NOTE: See also Teaching Blind Pedestrians to Cross at Complex Signalized Intersections

Crossing at Modern Signals
by Dona Sauerburger, COMSģ
Fall 2005 Newsletter, AER Orientation and Mobility Division

"Houston, we have a problem!"

Changes in intersection design have been insidious, and O&M specialists who are alert and teach blind people to travel in the suburbs as well as in cities are aware that sometimes our consumers, correctly using all the techniques we had been taught (see "traditional techniques") can find themselves in life-threatening situations.

The prevalence of some of the problems and their consequences was documented in a recent study in which 16 independent blind travelers in each of 3 cities (Charlotte, Portland, and Cambridge -- a total of 54 subjects) were observed crossing intersections with modern features such as protected turns, split phases and lead pedestrian intervals (Barlow, Bentzen and Bond, 2005).

These are all features for which our traditional street-crossing techniques are NOT sufficient, and are potentially dangerous. As a result, in one out of 20 crossings, the researchers had to intervene as the subjects started to cross because they were in immediate danger, often because they were crossing at the wrong time, such as when left-turning traffic which had the right of way was fast approaching. At several of the crossings in Charlotte and Portland, pedestrians were required to push a button in order to get a walk signal and sufficient time to cross, but pedestrians pushed the button in only 16.3% of the crossings in Portland and in none of the crossings in Charlotte. As a result, in almost half of those crossings (45.4%) the subjects were still in the street when the signal changed and the perpendicular traffic started! Often the subjects were unaware that anything was wrong or that they were at risk.

I can remember the first time I was aware that our traditional techniques can expose our consumers to life-threatening situations. It was when I sent a sweet elderly lady into the street when the traffic first started on the parallel street to her left, even though the drivers had a green arrow that gave them the right of way. I figured that since our reliable, traditional techniques had worked so well over the past 5 decades for people who canít see the signals, blind people could safely continue to use them even though they canít see the walk signal and canít see that the drivers have a green arrow to cross their path.

So my client started to cross when the traffic on the parallel street to her left surged forward, as I had directed her. Even though she prominently used a white cane, the left-turning drivers honked and glared and almost ran her down.

I became determined to figure out how blind people can know when itís their turn to cross at these modern intersections.


The rest of this article will describe the modern features that make our traditional street-crossing strategies dangerous and ineffective, then explain one strategy which seems to work. I will then describe strategies that O&M specialists or advocates have developed which, if relied on, can lead blind people to cross at the wrong time (when they donít have the right of way). The well-intentioned instructor who teaches these misguided strategies might be considered liable for any injury or death that results from using them.

Modern features that render our traditional techniques potentially misleading:

Traffic patterns at modern signals may include the following features:

  • Split phases: traffic in one direction of a street is allowed to go straight-through, right or left while traffic in the opposing direction waits (to allow for drivers to turn left without opposing traffic).

  • Protected left-turn phases: left-turners proceed while opposing traffic is held with red light (often this is accompanied with giving the right of way with green arrows to right-turning traffic from the other street)
    NOTE: Split phases and protected left phases are used when there is heavy left-turning traffic. In these situations, the left-turning drivers have a green arrow which gives them the right of way over the pedestrians (if there is a pedestrian signal, it indicates "Donít Walk")

  • Lead Pedestrian Intervals (LPI): pedestrian Walk signal starts 3-10 seconds before vehicular green signal

  • Exclusive pedestrian phases: provides for pedestrians to cross with all vehicular traffic stopped (except right-turn-on-red may be permitted)
    NOTE: LPIís and exclusive pedestrian phases are often installed where there is heavy right-turning traffic, to avoid or reduce conflicts with pedestrians.

    Probably the most significant features of modern intersections which affect blind travelers are actuation, and the ability of the engineers to readily change the timing of signals. These features make it impossible to predict the pattern or timing of the signal phases.

  • Actuation: Actuated signals respond to the traffic present at the intersection, so that certain phases of the cycle can be shortened or even skipped altogether if there is little or no traffic waiting in those lanes.

  • Variable timing: The timing of the different phases of the traffic signal can be changed in a twinkling, done either in the computer box (controller) at the intersection, or at a central office miles from the intersection. For example if, one evening during rush hour, the engineers notice on their monitors that the traffic is starting to back up on one street, CLICK! The timing on that street can be extended for each cycle, and later changed back again. If traffic is steadily increasing or decreasing on certain legs of an intersection (for example because of a new business or residential development), the engineers may drive out and tweak the controller at the intersection to extend the minimum timing for that street. Some signals are fixed-time during the day and actuated at night, or set with one timing for certain times of the day and another timing at other times, such as rush hour vs. the evening. Emergency vehicles can get an "emergency pre-emption" which changes all the signals in their favor (including shortening the pedestrian phase!).

