refuge areas may be delineated by markings on the roadway or raised above the surface of the street.

Bowman and Vecellio in 1994 reported comparisons of several kinds of medians, including undivided multilane roadways, TWLTL, and raised curb medians.(54,112,113) Raised curb facilities were associated with lower pedestrian crash rates, but the authors reported that both raised and TWLTL medians significantly reduced the number and severity of vehicular crashes at the study sites. In general, raised curb medians may be better than TWLTL medians which are, in turn, better than undivided highways, but the literature search did not conclusively find that medians improved pedestrian safety.(113)

A study by Bacquie et al. compared median refuge islands and split pedestrian crossovers in an analysis of crash reports at 10 crossing locations in Toronto, ON.(114) The split pedestrian crossover treatment includes a median refuge island with pedestrian-activated signal control. The crash data were not normalized by exposure data, but some indication was given about pedestrian and vehicle exposure for the two treatments. The study found that pedestrians were seldom struck while standing on the refuge island and were more often struck while crossing, due to poor gap judgment or improper driver yielding. Vehicle rear-end collisions were higher at the split pedestrian crossovers, and researchers surmised it was because it was a less common treatment than traditional intersection signals. The authors indicated some drivers did not act uniformly when approaching the split pedestrian crossovers because the drivers might not know when to stop or whether other drivers would stop in front or behind them.

Huang and Cynecki evaluated five refuge islands in Corvallis, OR, and Sacramento, CA, using a before-and-after study approach.(76) The Corvallis site was on a four-lane urban arterial with a center left-turn lane, while the Sacramento sites were on intersections of two-way, two-lane residential streets. The authors reasoned that, because refuge islands constrict the roadway and slow vehicle speeds, the islands would increase the number of motorists yielding to pedestrians. In other words, more pedestrians would have the benefit of motorists yielding to them. However, none of the treatments had a statistically significant effect on motorist yielding, as shown in
table 197.

Table 197. Pedestrians at refuge islands for whom motorists yielded.(76)

Beforea (percent (sample size))
Aftera (percent (sample size))
Corvallis, OR
5.7 (35)
7.5 (53)
Sacramento, CA
32.6 (46)
42.1 (38)

aSample size in parentheses.
bNo = Not significant at 0.10 level; Small = Small sample size.

Median refuge islands were installed at two signalized intersections in San Francisco and a midblock location in Las Vegas. Pécheux et al. reported that there were no measurable changes in the percentage of pedestrians trapped in the roadway, the percentage of pedestrians that were diverted to the crosswalk, or the percentage of pedestrian-vehicle conflicts at any of the sites where data for these MOEs were collected.(4) They also found no significant impacts on drivers’ yielding behavior at the intersection locations, but yielding increased significantly at the midblock location, as shown in table 198. The researchers surmised that the installation of a median refuge island at a midblock location was effective in increasing driver yielding to pedestrians and reducing pedestrian delay, while the median refuge islands at the signalized intersections in San Francisco appeared to be less effective in altering driver and pedestrian behaviors.

Pécheux et al. also reported on an offset pedestrian opening in two other median islands.(4) The offset is a type of channelization that encourages pedestrians to turn and walk parallel to the traffic they are crossing; it provides refuge for pedestrians in terms of physical separation from traffic and ensures they are facing the traffic before crossing the second half of the roadway. The crosswalk was created using waist-high bollards and raised medians; the offset at the other study site was developed through median cutouts in an existing raised median, and a new marked crosswalk was added. At both locations the percentage of pedestrians trapped in the roadway fell significantly, particularly at the Lake Mead site with a 57 percent decrease (see table 199). Researchers suggested that the large percentage of pedestrians trapped at the Lake Mead site in the before condition was likely caused by the absence of a marked crosswalk. The research team also measured large, significant increases in driver yielding at both sites as shown in table 200.

