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


Efforts during the initial phase of this project included a comprehensive literature review of pedestrian treatments being used at unsignalized pedestrian crossings. The appendix to the report contains the literature review. Certain parts of the literature review were updated or additional literature reviews were conducted as needed to support work done in later tasks of this project. The following sections contain those updated materials.


The RRFB flashes in an eye-catching sequence to draw drivers’ attention to the sign and the need to yield to a waiting pedestrian. It is located on the side of the road below pedestrian crosswalk or school crossing signs or overhead with a sign, and can be activated by a pedestrian either actively (pushing a button) or passively (detected by sensors).

Original FHWA Study on RRFB

An FHWA study evaluated RRFBs at 22 sites in St. Petersburg, FL; Washington, DC; and Mundelein, IL.(3) The RRFBs produced an increase in yielding behavior at all locations. During the baseline period before the introduction of the RRFB, yielding for individual sites ranged between 0 and 26percent. The average yielding for all sites was 4 percent before installation of the RRFBs. Within 7 to 30 days following installation of an RRFB, the average yielding increased to 78percent from the baseline condition, a statistically significant increase. Similar yielding values were observed during the remainder of the study period.

Data collected over a 2-year period, at 18 of the sites, confirmed that the RRFBs continue to be effective in encouraging drivers to yield to pedestrians, even over the longer term. By the end of the 2-year follow-up period, the researchers determined that the introduction of the RRFB was associated with yielding that ranged between 72and 96percent.

During the baseline measurement phase, the researchers installed advance yield markings to reduce the risk of multiple-threat crashes, which occur when a driver stopping to let a pedestrian cross is too close to the crosswalk, masking the pedestrian from drivers in the adjacent lane. The advance yield markings, which are recommended with an RRFB installation, were typically placed 30 ft in advance of the crosswalk unless a driveway or other issue was present, in which case they could be placed up to 50 ft in advance of the crosswalk. The posted speed limit at the sites ranged from 30 to 40 mi/h.

The observers scored the percentage of drivers yielding and not yielding to pedestrians. Researchers scored drivers as yielding if they stopped or slowed and allowed the pedestrian to cross. Conversely, researchers scored drivers as not yielding if they passed in front of the pedestrian but would have been able to stop when the pedestrian arrived at the crosswalk.

2009 FHWA Study

A 2009 FHWA report presented the results of an evaluation of RRFBs at two sites in Miami, FL.(4) The study team used the following measures of effectiveness (MOE) to assess the effect of the RRFB on pedestrian and driver behavior: 1) the percentage of pedestrians trapped in the roadway, 2) the percentage of drivers yielding to pedestrians, and 3)the percentage of pedestrian-vehicle conflicts. The researchers found statistically significant improvements in all of the MOEs as shown in table 1. For percent drivers yielding, only 4.2 or 4.1 percent of the drivers yielded at the two sites in the before condition. After the installation of the RRFB, yielding at the two sites increased to either 55.2-percent driver yielding or 60.1-percent driver yielding, depending on the site. The researchers concluded that the RRFB offered clear safety benefits, and it was placed into the category of highly effective countermeasures.

Table 1. Measures of effectiveness for RRFBs with pedestrian crossing signs, Miami, FL (compiled from reference 4).

Measure of effectiveness

Percent drivers yielding (staged crossings, daytime)

NW 67th and Main Street
4.2 (n=2,330)
55.2 (n=2,131)
0.01 (daytime and nighttime combined at this site)
S. Bayshore and Darwin
4.1 (n=2,075)
60.1 (n=1,361)
0.01 (daytime and nighttime combined at this site)

Percent drivers yielding (staged crossings, nighttime)

NW 67th and Main Street
4.4 (n=703)
69.8 (n=223)
0.01 (daytime and nighttime combined at this site)
S. Bayshore and Darwin
2.5 (n=139)
66.0 (n=225)
0.01 (daytime and nighttime combined at this site)

Percent drivers yielding (resident crossings)

NW 67th and Main Street
12.5 (n=137)
73.7 (n=259)
S. Bayshore and Darwin
5.4 (n=200)
83.4 (n=111)
Percent of pedestrians trapped in roadway
NW 67th and Main Street
< 0.01

