Road User Understanding of Bicycle Signal Faces on Traffic Signals (2020)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

CHAPTER 2 Background and Literature Review Relevant literature was identified by searching TRID (Transport Research International Documentation) for the appropriate keywords. This literature was supplemented by reviewing the list of references identified for the upcoming update to the chapter on bicycle signals in the AASHTO Guide for the Development of Bicycle Facilities led by project team member Toole Design Group. The research team also contacted the MUTCD office to get access to any Request to Experiment (RTE) and final evaluation reports related to bicycle symbols in signal faces. In addition, the team searched the grey literature for published evaluations, blog posts, and popular press stories using standard Google searches. Basic Human Factors Concepts The MUTCD provides a variety of guidance and support associated with the principles of traffic control devices (TCDs). Explicitly, the MUTCD (FHWA, 2009) states that “The proper use of traffic control devices should provide the reasonable and prudent road user with the information necessary to efficiently and lawfully use the streets, highways, pedestrian facilities and bikeways.” This report places particular emphasis on the bicyclist as the road user of concern operating on streets and bikeways that cross through signalized intersections. The following subsections provide content on the visibility, comprehension, compliance and human error, and evaluation methods for traffic control devices focused on traffic signals. Visibility Traffic signals must have an acceptable legibility distance for the intended road user. Legibility distance is defined as the distance from which the road user can detect the message conveyed by the traffic signals. The distance must be sufficient for the road user to comprehend the message and initiate the correct response to classify the traffic signals. Traffic signals should be placed in a conspicuous location with a clear line of sight to the road user but also in a way that is consistent with road user expectancy (Borowsky et al., 2008a). Many studies of visibility have dealt with characterizing the role of top-down and bottom-up attentional processes in controlling human attention under various circumstances. Wickens et al. (2001) suggested the salience, effort, expectancy, and value (SEEV) model for describing the human selective attention allocation. Wickens’s model is based on the general principle of the two attentional-perceptual processes (i.e., top-down and bottom-up). Following Wickens’s model, SEEV are the factors that explain how people allocate their selective visual attention. Salience, or capturing the properties of events, and effort, or the movement of attention across longer distances, are the bottom-up components of the model and expectancy, or the likelihood of seeing an event at a particular location, and value, or the importance and relevance of tasks served by the attended event, are the top-down components of the model (Wickens et al., 2001). This model, and its specific components (e.g., expectations) were frequently investigated in studies of driving 7 

behavior (Borowsky et al., 2008b; Horrey et al., 2006; Langham et al., 2002; Richard and Lichty, 2013; Werneke and Vollrath, 2012; Wickens and McCarley, 2008). The first step in creating a visible traffic signal relies on it being easily and rapidly detected by the road user. The signal should be positioned in such a way that would make it easy for the road user to detect and understand it. To optimize placement, both top-down and bottom-up attentional- perceptual processes should be supported. Target detection, visual search, and attention allocation are driven by both top-down and bottom-up processes. These processes are based on stimulus and sensory input, and are influenced by the user’s experience and knowledge-based, contextual, and mental schemes. To ensure that the traffic signals are visible to the road user, designers should consider the size, design, and placement of the device. A road user’s cone of vision can be defined as excellent from three to five degrees. At 10 degrees, road users have a clear vision where texture, shape, size, color, shading, and other visibility parameters can be distinguished easily. At 20 degrees, road users maintain satisfactory vision where regulatory and warning traffic control devices can be well perceived. At 70 to 90 degrees defines the cone of peripheral vision, where road users primarily see movement (Schieber et al., 2009). Expectation is an important factor that predicts where drivers will focus their attention while searching for valuable information on the road, such as oncoming traffic or traffic signals. A road user’s expectation can be derived from a short-term situational context or from mental schemas that are based on long-term knowledge and experience. As an example, in short-term, situational- context when a task-relevant events’ stream (e.g., stream of traveling vehicles from a specific direction) is higher in a particular place, the likelihood (i.e., expectation) of seeing a relevant event (e.g., an arriving vehicle) at that location will increase, eventually resulting in higher attention to that location (Werneke and Vollrath, 2012). The detection of traffic signals will often rely more on expectations that are derived from long-term knowledge and experience than on contextual events. Thus, placement of traffic signals faces in a way that is not consistent with drivers’ mental schemes can decrease the possibility of drivers’ correct and timely identification of traffic signals significantly. Research by Borowsky et al. found that incorrect placement of traffic signals can decrease the chance of correct identification by approximately 50% and extend the total fixation time (the total duration of time spent looking at a specific location) needed on the traffic signals until correct identification by several hundreds of milliseconds (Borowsky et al., 2008a, 2008b). Many of these concepts are embedded in the MUTCD which describes the requirements for placement of the primary vehicular signal faces. As defined by the MUTCD (2009) on signalized intersection approaches with 85th-percentile speeds of less than 45 mph, the minimum distance of signal faces of any diameter from the stop line is 40 feet. However, the maximum distance from the stop line to an 8-inch signal face is 120 feet, and the maximum distance from the stop line to a 12-inch signal face is 180 feet. Figure 2 summarizes this information. This information has not been explored for validity or for variations when adapted for persons on a bicycle. 8 