Table 198. Drivers yielding to pedestrians at median refuge islands.(4)


Site (location)

Percent of drivers yielding to pedestrians

Percent change



San Francisco

Geary & Stanyan (Intersection)
(n = 158)
(n = 164)
Geary & 6th (Intersection)
(n = 186)
(n = 262)
Las Vegas
Harmon: Paradise Rd. to Tropicana Blvd. (Midblock)
(n = 77)
(n = 284)
< 0.001

Table 199. Trapped pedestrians at offset median openings.(4)



Percent of drivers yielding to pedestrians

Percent change


Las Vegas
Maryland Pkwy & Dumont
(n = 631)
(n = 198)
< 0.001
Las Vegas
Lake Mead: Belmont to McCarran
(n = 61)
(n = 123)
< 0.001

Table 200. Drivers yielding to pedestrians at offset median openings.(4)



Percent of drivers yielding to pedestrians

Percent change


Las Vegas
Maryland Pkwy & Dumont
(n = 432)
(n = 246)
< 0.001
Las Vegas
Lake Mead: Belmont to McCarran
(n = 296)
(n = 117)
< 0.001


A road diet involves narrowing or eliminating travel lanes on a roadway to accommodate pedestrians and bicyclists. While there can be more than four travel lanes before treatment, road diets are often conversions of four-lane, undivided roads into three lanes—two through lanes plus a center turn lane (see figure 104 and figure 105). The fourth lane may be converted to a bicycle lane, sidewalk, and/or on-street parking. Thus, the existing cross-section is reallocated. A recent HSIS report documented an empirical Bayes analysis of road diet installations in Iowa, California, and Washington.(115) Researchers estimated the change in total crashes resulting from the conversions in each of the two databases and combined these estimates into a crash modification factor (CMF). The empirical Bayes evaluation of total crash frequency indicated a statistically significant effect of the road diet treatment in both datasets and when the results were combined. Table 201 shows the results from each of the two studies and the combined results—the CMFs and their standard deviations.

Figure 104. Photo. Four-lane configuration before road diet. A four-lane undivided roadway with on-street parking on both sides of the street, from a driver’s perspective on the centerline of the street.

Source: Pedestrian & Bicycle Information Center and FHWA.

Figure 104. Photo. Four-lane configuration before road diet.(115,116)


Figure 105. Photo. Three-lane configuration after road diet. The same street as shown in figure91, but it has a three-lane cross section with one travel lane in each direction and a two-way left-turn lane in the center. Bicycle lanes and marked spaces for on-street parking are also provided on each side of the street.

Source: Pedestrian & Bicycle Information Center and FHWA.

Figure 105. Photo. Three-lane configuration after road diet.(115,116)


Table 201. Results of EB analysis on four-lane to three-lane road diets.(115)

State/site characteristics
Accident type
Number of treated sites (roadway length)
CMF (standard deviation)
Iowa: Predominantly U.S. and State routes in small urban areas (average population of 17,000)
Total crashes
15 (15 mi)
0.53 (0.02)
California/Washington: Predominantly corridors in suburban areas surrounding larger cities (average population of 269,000)
Total crashes
30 (25 mi)
0.81 (0.03)
All sites
Total crashes
45 (40 mi)
0.71 (0.02)


Tobey et al. investigated the safety effects of sidewalks.(117) The researchers found that sites with no sidewalks or pathways were the most hazardous for pedestrians, with pedestrian hazard scores of +2.6. These scores indicate that crashes at sites without sidewalks are more than twice as likely to occur as expected. Sites with sidewalks on one side of the road had pedestrian hazard scores of +1.2, compared with scores of -1.2 for sites with sidewalks on both sides of the road. Thus, according to Tobey et al., sites with no sidewalks were the most hazardous to pedestrians, and sites where sidewalks were present on both sides of the road were least hazardous.

Sidewalks separated from the roadway are the preferred accommodation for pedestrians.(118) Providing walkways for pedestrians dramatically increases how well pedestrians perceive their needs are being met along roadways. The wider the separation is between the pedestrian and the roadway, the more comfortable the pedestrian facility. One recent study indicated that roadways without sidewalks are more than twice as likely to have pedestrian crashes as sites with sidewalks on both sides of the street.(118,119) By providing sidewalks on both sides of the street, numerous midblock crossing crashes can be eliminated.


Zigzag lines are applied at midblock pedestrian crossings to restrict parking, stopping, and overtaking to improve pedestrian conspicuity. They are used in New Zealand, Canada, Europe, Trinidad, Great Britain, South Africa, Hong Kong, and Australia. A 2010 paper reviewed the literature to discuss how different countries use and interpret the meaning of zigzag pavement markings (lines) used at midblock crossings.(120) The review indicated that zigzag lines at pedestrian zebra crossings are misunderstood by most Trinidadian and Australian drivers as well as some researchers in North America. Such misunderstanding is associated with frequent vehicles parking, stopping, and overtaking in the vicinity of the pedestrian crossing. More education and public information on the crossing features and its use is needed.