Percent of vehicle-pedestrian conflicts

NW 67th and Main Street
< 0.05
S. Bayshore and Darwin
< 0.01

2009 Florida Study

A 2009 report summarized the effects of installing a pedestrian-activated RRFB at the location of one uncontrolled trail crossing at a busy (15,000 average daily traffic (ADT)), four-lane urban street in St. Petersburg, FL.(5) The researchers used a mounted video camera to collect pre- and post-treatment data about pedestrian and driver interactions at the trail crossing. An analysis of the data showed a statistically significant increase in driver yielding (from 2percent pretreatment to 35 percent post-treatment and 54 percent when the beacon was activated) and ability of pedestrians to cross the entire intersection (from 82 percent pretreatment to 94 percent post-treatment).

2011 Texas Study

A 2011 before-and-after study looked at the effectiveness of RRFBs at an uncontrolled crossing in Garland, TX.(6) The school crosswalk on a five-lane arterial had continental crosswalk markings, supplemented by school crossing signs on either side of the roadway. Before installation, city engineers had observed driver compliance with the crosswalk was poor and planned to install overhead and side-mounted RRFBs to improve compliance and facilitate pedestrian crossing maneuvers. In this study, researchers observed drivers’ yielding behavior for crossing-guard-controlled crossings and staged pedestrian crossings, both before and after installation of the RRFBs (see figure 1). Researchers found that while yielding to school-related crossings with a crossing guard remained fairly constant (with yielding rates about 90 percent), drivers’ responses to staged crossings in non‑school-zone time periods improved from 1 percent before the installation to 80 percent after installation.

Figure 1. Photo. School crosswalk with RRFBs in Garland, Texas. A pedestrian crosswalk in a school zone. The crosswalk is supplemented by rectangular rapid flashing beacons, one set of post-mounted beacons on each side of the road, and one set above the center of the road on a mast arm.

Source: Texas A&M Transportation Institute.

Figure 1. Photo. School crosswalk with RRFBs in Garland, TX.(6)

2011 Oregon Study

A 2011 Oregon Department of Transportation report evaluated RRFB installation at three crosswalks in Bend, OR.(7) For two of the locations, the highway has a 45-mi/h posted speed limit and is a four-lane roadway with a center median, bike lanes, and sidewalks. Because the posted speed limit of 45 mi/h was greater than most locations where RRFBs have been installed in Oregon, the plans for the RRFB installations included additional features to increase the visibility of the crosswalks and the pedestrians and bicyclists using them. These include RRFB assemblies at three locations: on the side of the road, on the median at the crosswalk, and 500 ft in advance of the crosswalk. Pavement markings included ladder bars with a continental crosswalk, a stop line 50ft in advance of the crosswalk, and double white solid no-lane-change lines, as well as the legend “PED X-ING” on the road as vehicles approach the intersection. The signs in the RRFB assembly were 48 inches, and there was a sign in advance of the crosswalk with the legend “Stop Here for Pedestrians.” Before the installation of the RRFBs, motorist yield rates were 23and 25 percent at the 45-mi/h intersections and 6 percent at the third crossing. These rates increased to between 74 and 83 percent following treatment. The researchers concluded that “RRFBs should be considered for installation on high-speed facilities where there are posted speeds greater than 35 mi/h if there are pedestrians and bicyclists using the facility and a history of crashes or the potential for them.”(7)

2013 California Study

A study of two sites in Santa Monica, CA, compared the effect of an RRFB and a circular rapid-flashing beacon (CRFB) on yielding behavior at two crossings.(8) The RRFB was installed at one site and the CRFB at the other, and after an evaluation period, they were switched and evaluated again. The study evaluated driver yielding rates both when the beacons were activated and when they were not activated. At both sites, the beacon that was installed first showed better yielding rates than the beacon that was installed second. At site 1, the RRFB resulted in 85-percent yielding when activated and the subsequent CRFB showed 63 percent yielding, while at site 2, the CRFB was installed first and produced 92-percent yielding compared to 80-percent yielding for the RRFB. In all cases, driver yielding rates were higher (between 7 and 22 percentage points) when the beacon was activated than when it was not.

2013 Canada Study

A 2013 pilot project in Calgary, AB, Canada, assessed motorists yielding behavior before and after installation of RRFBs.(9) Overall, the installation of the RRFB improved yielding compliance from 83 percent to 98 percent, which was statistically significant.