Source: MUTCD Figure 2. MUTCD Figure 4D-4 on Lateral Placement and Visibility of Primary Traffic Signal Faces Comprehension After the road user detects the traffic signals, the user must recognize, identify, and comprehend its meaning. With correct comprehension, the road user can initiate a correct action by complying with the directive (e.g., stop in response to a circular red indication). The comprehension of traffic signals is critical for road user safety. Thus, designing traffic signals that are easily recognized and understood is crucial, especially in the case of new, unfamiliar, or uncommon signals. To significantly increase the probability of correct recognition and comprehension, signal design should follow the ergonomic principals for display design (e.g., compatibility, familiarity, and standardization) (Sanders and McCormick, 1993; Ben-Bassat and Shinar, 2006; Shinar et al., 2003). Comprehension of traffic signals by road users is a critical factor in compliance and, ultimately, in the device operating correctly. Even though the meaning and state of traffic signal indications 9 

can be easily interpreted, they can still be misunderstood for the same reasons as traffic signs, especially uncommon or newly introduced traffic signal indications. For traffic control devices, the comprehension of traffic signs can be more challenging than the comprehension of traffic signals and is attributed to the nontrivial meaning and greater variance of messages that are communicated through traffic signs (Dissanayake and Lu, 2001). In a cross- cultural traffic signs comparison study, Shinar et al. (2003) suggested that traffic signs that follow good ergonomic design principals are more likely to be fully comprehended than signs that violate these guidelines, which was validated in a later in-depth study (Ben-Bassat and Shinar, 2006). Ben-Bassat et al. (2006) examined relevant ergonomic design principles (Sanders and McCormick, 1993) for the purpose of increasing sign comprehension rates. This study identified physical and conceptual compatibility, standardization, and familiarity as relevant principals that, when applied correctly, can increase the comprehension of a traffic sign by road users. In accordance with those principals, a traffic sign should be consistent with what it represents to facilitate the mental associations of road users (e.g., traffic signal picture represents the presence of nearby traffic signal). A traffic sign should also follow the same norms (i.e., colors, symbols, shapes, sizes, etc.) used in existing similar traffic symbols (e.g., in a traffic signal the color red should communicate “stop” and green “go”). Compliance and Human Error Once the road user recognizes the traffic signals and understands their meaning, the user is required to comply with the directive information in a timely manner. For that reason, the traffic signal should encourage the desired behavior from the road user by design. Non-compliance can stem from two reasons: 1) intended violation or 2) human error (e.g., not seeing the traffic signal, misreading, misunderstanding, or mistaking it with another traffic signal. Errors can be defined as occasions where the user’s intended performance was acceptable, but it fell short (such as intending to drive at or below the speed limit, but accidentally pressing the accelerator pedal too far (a slip), forgetting the speed limit (a lapse), or thinking that the speed limit is 70 mph when it is actually 60 mph (a mistake)). In contrast, intentional violations may be defined as occasions where the driver intended to perform the action, such as deliberately exceeding the speed limit. Driver error has been identified as a direct cause of at least two-thirds of the crashes, according to some estimations from the U.S. ( Hankey et al., 1999; Wierwille et al., 2002). To improve TCD compliance, engineers need to understand the underlying psychological mechanisms that lead drivers to make an error. The analysis of a human error in general, and in driving more specifically, rely on taxonomies and theories of psychological mechanisms. Various classification methods and theories have been proposed to describe human error (e.g., Norman, 1981; Rassmussen, 1986; Reason, 1990; Wickens and Hollands, 2000). Norman relates his classification of human error to a scheme-based human behavior theory. The errors that Norman describes are a result of an unintended action (e.g., mode error - the wrong scheme gets executed due to misperception of the situation), similar to a slip type error, which is an error that occurs while trying to execute a predetermined approach to achieve an objective (Reason, 1990). Reason presented a more elaborate human error taxonomy with four possible categories: slip, lapse, mistake, and violation. Wickens's (1992) ties the human errors with each of the basic stages of the human information processing model, which is shown in Figure 3: perception stage (i.e., assessment of the situation), cognitive stage (i.e., a plan for action is created), and last stage of action execution. 10 

Source: Wickens, C.D., Hollands, J. Banbury, S. Parasuraman, Engineering Psychology and Human Performance, Fourth Edition, Routledge, Taylor and Francis Group, 2013. Used by permission. Figure 3. Human Information Processing Model. In this model, mistakes are a consequence of failures in the first two stages (understanding of the situation and/or planning of the action), lapses are a result of a poor cognition process (specifically, failure in retrieval from memory) and slips are a result of failures in the execution of the action. Reason (1990) also suggested that errors can be attributed to each of the three levels of the model of cognitive control. Thus, errors can be a result of failures in actions that are skill, rule, and/or knowledge-based. Tasks that the human is very skilled with, as a result of vast experience, will be executed almost automatically (i.e., no need for thought), and thus, failures would often be a result of bad execution of good intention (i.e., slips and lapses). Less common tasks will require more cognitive effort of the human, either to recall a preferred and known response (i.e., rule- based) or, in the less common case, to plan a course of action based on individual knowledge; in this case the failure can be a consequence of an incorrect assessment of the situation or bad planning (i.e., mistake). Evaluation Methods for Traffic Control Devices Methodological approaches to TCD evaluation can take many forms, including surveys, laboratory testing, driving simulators, test tracks, and in-field observations (Figure 4). Each method has inherent advantages and limitations. Generally, as ones moves from left to right along the continuum, the realism of the setting is improved. However, with each incremental improvement in realism potentially uncontrolled and confounding variables are introduced into the evaluation. As one moves from right to left, additional experimental control is improved, helping to isolate the effects of interest (Chrysler et al., 2011). Ultimately, robust human factors research leverages triangulation amongst different experimental mediums to validate research findings and increase the transferability of research findings into practice. 11 