    The effect of actuation and being able to change signal timing easily is that it is impossible to predict the pattern or the timing of traffic signals.

    A solution that seems to work:

    Many of us have struggled to develop strategies for determining when to cross at complex intersections.

    One strategy that seems effective for avoiding crossing when the left-turners have the right of way (such as at signals with split phases and protected left phases) was explained by Frieswyk (2005) and Barlow et al (2005). This strategy is to cross with the nearest parallel traffic. That is, if crossing either to or from the NW and SW corners, start to cross with the southbound traffic, as shown in Figure 1.

    Drawing shows a 'plus-shaped' intersection with pedestrians trying to cross the east-west street from the top left and bottom left corners (NW and SW corners).  The large arrow representing the traffic they cross with is starting next to the upper left (NW) pedestrian and going down (south) to pass the lower left (SW) pedestrian.

    Figure 1: To avoid starting to cross when left-turners have the right of way, cross with the nearest parallel traffic. In this example, if crossing Main Street from the NW or SW corners, the nearest parallel traffic is the southbound traffic, as shown with the large arrow.

    The only exception Iíve seen for this rule is at an intersection where the traditional rule of crossing with parallel traffic also doesnít work, because during the entire time that the parallel traffic is moving, it has an arrow to turn right, and pedestrians get a ďdonít walk" signal and must wait. Pedestrians get their walk signal at another time.

    Although the exceptions to this strategy are rare, blind pedestrians should always analyze unfamiliar intersections and be prepared for unusual situations like this one.

    Dangerous strategies!

    The aforementioned strategy seems to work for split and protected left phases, but what about situations where no parallel traffic is moving when itís time to start crossing? For example, at lead pedestrian intervals and at exclusive pedestrian phases there is no parallel traffic movement permitted when the pedestrians should start crossing. And of course there are many situations where the surge of traffic on the parallel street, which our traditional strategy uses as a cue to indicate when itís time to cross, doesnít exist or cannot be heard. The only suggestion from Barlow et al (2005) is an accessible pedestrian signal (APS).

    Many well-meaning O&M specialists and advocates have come up with alternative strategies which seem to be effective but which, upon more careful analysis, have each turned out to be dangerous because they are potentially misleading. These strategies involve:
  • observing the 1) timing or 2) sequence of the signal in order to predict when the walk signal or green light will begin, and/or
  • starting to cross when the perpendicular traffic stops.

    However, these strategies are all unreliable because at actuated intersections itís not only impossible to predict the timing, itís also impossible to predict the pattern of traffic, because any of the phases can be skipped if there is no one waiting in those lanes. Even at fixed-time signals, the timing and/or sequence can be changed in an instant and when they make these changes, the engineers do not send out announcements nor take the responsibility to contact all blind travelers whose safety depends on this knowledge and who may assume that the sequence or timing of the phases is the same as it was when they were taught to rely on it for their safety and their lives.

    Using stopped perpendicular traffic as a cue is also potentially misleading and dangerous. Vehicles can stop for reasons other than a red light, such as to wait to turn left, or drop someone off, etc. Often, by the time a group of vehicles have stopped, itís too late to start the crossing. More importantly, there are traffic patterns which have some of the traffic on a street stop while other traffic on that street has the green light. Sometimes these traffic patterns will occur unexpectedly, and cannot be anticipated no matter how long one observes the pattern, as I will explain in my first example below.

    Examples of using misleading strategies or cues:

    What follows are some examples of situations which illustrate how the strategies of timing, predicting traffic patterns, and using stopped perpendicular traffic as a cue can backfire and lead people to start the crossing when they are at risk because they are crossing against the light.

    Faulty strategy #1:
    Predicting crossing time by observing the traffic pattern / signal cycle:

    One of the hairiest examples Iíve observed where relying on predicting the traffic pattern was potentially lethal was at an intersection where my client wanted to cross a 7-lane street (Piney Orchard Parkway) with a 4-lane street to his right (Waugh Chapel Road). Piney Orchard is a north-south street and he wanted to cross it from the SE corner to the SW corner. Although Piney Orchard is the main street, Waugh Chapel is also a major route for thousands of residents coming home from shopping or work. There is normally traffic on it at all times of the day.

    We observed the traffic over an extended period of time, and noted that the traffic pattern used the following sequence:

    1st: Piney Orchard: left-turning traffic (there was always southbound traffic waiting to turn left from Piney Orchard to go east);
    2nd: Piney Orchard: all traffic in both directions;
    3rd: Waugh Chapel: left-turning traffic (if there is any);
    4th: Waugh Chapel: all traffic in both directions.