2014 Michigan Study

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 the presence of a sharp curve, the posted speed limit was 25mi/h, and the site had 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 80-percent 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 RRFB 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 RRFB.

2014 Texas Study

A Texas Department of Transportation (TxDOT) study examined the effectiveness of the following traffic control devices used at pedestrian crossings: traffic control signals, pedestrian hybrid beacon, and RRFBs.(11,12) The 22 RRFB sites had School Crossing signs with the RRFB. While there are some RRFB sites in Texas with Pedestrian Crossing signs, all sites used in the Texas study had School Crossing signs. The FHWA Interim Approval for the RRFB states that when used, two Pedestrian Crossing or School Crossing signs shall be installed at the crosswalk, one on the right-hand side of the roadway and one on the left-hand sign of the roadway.(1) On a divided highway, the left-hand side assembly should be installed on the median, if practical, rather than on the far left side of the highway. A later interpretation indicated that overhead mounting is appropriate, and that if overhead mounting is used, a minimum of only one such sign per approach is required and it should be located over the approximate center of the lanes of the approach.(13) In Garland, some of the sites had the School Crossing signs located over the roadway on a mast arm along with the roadside installation (see figure 2). The overhead placements were on undivided roadways such as those with four through lanes and a two-way left-turn lane (TWLTL) or multilane one-way roads. The side mounts were used on divided roadways when the second sign could be placed in the median. The City of Garland was concerned that the RRFB would be outside the driver’s cone of vision or it could easily be obscured by a truck going in the opposite direction when located on the left side of an undivided roadway. The medians on the divided roadways allow a left-side installation next to traffic going in that direction. When the median was less than 4 ft wide, the city used an overhead installation.

Figure 2. Photo. Study site from TxDOT study showing overhead RRFB installation. A pedestrian crosswalk that is supplemented by rectangular rapid flashing beacons. One set of post-mounted beacons is on each side of the road, and one set is above the center of the road on a mast arm.

Source: Fitzpatrick et al.

Figure 2. Photo. Study site from TxDOT study showing overhead RRFB installation.(11)

Figure 3 shows the plot of driver yielding versus posted speed limit, and figure 4 shows the plot of driver yielding to total crossing distance for the data from the Texas study.(11) These plots provide an overview of the findings and relationships with the use of average site yielding values. The data for individual crossings were used in the analysis. The modeling results for the RRFBs are shown in table 2. For RRFBs, posted speed limit, total crossing distance, one-way versus two-way traffic, and city were all significant. RRFB sites with higher posted speed limits were associated with higher driver yielding values (see figure 3). As shown in figure 3, the two Waco sites with a 30-mi/h posted speed limit have very low driver yielding (below 40 percent). These sites had low traffic volumes during data collection, which resulted in several crossings having no vehicles yielding. Even when these two sites are removed from the model, the trend of higher driver yielding for higher speed was still present and statistically significant. A closer review of the data reveals that while driver yielding is higher for the 40-mi/h sites compared with the 35-mi/h sites, the overall difference is very small (only 1 percentage point between the two averages). Therefore, while there may be a statistically significant increase in driver yielding by speed limit, the difference is not of practical significance.

Figure 3. Graph. RRFB: driver yielding to posted speed limit plot from 2014 Texas study. Percent of drivers yielding to crossing pedestrians at each study site is on the y-axis, and the posted speed limit in mi/h is on the x-axis. Different symbols are used for each of the three cities represented in the study. The chart shows that the two Waco sites with 30-mi/h speed limits had between 30- and 40-percent yielding. The 13 sites in Garland with speed limits of 35, 40, or 45mi/h had yielding rates between 80 and 100 percent, and the single Frisco site, posted at 45mi/h, showed a yielding rate of about 78 percent.

Source: Fitzpatrick et al.

Figure 3. Graph. RRFB: driver yielding to posted speed limit plot from 2014 Texas study.(12)


Figure 4. Graph. RRFB: driver yielding to total crossing distance plot from 2014 Texas study. Percent of drivers yielding to crossing pedestrians at each study site is on the y-axis, and the total crossing distance in feet is on the x-axis. Different symbols are used for each of the three cities represented in the study. The chart shows that except for the two Waco sites with yielding rates between 30 and 40 percent, the yielding rate declines from nearly 100 percent at a crossing distance of less than 50 ft to a rate of about 78 percent at a crossing distance of approximately 120 ft.