Source: Chrysler et al., 2011. Used by permission. Figure 4. Types of Human Factors Evaluation for Traffic Control Devices Visibility and Comprehension of Bicycle Signal Face No published research studies were found that have directly addressed the visibility of the bicycle signal face. Visibility includes placement for optimal detection by road users, conspicuity of the lens, and detection distances. There are two separate issues related to the comprehension of the bicycle symbol in the signal face: 1) recognizing that the symbol face denotes the signal as exclusive for bicycles, and 2) knowing which movements are allowed by the indications given by the bicycle signal. No published research studies were found that have directly addressed comprehension of the bicycle symbol in the signal face, either for bicyclist or drivers. The use of the bicycle symbol in signs, pavement markings, and signal faces, however, is a widespread and international practice. In a review of signs and signals for cyclists and pedestrians in 13 countries (Austria, Belgium, Denmark, France, Germany, Italy, Norway, Poland, Russian Federation, Spain, Switzerland, United Kingdom and the U.S.) for the United Nations, Hiron et al. (2014) found that nearly all symbols feature a similar version of the bicycle (although sometimes a person is shown riding the bicycle). The study notes that most of the countries reviewed also have three-section faces with bicycle symbols in the lens. Figure 5 shows a variety of bicycle signal faces in international use. All of the symbols are very similar, though the faces from the Utrecht, Netherlands, and Shanghai, China, include an arrow in the bicycle symbol face. 12 

Beijing, China Lima, Peru Credit: D. Hurwitz, Oregon State University, used by permission Credit: A. Clarke, Toole Design Group, used by permission Shanghai, China Utrecht, Netherlands Credit: D. Hurwitz, Oregon State University, used by permission Credit: A. Clarke, Toole Design Group, used by permission Vancouver, B.C. Canada London, United Kingdom Credit: C. Monsere, Portland State University, used by Credit: S. Kothuri, Portland State University, used by permission permission Figure 5. Examples of International Bicycle Signal Faces 13 

Published Evaluation Reports While no published research studies were found regarding comprehension on the use of the bicycle signal indications, several published reports include brief assessments of visibility and comprehension of the bicycle signal face. As mentioned in the introduction, the city of Davis, CA, is believed to have installed the first bicycle signal in the U.S. in 1994. A published evaluation report describing the evaluation of the Davis signal was prepared for the California Traffic Control Devices Committee by Pelz et al. (1996). The bicycle signal heads, consisting of red-yellow-green 12-inch circular displays with the bicycle symbol in the face were installed at the 3-leg intersection of Russell Boulevard and Sycamore Lane near the University of California, Davis campus. The geometry of the intersection and the location of the bicycle signal heads are shown in Figure 6. The south leg of the intersection is a multiuse path and there are no northbound vehicles. Modifications to the signal phasing provided for the exclusive north and south movement of bicycle traffic. For the southbound left- and right-turn vehicle movements and bicycle movements, both the vehicular and bicycle signal faces were visible to each road user. The evaluation included a before-after survey of users and review of crash and citation data. In the after survey, a question was asked whether the respondent thinks that “seeing the round red signal with the green bicycle signal is confusing to drivers?” A total of 191 persons responded to this question and 33% (n=64) indicated “Yes.” The crash and citation data revealed no issues. In the opinion of the authors, placing the bicycle signal in locations visible to motorists resulted in a clear understanding of the bicycle signal by motor vehicle users. The evaluation did note a learning curve for drivers (early in the evaluation period some drivers would go during the green bike phase). In conclusion, the study noted that over the long-term there were no issues and that “once the signal has become operational the signal is easy to understand by both cyclist and motorists.” Of official Requests to Experiment (RTE) with bicycle signal faces conducted before IA-16 was issued that are listed on the MUTCD website, only the final report from the City and County of Denver was available (Denver, 2009). The experiment evaluated signal compliance at an intersection with a shared path. The evaluation consisted of three phases – pre-installation or baseline condition, post-installation, and post-removal of the bicycle traffic signal head. In the pre- installation phase, the data was collected with the presence of a conventional pedestrian signal and no bicycle signal. In the post-installation phase, data were collected after the installation of a bicycle signal. In the post-removal phase, data were collected after the removal of the bicycle signal and with the presence of a pedestrian countdown timer. Data was collected pre-installation, one week, one month and two months after the bicycle signal was installed and removed to examine changes in behavior and signal compliance. A total of 8,619 observations over 59 hours were made during the three phases. On-site observations were employed to study bicyclist behavior. The pre-and post-installation analysis revealed that bicyclists on the trail were more likely to enter the intersection during a compliant portion of the traffic signal cycle when the bicycle signal was present and the capacity of the signal to accommodate the compliant bicycles increased. No negative effects on bicyclist behavior were found due to the presence of the bicycle signal head, and pedestrians were less likely to non-comply when the bicycle signal was operating. The report writers concluded that the bicycle signal did not lead to pedestrian confusion. One conflict between motor vehicles and trail users was observed in 59 hours, therefore leading the study to conclude that the bicycle traffic signal did not lead to driver confusion. During the post- removal phase, bicyclists were observed to be more likely to enter the intersection during a non- compliant phase and the capacity of the traffic signal to accommodate the compliant bicyclists was reduced. Statistical analysis of the 8,619 observations revealed little to no change in crossing behavior for bicyclists on the trail when comparing the data from all three phases. 14 