    Because the first traffic movement from his parallel street (Waugh Chapel) might be the traffic beside him with a green arrow to turn left across his path, my client used the strategy of starting to cross with the traffic in the nearest parallel lanes -- that is, the Waugh Chapel traffic coming east from across Piney Orchard. This is the strategy that was explained above, and it is effective if done correctly.

    However, there was one major problem: he had to hear the surge of traffic from vehicles that were waiting on the other side of Piney Orchard -- more than 7 lanes away. At very wide intersections like this one, those vehicles (coming from the parallel street) often sound the same as vehicles turning left from the perpendicular street (see Figure 2).

    Drawing shows a plus-shaped intersection, with a pedestrian in the bottom left corner.  Arrows indicating the two sources of traffic that sound the same come from the top of the drawing and pass beside the pedestrian, and come from the street to the right, turn left, and pass beside the pedestrian in the same place that the other traffic passed.

    Figure 2 [left]: As pedestrians stand on the corner (the dot shows a pedestrian standing on the bottom left corner), vehicles passing beside them often sound the same, regardless of whether they came from straight across the street or they turned left from the perpendicular street.

    So to distinguish whether the traffic was coming from across the street or turning left from Piney Orchard, we used our knowledge of the traffic pattern to predict when to expect his parallel traffic to start. Normally, it would start after all the traffic on Piney Orchard stopped, so when he heard the traffic on Piney Orchard stop, he would start to cross as soon as he heard traffic that sounded like it came from Waugh Chapel on the other side of Piney Orchard. Sometimes this was the first parallel movement.

    However, we were astounded to realize that sometimes all three lanes of traffic close in front of him stopped on the perpendicular street (Piney Orchard) and then he heard what sounded like traffic coming from his parallel street, Waugh Chapel, but it wasn't! It was traffic turning left from Piney Orchard! He couldn't hear that the traffic on Piney Orchard going south was NOT stopped, and he didn't know that the traffic in front of him was about to get the green signal again, and so he started crossing seven lanes of traffic AGAINST A RED LIGHT!

    After we studied it, we realized what was happening. Whenever Piney Orchard traffic had been moving for its minimum timing (about 45 seconds) and no vehicles had approached from Waugh Chapel, if a southbound vehicle on Piney Orchard approached from the north to turn left into Waugh Chapel, the northbound traffic on Piney Orchard would be stopped for just a few seconds and allow the left-turner to proceed onto Waugh Chapel, then they would resume again (see Figure 3).

    Drawing shows Piney Orchard going left to right (north to south) and the pedestrian is standing on the bottom left (SE) corner.  Traffic coming from the right (north) on Piney Orchard is going straight to the left (south) and also turning left and passing next to the pedestrian.  Traffic in front of the pedestrian on Piney Orchard is stopped and waiting while the opposing traffic turns left.

    Figure 3 [left]: If southbound traffic wants to turn left when no one wants to cross Piney Orchard, northbound traffic is stopped to allow the left turn, as shown here in the drawing (my client is standing at the lower left / SE corner, indicated by the blob).

    So using the sequence of the traffic was unreliable for distinguishing which traffic he heard. We worked hard to have him learn to distinguish the sounds of left-turning perpendicular traffic from that of parallel traffic, and eventually he was able to do it.

    The problem with trying to predict traffic movement based on our understanding of the system is that we donít have engineering degrees and even if we did, we donít know how these engineers designed this signal. Traffic engineers have explained to us that there is no way that even they could predict what will happen at any given intersection based only observation, no matter how long the observation is. Using traffic patterns to determine or predict what's happening is unreliable and therefore dangerous.

    Faulty strategy #2:
    Predicting crossing time by observing the timing of the signal

    Some O&M specialists have resorted to timing the signal phases to predict when itís time to cross. In general, the strategy works like this. First, you observe that your walk signal or green light comes on at a certain time after another event, for example 17 seconds after the left-turning traffic on the parallel street starts, or 25 seconds after the perpendicular traffic starts. Then, when the observed event occurs, you start counting seconds until you assume your signal starts, and you begin crossing.

    In addition to requiring the ability to accurately measure the passage of time, this strategy requires that the timing of the cycle is reliably consistent and predictable. There are two problems with this strategy:

    1) no traffic signal cycle is reliably consistent and predictable.

    Of course the timing at actuated signals will vary with each cycle, depending on how many vehicles are in the various lanes. But even fixed-time signals can be programmed to vary at different times of the day or week to handle different traffic loads.

    And even if you had assurance from the traffic engineers that the signal is fixed-time and doesnít vary during the day or week, it is still not reliable because the timing can easily be changed without your being aware of it.