Source: Fitzpatrick et al.

Figure 4. Graph. RRFB: driver yielding to total crossing distance plot from 2014 Texas study.(12)

The data revealed a trend of lower driver yielding rates for wider crossing distances compared with shorter crossing distances (see figure 4). Perhaps drivers believe that the greater distance between their vehicles and the pedestrian presents the opportunity to not stop for the waiting pedestrian. For example, a driver on a six-lane road has multiple lanes in which to adjust position, perhaps feeling that leaving a full traffic lane between the car and the crossing pedestrian is sufficient.

The model shown in table 2 uses Frisco as the base city and provides odds ratios for the two other cities. The driver yielding rate for Waco is lower compared with Frisco (not statistically significant), while driver yielding is higher for Garland (statistically significant). The greater number of the devices in Garland may contribute to drivers being more familiar with the treatment, which could be contributing to the better driver yielding behavior.

Table 2. RRFB total driver yielding model results.(12)

Standard error
z value
Pr(>|z|) b
Reference Levelc
M.O_T: two-way
City: Garland
City: Waco
  • aColumn headings are defined as follows:
    • Coefficients = variables included in model.
      • M.PSL = posted speed limit on major roadway.
      • M.O_T = one-way or two-way operations on major roadway.
      • Total_CD = total crossing distance.
      • City = city where RRFB is located (Frisco, Garland, or Waco).
    • Estimate: natural logarithm of the ratio: Odds (coefficient level)/Odds (reference level).
    • Standard error: standard error of estimate.
    • z-value: Standard normal score for Estimate, given the hypothesis that the actual odds ratio equals one.
    • p-value: Probability that the observed log-odds ratio is at least as extreme as Estimate, given the hypothesis that the actual odds ratio equals one.
  • bSignificance values are as follows: p < 0.10; * p < 0.05; ** p < 0.01; and *** p < 0.001.
  • cReference Level in the model has the following conditions:
    • Categorical variables base value: City = Frisco and M.O_T = one-way.
    • Continuous variables range: M.PSL = 30 to 45 mi/h and Total_CD = 38 to 120 ft.


Another area of investigation for this FHWA research became the detection distance to objects located beyond flashing beacons.

Previous studies have investigated detection distance to objects under various conditions. In the 1990s, a National Cooperative Highway Research Program (NCHRP) study investigated stopping sight distance issues, including examining driver capabilities in detecting objects in the roadway.(14) Using the closed course at Texas A&M University, the researchers had 20 drivers detect 7 objects during nighttime conditions. Drivers indicated when they could detect and recognize the objects. With low-beam headlamp illumination, the researchers found that the rear of a vehicle was detected at a range of 725 and 1,000 ft and then recognized between 550 and 725 ft. For high-beam illumination, the recognition distances started at about 1,100 ft, and the detection distances extended to almost 1,800 ft. For their pedestrian, who was a mannequin dressed in dark clothing, recognition under low-beam headlamp illumination was about 100 ft, and the detection distance was about 225 ft. Under high-beam headlamp illumination, the recognition distance was about 300 ft, and the detection distance reached a maxium of almost 500ft.

A 2012 TxDOT study also used the Texas A&M University closed course to investigate nighttime detection of various objects.(15) The objective of the study was to investigate whether very bright traffic signs in rural conditions limited sight distance beyond the sign. The study included observations using both low-beam and high-beam headlamp illumination. It also included detection tasks without a sign present and with signs made of different retroreflective materials. The objects included a small gray wooden plaque, a pedestrian in blue medical scrubs, and the rear of a parked car. Each of the objects was placed outside the travel lane within 3.2 ft of the right edge line pavement marking. For the data without signs, the detection distance to the pedestrian was about 380 ft with low beams and 550 ft with high beams.