Source: Pelz, 1996 Figure 6. Signal Location and Phasing, Russell and Sycamore Lane, Davis, CA A brief report on the installation of a bicycle signal in San Francisco at the intersection of Masonic and Fell is published on the NACTO case studies website (NACTO, n.d.). The installation, in 2008, was the first bicycle signal installed by San Francisco Municipal Transportation Agency. The signal separates left-turning vehicles from bicyclists in a left-side bicycle lane and pedestrians, the majority of whom are entering a park. The existing infrastructure at the intersection required the vehicle and bicycle signals to be placed on the same mast arm. Non-compliance of left-turning vehicles with the red turn arrow was a problem and required some phasing modifications and louvers. After these modifications, operations and compliance by motor vehicles improved. Blog and News Posts While blog and news posts are not peer-reviewed research, they do provide some anecdotal observations of potential issues. Recent installations of bicycle traffic signals in Seattle, WA, Brooklyn, NY, and Chicago, IL have drawn blog posts and news stories about driver confusion with bicycle signal displays. In Brooklyn, signal heads at Third Street and Prospect Park West are used to control vehicle right turns on to a one-way street and bicycle traffic’s connection to a park and a left-side two-way separated bike lane. The six-section signal head has vehicle and bicycle signal faces mounted adjacent to each other (Figure 7). The bicycle symbol face is the only differentiating element of the bicycle signal, as an accompanying “Bike Signal” sign is not present. A brief news story and accompanying video show drivers turning right when the vehicle signal is red and the bicycle signal is green (Mixson, 2018). It is unclear whether the non-compliance is 15 

related to confusion with the signals or intentional non-compliance, as right-turn-on-red is not allowed in New York City. Bicycle Signal Face Source: Google Streetview, 2019 Figure 7. Traffic Signal at Third Street and Prospect Park West (Brooklyn, NY) In Seattle, installation of a two-way separated bicycle lane on 2nd Avenue in downtown required the use of bicycle signals to safely separate the contra-flow bicycle traffic from left-turning vehicles. The initial installation had all traffic signal faces post-mounted on the left side. A news story about the project discussed driver confusion with the design (McNichols, 2014). A subsequent project included upgraded signal infrastructure with mast arms that allowed the separation of the signal heads. A blog post on “Seattle Bikes” notes improved driver understanding of the signal displays (Fucoloro, 2016). Another blog post by Michael Andersen of the Green Lane project describes the driver non- compliance of no-right-on-red and proceeding through the bicycle green at the intersections along a two-way protected bike lane on the Broadway corridor in Seattle (Anderson, 2014). The corridor also includes a streetcar and two-way vehicle traffic. Anderson hypothesizes that driver confusion may be related to right-turning drivers seeing the bicycle signal face and assuming it was for right- turning traffic (i.e., not detecting or comprehending the bicycle symbol within the signal lens). A blog post from a graphic design firm in Chicago describes the confusion of drivers with bicycle signals, especially with drivers illegally turning during a bicycle green following recent installations of bicycle signals. The author proposes three potential solutions to mitigate the confusion – simplifying the number of lines in the bicycle symbol in the lens for improved visibility, introducing a bike-familiar abstract shape (chevron) in the lens, or using words (BIKE) instead of the bicycle symbols in the lens (Gunderson, 2017). Compliance There is more literature on cyclist compliance at signalized intersections, though most of the studies document compliance at general traffic signals. Compliance, however, is not always a proxy for comprehension and varies significantly in the studies reviewed. 16 