The Virginia Department of Transportation studied the effectiveness of zig-zag pavement markings (shown in figure 106) in Loudoun County where the Washington and Old Dominion Trail crosses Belmont Ridge Road and Sterling Boulevard.(121) Effectiveness was defined in three ways: 1) an increase in motorist awareness in advance of the crossing locations, 2) a positive change in motorist attitudes, and 3) motorist understanding of the markings. Motorist awareness was measured by computing the difference in vehicle speeds before and after the installation of the markings. Attitudinal changes were measured through a survey targeting motorists, pedestrians, and bicyclists familiar with the markings. Motorist understanding was measured through another survey administered elsewhere in the State that targeted motorists unfamiliar with the zig-zag markings in Loudoun County.

The study found that the zig-zag markings installed in advance of the two crossings heightened the awareness of approaching motorists. This was evidenced by reduced mean vehicle speeds within the marking zones; speed reductions were largely sustained at observations 6 and 12months after installation, compared with 1 week after installation. Further, the majority of survey respondents indicated an increase in awareness, a change in their driving behavior, and a higher tendency to yield than before. The study also found that motorists had limited understanding of the purpose of the markings. When seen with context, motorists’ correct interpretations of their meaning increased, but not to levels compatible with guidance set forth in the MUTCD.(2) Researchers concluded that public information and education campaigns would help to increase understanding of the zig-zag pavement markings further.

Figure 106. Diagram. Schematic of zig-zag pavement marking design. A plan view of a section of a two-lane street, approximately 36 ft in length and 24 ft in width. The northbound lane, shown on the bottom half of the diagram, contains a zig-zag pavement marking the entire 36-ft length. The marking is shown as 6 inches wide and is located in the middle 4 ft of the travel lane. The marking is directed from the driver’s left to the driver’s right for 12 ft, then reverses direction for the next 12 ft, and then repeats the initial section for 12 ft.

Source: Dougald, Dittberner, and Sripathi/Transportation Research Record 2299.

Figure 106. Diagram. Schematic of zig-zag pavement marking design.(121)


A 2009 paper reported on the effectiveness of engineering countermeasures toward crash reductions at eight corridors in Miami-Dade, FL.(121) A before-and-after study was used to compare the sequential implementation of a 3-year large-scale NHTSA project. The project focused primarily on education and enforcement components followed by a large-scale FHWA engineering countermeasure project that was added to the NHTSA project along specific corridors. Results showed that the NHTSA pedestrian safety project reduced countywide pedestrian crash rates by 13 percent along the targeted corridors, and the FHWA engineering safety project produced a further reduction to 50 percent of the baseline level. These results translate to 50 fewer pedestrian crashes annually along the treated corridors. Countermeasures implemented included the following:

  • Reduced minimum green time at midblock crosswalks controlled by a traffic signal.
  • Advance yield markings at crosswalks with an uncontrolled approach.
  • Recessed or offset stop lines for intersections with traffic signals.
  • Leading pedestrian intervals.
  • Pedestrian pushbuttons that confirm having been pressed.
  • “Turning Vehicles Yield to Pedestrians” symbol signs for drivers.
  • Elimination of permissive left turns at a signalized intersection.
  • In-street pedestrian signs.
  • Pedestrian zone signs.
  • Midblock traffic signal.
  • Intelligent transportation system (ITS) video pedestrian detection.
  • RRFB for uncontrolled multilane crosswalks.
  • ITS smart lighting at crosswalks with nighttime crashes.
  • ITS “No Right Turn on Red” signs.
  • Pedestrian countdown timers.
  • Speed trailer.

In 2005, the Chicago Department of Transportation reported on the effects of a combination of traffic control devices and calming measures used to slow traffic and improve safety around schools.(123) These measures included the following:

  • Installation of speed humps along local street frontages of schools.

  • Variable speed indicator signs giving interactive speed indication to motorists passing by schools on arterial streets.

  • Installation of traditional school crossing warning signs and school zone 20 mi/h speed limit signs.

  • Experimental use of strong yellow/green pavement marking materials to mark crosswalks, “school” legends, speed humps, center lines, and stop bars in the blocks adjacent to schools.