Comparing these two studies, the vehicle detection results look quite similar, whereas the pedestrian detection distances from the TxDOT study are slightly longer than those found in NCHRP Report 400. There are two likely causes. One is that the pedestrian used for the NCHRP Report 400 work was described as wearing dark clothing, while blue medical scrubs, which are not as dark, were used in the TxDOT 2012 study. Another reason could be the evolution of headlamp technologies. The NCHRP Report 400 work was conducted with a vehicle with sealed beam headlamps, while the TxDOT work completed with modern-day tungsten-halogen headlamps. Also confounding the results is the fact that the participants in the 2012 TxDOT study were drivers while in the NCHRP 400 study they were observers in the front seat.

A study done in 2011 looked at the impact of color contrast in the detection and recognition of objects in a road environment.(16) The investigation compared the nighttime object detection distance to several objects under three lighting systems: two LED systems with differing color temperatures and a fluorescent system. The objects included blue-clothed and black-clothed pedestrians. The results showed that the LED lighting types are significant in terms of the average distance at which a driver can identify a pedestrian and provide a longer detection distance than the fluorescent lighting. The mean detection distance was about 475 ft with the fluorescent system and about 600 to 675 ft for the LED systems.


There is extensive research into how various characteristics of road signs such as sign size, letter size, contrast, luminance, conspicuity, and others affect sign legibility distance. (See, for example, references 17, 18, 19, and 20.) Legibility distance is the location upstream of a sign where a driver can correctly read all of the words on a sign or correctly identify the symbol on a symbol sign. Examples of recent research on road sign symbol recognition are summarized in the following paragraphs.

Paniati conducted a laboratory experiment to determine the relative legibility distance and driver comprehension of 22 symbol warning signs that were in use in the United States in the 1980s.(21) The results showed that legibility distance decreases with participant age and that bold symbols of simple design provide the greatest legibility distance for all age groups. Data were collected using a zoom lens on a slide projector. The participants were presented with a randomly selected slide beginning at a simulated distance of 1,000 ft and moving equivalent to a driving speed of 32 mi/h. The participant pressed a button when he or she could identify the symbol on the projected sign. This study included 16 participants under 45 years old and 16participants over 55years old. Paniati noted that, as expected, there was a significant difference between the legibility distances for many of the symbols. The results indicated signs with color cues (e.g., signal, stop, yield ahead) or of simple design (e.g., crossroad, right turn) provided the greatest legibility distance. As increased complexity is added to the symbol (e.g., added lane, winding road, reverse curve) the legibility decreases. As the details of the image to be resolved become finer (e.g., slippery when wet, narrow bridge, pavement ends) or the long-distance appearance of the images begin to resemble one another (e.g., pedestrian, worker, school crossing), the legibility distance continues to decrease.

A study conducted in the early 2000s had participants walk toward a calibrated, fixed-size sign projected on a large projection screen.(22) Two levels of performance were assessed: maximum recognition distance (threshold) and the distance at which the symbol types could be recognized with ease (confident). A total of 40 subjects, half of whom were young and half older, participated in the study. The traffic sign background luminance, luminance contrast, and symbol type were found to be statistically significant in affecting the symbol recognition distance. Observer age and background complexity were statistically nonsignificant.

Zwahlen and Schnell conducted an exploratory daytime and nighttime sign recognition and legibility field driving experiment involving 11 signs and 10 subjects.(23) The instructions emphasized that the subjects were to say aloud the information on the traffic sign at the point during their approach when they could clearly (with near 100-percent certainty) identify all visual details of the message in the symbol. The average daytime legibility and recognition distances were about 1.8 times longer than the average nighttime legibility and recognition distances. One of the signs tested was the Curve Arrow (black paint on yellow engineer grade background) which had an average legibility/recognition distance of 1,045 ft in the daytime and 743 ft at night.

Conspicuity is the capacity of a sign to stand out or be distinguishable from its surroundings and thus be readily discovered.(24) For a sign to be conspicuous, the viewer must be able to differentiate it from the surrounding background. Variables affecting conspicuity include luminance, luminance contrast, and color contrast. The addition of beacons to a roadway sign can improve the conspicuity of a sign. Literature that specifically addresses the legibility distance for symbol signs when used with supplemental beacons was not identified. Therefore, this FHWA study provides an opportunity to gain insights into the situation when beacons are used to supplement a roadside sign. Another unique aspect of this FHWA study is that participants were driving the vehicle while searching for the signs.



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