Bicycle-Specific Traffic Signals In general, most of the studies about compliance at bicycle traffic signals suggest a link to intended non-compliance rather than poor comprehension of bicycle traffic signals. Monsere et al. (2013) investigated cyclist compliance at signalized intersections equipped with and without bicycle signals in Oregon. Two types of cyclist compliance were evaluated, those that moved straight through the intersection violating the red signal or those that made an illegal right turn. Overall, there was high compliance and no difference between behaviors at bicycle signals and general traffic signals, suggesting good comprehension of the bicycle symbol in the signal face. As part of an evaluation of new bicycling facilities in Washington, D.C., Goodno et al. (2013) studied compliance at locations with bicycle-specific signals. They found compliance, which ranged from 80% to below 20% at some intersections, was strongly related to crossing traffic and somewhat related to delay or progression for cyclists (i.e., low cross traffic and delays contributed to non-compliance). Monsere et al. (2014) studied user behavior at signalized intersections as part of a larger project studying intersections in Portland, OR, San Francisco, CA, Chicago, IL and Washington D.C. Figure 8 summarizes the compliance, which ranged from 67% to 98%, of bicyclists at all of the intersections where video data collection was conducted. At the L Street locations in Washington, D.C., cyclists were using the Leading Pedestrian Interval to obtain an early start (now allowed by ordinance). At the three intersections studied in Chicago on Dearborn Avenue, road user compliance with the signals was nearly identical. A range of 77-93% of observed bicyclists complied with the bicycle signal, which compared to about 84-92% of observed motorists who complied with the left-turn signal separating their movement from the two-way bicycle traffic at the same intersections. Clifford et al. (2018) studied the impacts of new infrastructure innovations for cyclists – “Hold the Left” and “Early Release” at signalized intersections in London, UK. Video imaging was used to observe behavior and surveys were utilized to determine user perceptions of these treatments. The “Hold the Left” treatment is implemented on a cycle track or bike lane, which is equipped with a bicycle signal for cyclists. This treatment separates the vehicular left-turning movements while cyclists and through vehicles are allowed to proceed through to minimize conflicts — the compliance rates at four intersections varied between 78% and 92%. “Early Release” is the same as a leading bicycle interval and is implemented with bicycle signals. The Early Release treatment was tested at three intersections, where bicycles were provided with four seconds early release. The proportion of cyclists who were able to take advantage of the lead interval ranged from 81% to 97%. Additionally, the behavior of drivers who also took advantage of the cyclists lead interval were also observed. When a cyclist was present during the lead interval, the proportion of vehicles that took advantage of the early release ranged from 0% to 7%. When a cyclist was not present, the proportion ranged from 0% to 4%. These represent instances of motorists taking cues from the bicycle signals. 17 

Waited for green/legal right-turn on red Proceeded illegally on red Fell/ Baker 80% 20% Oak/Broderick (Leading Bicycle Signal) 80% 20% Oak/Divisadero (Traffic Signal) 98% 2% Multnomah/7th (Traffic Signal) 96% 4% Multnomah/11th (Traffic Signal) 96% 4% Multnomah/9th (Traffic Signal) 92% 8% L Street/ Connecticut(Traffic Signal w/ LPI) 67% 33% L Street/ 15th Street (Traffic Signal w/ LPI) 79% 21% Milwaukee/ Desplaines (Traffic Signal) 76% 24% Milwaukee/ Elston (Bicycle Signal) 84% 16% Dearborn/ Randolph (Bicycle Signal) 92% 8% Dearborn/ Madison (Bicycle Signal) 77% 23% Dearborn/ Congress (Bicycle Signal) 93% 7% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Source: Monsere et al. 2014 Figure 8. Observed Cyclist Compliance with Traffic Signals Greenshields et al. studied the impact of “Cycle Gates” used in the United Kingdom, which included the provision of separate stop line and bicycle signals for cyclists and a separate stop line for vehicles to prevent left-hook collisions (Greenshields et al., 2018). The bicycle signals allow the cyclists to enter a reservoir area ahead of the other traffic and wait at the stop line. When the cyclists in the reservoir area are presented with a green-signal indication, other cyclists behind the cycle gate are shown a red indication preventing their entry into the reservoir area. The cyclists in the reservoir area are allowed to proceed on the green indication, which is given a few seconds before the vehicular green indication, thus allowing cyclists to clear the conflict area before the left-turning vehicles start their maneuver (see Figure 9 for an annotated description). The usage of the cycle lane and gate was 97% and 61.5%, respectively, at the two intersections as compared to the general traffic lane. Twenty-two percent of the cyclists using the cycle lane and 6.8% of the cyclists using the general lane were non-compliant at the red signal near the first stop line if the downstream reservoir signal was green. Similarly, the non-compliance rate at the other location was 8.8% for cyclists using the cycle lane and 38.7% for cyclists using the general lane. The overall non-compliance rate (cyclists disregarded red signals at both stop lines) was 1.7% and 6.1%, respectively. 18 

Source: Greenshields, 2018 Figure 9. Annotated Image of a Cycle Gate General Traffic Signals Many studies have examined factors affecting cyclist compliance at general traffic signals using observational studies and online surveys (Johnson et al., 2011; Johnson et al., 2013, Mirabella and Zhang, 2014; Pai and Jou, 2014; Richardson and Caulfield, 2015; Casello et al., 2017). Their findings revealed that direction of travel, presence of other road users, gender, age, helmet use, previous crash experience, detection reliability and presence of pedestrian crossings all had an effect on cyclist compliance. Other factors that impacted positively impacted cyclist compliance included implementation of signal timing features such as rest in walk and pedestrian recall, 19 