The following summary was provided(123):

The analysis conducted was limited by the absence of control locations where similar marking treatments might have been installed using standard white pavement marking colors for crosswalks, “SCHOOL” legends, stop bars, and speed hump markings. The program analysis also generally was limited to assessing the combined effect of yellow/green markings, improved signing, and speed humps (on local streets), rather than analyzing the effect of individual traffic control measures. Understandably, it was the City’s intent to maximize the impact on motorists to increase their awareness, slow traffic, and improve overall safety in the school zones, rather than simply conduct a limited experiment on alternating color pattern crosswalks using a combination of white and strong yellow/green pavement marking materials.

The usefulness of the crash analysis was somewhat limited by only having one year of After-condition data available for the 2002 Program installation locations. No After-condition analysis was possible for the 2003 Program locations, nor, obviously, for the 2004 Program schools.

The results of the analysis suggest that the use of strong yellow/green pavement markings did not seem to have a significant effect on traffic speeds or crash experience. On arterial streets, the change in aggregate mean speeds, the aggregate percentage of traffic exceeding the speed limit, and the mode and median values of peak hour 85th percentile speeds was minimal. The use of speed indicators, which have proven effective in reducing speeds in other locations throughout the country, did not have a large effect on either speeds or crashes during school peak hours. The combined use of speed indicators and strong yellow/green markings also did not have a major impact on reducing speeds or crashes.

On local streets, the locations studied all had a combination of speed humps and strong yellow/green pavement markings. Most of these locations already had all-way stop control at adjoining intersections, thus already limiting the speeds on those streets. While the change in aggregate mean speeds and the aggregate percentage of traffic exceeding the speed limit was minimal, there did appear to be a reduction in the mode and median values of peak hour 85th percentile speeds. However, it seems reasonable to conclude that this reduction may have been largely attributable to the installation of speed humps rather than the yellow/green markings or upgraded school zone signing. This conclusion was reflected by the perception of survey respondents on the relative effectiveness of speed humps versus yellow/green markings. (pp. 9–10)

The city of Los Angeles, CA, has developed what it refers to as a “Smart Pedestrian Warning” system that includes the following multiple pedestrian crossing treatments(67):

  • Advance pavement messages (“PED XING”).
  • Advance warning pedestrian signs.
  • Extended red curb.
  • Double posting of intersection pedestrian signs.
  • Ladder-style crosswalk markings.
  • Automated pedestrian detection (video imaging).
  • Actuated alternating flashing overhead amber beacons.

This pedestrian crossing design and its various elements have evolved over the past several years based on experimentation and testing. To date, about 25 pedestrian crossing warning systems have been installed in Los Angeles. Fisher, in an undated paper, reports on informal evaluations by city engineering staff that indicate that this pedestrian warning system has improved motorist yielding to pedestrians from 20 to 30 percent to the 72- to 76-percent range.(67) Their evaluation also indicates that, of the 24 to 28 percent of motorists who did not yield, at least they traveled more slowly when approaching the enhanced crossings. For example, limited data indicate that 85th percentile vehicle speeds were reduced from 2 to 12 mi/h.

A study by Chen, Chen, and Ewing sought to evaluate the relative effectiveness of five countermeasures in New York City—increasing the total cycle length, Barnes Dance, split-phase timing, signal installation, and high-visibility crosswalk—and examine potential trade-offs in their effectiveness in reducing pedestrian crashes and multiple-vehicle crashes.(124) They adopted a rigorous two-stage design that first identified a comparison group of intersections, corresponding to each treatment group, and then estimated a negative binomial model with the Generalized Estimating Equation method to further control confounding factors and within-subject correlation; the model also accounted for built environment characteristics. Researchers concluded that the four signal-related countermeasures were more effective in reducing crashes than high-visibility crosswalks, but they added that there are trade-offs between improving pedestrian safety and motorist safety. Treatments that indirectly resolve conflicts (e.g., increasing total cycle length and Barnes Dance) were more effective in reducing pedestrian crashes and yet less effective in reducing vehicle crashes than those that directly separate conflicts (i.e., split phase and signal installation). In the case of Barnes Dance, there was a potential increase in vehicle crashes. This finding suggests that selection of a specific countermeasure at a location highly depends on the characteristics of the location and the problem at hand. Researchers suggested that the types of conflicts and balance of time for different groups of road users at the intersections should be considered so that the improvement of the safety of one group does not compromise that of other groups.


Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000
Turner-Fairbank Highway Research Center | 6300 Georgetown Pike | McLean, VA | 22101