presence of bike boxes, two-phase left turns, turning lanes with advanced green phases, and arrival on green, while the presence of T-intersections and intersections with short red-phase duration negatively impacted compliance. The impact of type of bike on compliance was mixed with one study finding higher non-compliance rates among e-bike users (Pai and Jou, 2014), while another study did not find a statistically significant difference in compliance rates (Langford et al., 2015). Another set of studies explored cyclist behavior at traffic signals equipped with a blue-light feedback confirmation device (Boudart et al., 2015; Boudart et al., 2017) using observational studies and postcard intercept surveys. Their findings revealed that while differences in comprehension of the blue-light confirmation device were observed, the device, however, did not have a statistically significant impact on compliance. A modified pavement marking was also tested and it also did not have a statistically significant impact on compliance. Another study explored compliance of cyclists at signalized intersections with the modified 9C-7 pavement marking using observational data, and results showed high compliance rates with traffic signals, excluding the right-turn-on-red (Smith et al., 2018). Safety at Intersections with Bicycle Signals While there is deep literature on bicycle crash frequency and severity, few studies have examined the impacts of traffic control on bicycle-motor vehicle crashes. Rahimi et al. (2013) evaluated five design elements for left-hook crashes which included mixed traffic with left-turning motorists, left turns in the intersection for the motorists, bicycle signals, advance stop lines for bicyclists, bike boxes using video observation, and surveys along a route in Japan using 10 bicyclists and four drivers. Their results revealed a higher preference for bicycle signals based on comfort and safety. Wahi et al. (2018) examined bicycle-motor vehicle crashes in Queensland, Australia, between 2002 and 2014 at uncontrolled, stop control locations and signalized intersections. At signalized intersections, age, roadway characteristics (dip, the presence of driveways) and bicyclist behavior (movements that led them to be at fault during a crash) increased injury severity, while helmet use decreased severity at signalized intersections. Recently, the New York City DOT conducted a safety evaluation of bicycle-specific intersection treatments to provide guidance on the appropriate treatment (NYCDOT, 2018). Mixing zones, fully split phases (with bicycle signals), delayed turn (split LBI) and offset crossing (protected intersections) were evaluated in the study using crash, conflict and comfort analysis. Of these treatments, fully split phases, delayed turn and offset crossing used bicycle-specific traffic signals. While mixing zone and offset crossing are design treatments, delayed turn and fully split phase are signal timing treatments. With the delayed turn, bicyclists are provided a head start similar to a leading pedestrian interval, while turning movements are held before they are allowed to proceed concurrently with the bicyclists. In a fully split-phase treatment, the through bicyclists and turning vehicles are separated in time with bicycle signals. The study did not document any driver confusion with bicycle traffic signals. Kothuri et al. (2018) also studied the safety impacts of split LBI and mixing-zone treatments using an observational study with conflict analysis. With the split LBI treatment, while the conflicts were eliminated at the start of green, conflicts persisted during the start of the flashing yellow interval and continued through the stale green. Some user confusion (related to the merging behavior and where each entity needed to position themselves) was observed regarding the position of the bicyclists and drivers within the mixing zone. Qualitative guidance was also provided regarding the optimal treatment to use given a set of bicycle and turning vehicular volumes. 20 

Related Traffic Control Devices This section reviews research for comprehension, visibility, and compliance related to traffic control devices for vehicles, transit and pedestrian control with a focus on both research methods and issues that are similar to bicycle signal faces. Vehicular Traffic Signals Lens Size and Backplates for Traffic Signals Conspicuity of traffic signals has been cited as a factor in intersection collisions and improving their visibility can improve safety. Cole and Brown (1968) found that signal visibility was insensitive to lens size and depended only on intensity. They determined that greater visibility could be achieved by using a higher intensity lens. Other studies have found that the use of a larger signal lens improved visibility (Hulscher, 1975) and the use of backplates or backboards reduced the intensity required by 25-40% at distances of 300 feet (FHWA, 2000; Hulscher, 1975). King (1981) found that signal visibility during the day was affected by signal lens size and intensity, but not at night. Sayed et al. (2005) evaluated the safety impacts of improved signal conspicuity, which resulted from the addition of yellow micro-prismatic retroreflective sheeting along the outer edge of the signal at 17 intersections. Protected/Permissive Displays for Turns A number of studies have explored drivers’ comprehension of flashing yellow arrow (FYA) signal display indications for left turns (Asante and Williams, 1993; Bonneson and McCoy, 1993; Noyce and Kacir, 2001, 2002; Drakopoulos and Lyles, 2001; Brehmer et al., 2003; Noyce and Smith, 2003; Knodler et al., 2005, 2006a, 2006b, 2007; Henery and Geyer, 2008; Schlattler et al., 2013; Hurwitz et al., 2013; Marnell et al., 2013; Hurwitz et al., 2014). These studies have either utilized static surveys and/or observed behavior in the driving simulator to determine comprehension rates. The surveys were typically computer-based and were either administered independently or as a follow-up after the drivers completed the experiment in the driving simulator. They consisted of static images of intersections with various signal display alternatives and the responses were usually presented as multiple-choice options. The experiments in the driving simulator usually involved subjects driving in a grid and being presented with various signal display alternatives and their actions were recorded. The results of these studies demonstrated that simultaneous displays (green arrow and green ball, green arrow and red ball) were associated with lower driver comprehension rates than single indications alone (Noyce and Kacir, 2002). The results also showed that the FYA signal display indication for left turns was well understood by drivers and led to FYA being adopted for permissive left-turn indications. Boot et al. (2015) evaluated a new flashing pedestrian indicator (FPI) that alternated between a yellow arrow and a pedestrian symbol using online surveys. Drivers generally understood the meaning of FPI and it was associated with significantly more yielding to pedestrians; however, confusion was observed among drivers proceeding through the intersection. Though included in the MUTCD, there is minimal research on driver comprehension of the use of FYA for right turns. Recent studies have used web-based surveys, microsimulation models and driving simulator study to determine drivers’ comprehension on the use of FYA for right turns (Hurwitz et al., 2018; Ryan et al., 2018; Jashami et al., 2019). Results revealed FYA indication improves driver comprehension and behavioral responses to the permissive right-turn condition. Drivers were also 21 

observed to approach the intersection at slower speeds when they encountered a FYA than a steady circular green indication in the absence of a pedestrian. Supplemental Signs for Traffic Signal Faces When a traffic signal face is intended to control a specific movement or vehicle type, a supplemental sign is often used for additional clarification. Figure 10 shows the supplemental signage available in the MUTCD for signals, turn prohibition and lane control. In general, many of the studies show increased comprehension with the addition of a supplemental sign. Most of the studies evaluating the comprehension of signal indications with supplemental signs studied either protected permitted left turns or right turns (PPLT or PPRT) (Bonneson and McCoy, 1993; Drakopoulos and Lyles, 2001; Henery and Geyer, 2008; Schlattler et al., 2013; Hurwitz et al., 2018). These studies used surveys to understand driver comprehension of the traffic control devices with and without supplemental signs. Results revealed that the supplemental signs were beneficial in specific situations (e.g., R10-12 during the permitted phase) (Drakopoulos and Lyles, 2000) and increased driver comprehension (Schlattler et al., 2013; Hurwitz et al. 2018). One study also revealed higher comprehension rates for the R10-12 sign (94%) than the FYA indication (72.4%) (Henery and Geyer, 2008). However, in some of these studies, the comprehension measures were biased since the supplemental sign contained the desired response to the signal indication. R10-12 R10-10 R10-11 R10-11a R10-10b R3-2L R3-5R Source: MUTCD and Schlattler et al. 2013 Figure 10. Supplemental Signs: Signals, Turn Prohibition, and Lane Control Transit Signals Light Rail Transit Signals Similar to bicycles, there is often a need to separate the movements of light rail vehicles from other traffic at signalized intersections. Before the adoption of the guidance in the current MUTCD, a TCRP report reviewed 10 early LRT systems in North America and found no uniformity in signal displays across the systems (Korve et al., 1996). While some systems used 22 

standard traffic signals on a shared right-of-way, others used a monochrome bar, monochrome “T,” colored “T,” or colored “X” LRT signals. When these signals were installed in the motorist's line of sight they led to driver confusion, especially at night. As part of the research, guidance on size, shape, aspect and placement of LRT signals to avoid motorist confusion was developed. The report suggests that in locations where the LRT signals could cause motorist confusion, they should be positioned and shielded in a way that they are visible only to LRT operators. The TCRP report stated that the LRT signals should use a 12-inch lens; however, an 8-inch lens may be used in urban areas where space is tight. The recommended shape was rectangular or square with a dark color (black is preferred) and a visor for each lens. A monochrome bar was the recommended display indication, and a PROCEED indication for the train included a vertical lunar white bar placed near the bottom of the signal head. The STOP indication should consist of a horizontal lunar white bar placed near the top of the signal head. Between the PROCEED and STOP indications, a flashing white triangle should be used to indicate when the LRT should PREPARE TO STOP. The report also stated the primary signal be located on the near side of the intersection and they should be separated vertically and/or horizontally by at least 8 feet from the nearest traffic signal head or the pedestrian signal head for the same approach (Korve et al. 1996). The LRT signals should also be installed within the cone of vision of the LRT operators, which is 25 degrees on each side of the center track line for a total of 50 degrees. Bus Queue Jump Signals Bus queue jump lanes are used to reduce transit delay and increase reliability and combine short dedicated transit facilities with either a leading bus interval or active transit signal priority to prioritize transit (NACTO, 2016). To facilitate queue jumps, buses need to have access to a lane and move to the head of the queue at the beginning of the signal cycle (NACTO, 2016). In the typical design, a bus uses a shared right-turn lane with an adjacent near-side bus stop. When the bus is first in the queue, the right-turn signal is displayed while the other through traffic is shown a red indication. In both the NACTO’s Transit Street Design Guide and the TCRP Report 118 Bus Rapid Transit Guide (Kittelson et al., 2007), the authors suggest the possibility of motorist confusion, but no quantitative evidence is presented. In practice, louvered or visibility-limited green indications are used, which is only visible to the right-most lane and often accompanied by a sign indicating the signal face is for right turns “except buses.” No other published studies on the topic were identified. Pedestrian Signals Pedestrian signal indications are comprised of a steady walking person symbolizing the WALK indication, a flashing upraised hand symbolizing the pedestrian clearance interval (FLASHING DON’T WALK (FDW)) and the steady upraised hand symbolizing the DON’T WALK indication. During the WALK indication, the pedestrians are permitted to start crossing. During the pedestrian clearance interval, pedestrians are not supposed to start crossing, but those that are already in the crosswalk are expected to complete their crossing. During the steady DON’T WALK, pedestrians in most jurisdictions are not supposed to enter the roadway. Research has shown the FDW is poorly understood, with comprehension levels ranging from 31% to 50% (Mahach et al., 2002; Chicago DOT, 2002). Other research has also shown that pedestrians were more likely to start crossing during FDW (which is illegal in many states), run out of time while crossing, return to the starting location, or get caught in the middle of the crosswalk when the indication changes to solid DON’T WALK (Huang and Zegeer, 2000). 23 

Pedestrian Countdown Timers Countdown timers are clocks that display the remaining time for a signal indication, thus providing users with real-time information to make better decisions. In the U.S., they are most commonly seen for pedestrian operations. The pedestrian countdown signals were first approved and included in the 2003 MUTCD (FHWA, 2003). These countdown signals show the amount of time remaining in the clearance interval (FDW). A number of studies have reported a reduction in pedestrian-motor vehicle conflicts and improved pedestrian safety as a result of the pedestrian countdown timer installation (Huang and Zegeer, 2000; Markowitz et al., 2006; Chen et al., 2015; Lambrianidou et al., 2013; Schmitz, 2011; Scott et al., 2012; Vasudevan et al., 2011; Eccles et al., 2004). The pedestrian countdown timers were also found to improve driver safety (Kwigizile et al., 2015; Kitali et al., 2018). Drivers also used the pedestrian countdown timers to make informed decisions when approaching the intersection (Chen et al., 2015; Schmitz, 2011; Elekwachi, 2010; Nambisan and Karkee, 2010). One study examined the legibility and comprehension of the countdown signals without the flashing hand using digital video displays (Van Houten et al., 2015). Results revealed that pedestrians were more likely to consider crossing if they judged they had enough time with countdown pedestrian signals alone than with countdown signals plus FDW and this effect held across gender and age. Pedestrian Hybrid Beacons Pedestrian hybrid beacon (PHB, HAWK) is a traffic control device used at a pedestrian crossing to control traffic on the major approach. The PHB consists of two red indications and one yellow indication. In its base state, the PHB rests in a dark mode. When a pedestrian activates a pushbutton indicating an intent to cross, the PHB displays a flashing yellow indication for the driver for a few seconds, followed by a steady yellow indication and steady red indication requiring drivers to stop. A WALK indication is displayed for the pedestrians followed by a clearance interval (FLASHING DON’T WALK). During the flashing pedestrian clearance interval, an alternating flashing red indication is displayed to the drivers. During the flashing red indication, drivers are allowed to proceed after stopping if the pedestrians have cleared half the roadway (Fitzpatrick and Pratt, 2016). A specific PHB concern relates to driver behavior when during the dark mode and understanding of the flash mode. The concern among some professionals is that drivers may believe that the PHB is not working when it is operating in a dark mode similar to a traffic signal during a power outage and may treat it as a stop sign (Fitzpatrick and Pratt, 2016). However, two studies did not find any evidence of this behavior (Nassi and Barton, 2008; Fitzpatrick and Pratt, 2016). Some studies have demonstrated improved pedestrian safety and driver yielding in response to PHB installation (Turner et al., 2006; Nassi and Burton, 2008; Fitzpatrick and Park, 2009; Godavarthy and Russell, 2010; Hunter-Zaworski and Mueller, 2012; Fitzpatrick et al., 2013, 2014; Lincoln and Tremblay, 2014; Brewer et al., 2015; Pulugurtha and Self, 2015; Fitzpatrick and Pratt, 2016). Summary This chapter reviewed literature pertaining to basic human factors concepts of visibility, comprehension, and compliance, which are all critical characteristics associated with proper use of traffic control devices. Road users first need to see the traffic control device, correctly comprehend its meaning and respond accordingly. A review of bicycle-focused literature showed no published research studies that directly addressed visibility and comprehension of the bicycle signal face or the transferability of design 24 

assumptions from motor vehicle users. This reveals gaps in research that need to be further examined. Varying anecdotal reports have documented driver confusion with bicycle signals due to lack of separation between vehicular and bicycle traffic signal faces, and not recognizing that the bicycle signal indications were exclusive for bicycles. However, none of the more formal published evaluation reports found evidence of significant user confusion. Research on safety impacts at intersections with and without bicycle signals is also very limited. Safety of cyclists at intersections with and without bicycle signals may also warrant further research. Review of literature pertaining to traffic control devices for vehicles revealed the importance of visibility and conspicuity of traffic signals in reducing collisions. Studies also revealed extensive testing of various signal display alternatives to determine optimal displays and signal head configurations to communicate protected/permissive movements. These studies highlight the importance of substantial research into effective ways to communicate allowable movements for vehicles. In contrast, little to no research was found on tests for comprehension and best ways to communicate allowable movements by the bicycle traffic signal. Similarly, many studies have explored pedestrian comprehension and safety impacts for drivers and pedestrians with pedestrian countdown timers. Other studies have explored potential driver confusion when the pedestrian hybrid beacon is operating in a dark mode and found no evidence. Studies evaluating light rail transit signals found evidence of driver confusion due to non-standardization and use of different types of signals. This led to guidance on size, shape, aspect and placement of transit signals to avoid motorist confusion. In contrast, no research studies were found which explored bicyclist and driver comprehension and compliance based on size, placement, and orientation of bicycle signal faces. Research is therefore needed to understand this gap. 25