    2) itís practically impossible to anticipate every conceivable scenario that could undermine the strategy

    One example of a strategy for crossing which presumably considered every conceivable scenario but which overlooked an important factor was developed for crossing an arterial which is 7 lanes wide. The problem was insufficient traffic from the parallel side street to use as a cue to cross. The strategy carefully took into consideration the fact that the signal is semiactuated.

    It was observed that the arterial being crossed had a minimum time for its green signal, so that when the pedestrian button was pushed to cross it, the signal did not change until the arterial had its minimum time of 45 seconds. That is, if the pedestrian button is pushed less than 45 seconds after the signal for the arterial is red, the signal will not change to red again until the arterial road has had at least 45 seconds of green. So, when a minute passes without the signal for the arterial road being red, its signal will turn yellow as soon as the pedestrian button is pushed and, exactly 10 seconds later, the walk signal to cross begins.

    So the strategy that was developed to figure out when it was time to cross was for the pedestrian to wait until the arterial road had the green signal for at least a minute, then push the pedestrian button and count ten seconds while preparing to cross. At the end of ten seconds, the pedestrian assumes that the walk signal is on, and starts to cross.

    It would seem that everything was considered and addressed, such as the effects of actuation. Unfortunately, something that hadnít been anticipated could make it go very wrong.

    This strategy requires that pedestrians realize whenever the signal for the arterial changes to red, because they must wait at least a minute after the signal is red before they push the pedestrian button.

    But the signal can change to red without the blind pedestrian realizing it, as explained here: After the arterial road has had its minimum of 45 seconds of green signal, whenever a car passes over the sensors on the side street, the signal for the arterial road turns to yellow and then red. This happens even when the car is not coming from the side street, but rather is turning left from the arterial into the side street and cuts the corner so closely that it rides over the sensors. This happened several times during the time I was there with my client (about 6 hours altogether).

    Rather frequently, then, even when there is no car waiting on the side street, the signal for the arterial changes to red. This would be no problem if the traffic on the arterial road stopped so the blind pedestrian would realize the signal had changed. She would then know that she has to wait another minute before pressing the pedestrian button. However, if there is a gap in traffic on the arterial during the red signal, no cars will stop, and the blind pedestrian will not realize its signal had changed to red and back to green again.

    This could happen very easily. If the signal for the arterial turns red during a gap in traffic that is 10 seconds or longer, the blind person will never know the signal had changed. Because of traffic signals within a half mile in each direction along this arterial, there are many gaps in traffic on the arterial that are much longer than 10 seconds, even during busy times of the day. During one 15-minute observation of traffic on a weekday afternoon, there were 26 gaps -- almost two every minute. The gaps each lasted from 5 to 22 seconds, with the average gap being 13 seconds. Seventeen of the gaps were 10 seconds or longer, which is more than one such gap every minute.

    If a car from the arterial drove over the sensors at the beginning of one of these gaps (which is very likely if the driver was waiting for a gap to turn left into the side street), the signal would turn red for the arterial road without the pedestrian knowing it because there would be no cars on the arterial, so there would be no traffic stopping. If the pedestrian pushed the button within a minute of that car driving over the sensors, the walk signal would NOT come on in ten seconds, it would come after the arterial road had 45 seconds of green.

    So, because the pedestrian would be under faulty assumptions and using a strategy based on unpredictable features, she might start to walk across a seven-lane street when her signal is red and the traffic on the street she is crossing has a green signal!

    This is just one scenario in which assumptions about this signal could be erroneous. There are a number of ways by which actuated signals become dysfunctional and donít follow normal patterns. For example whenever the actuation wires in the street are cut during road repair or construction, the signal reverts back to fixed-time until it is repaired again, which often takes months. This 10-second strategy would not work then because the walk signal would not come 10 seconds after the button is pushed, it would come at regular intervals regardless of the button.

    We cannot predict or consider all the potential factors!

    I think very few of us -- even traffic engineers -- understand the workings of the signals enough to predict everything that could happen. I have studied, observed, and dealt with actuation and the modern traffic patterns for about 10 years, and attended sessions with traffic engineers to try to understand how the system works. Yet every time I think I finally understand actuation and how it works, I am dismayed to find an exception to the rule, a situation which I had not predicted. This happened just two weeks ago while I was presenting on this subject to the Northeastern chapter of AER, when I still complacently thought I could outwit these signals and develop reliable strategies to deal with them.

    Iíll start at the beginning. I had always taught that if you push a pedestrian button when your signal is green, the walk signal will not come on right away, it will come on at the beginning of the next cycle. One day as I was teaching this to a client, we pushed the button while our signal was green and oops! The walk signal came on immediately!

    I called the engineer and learned that this can happen when you cross secondary streets at fully actuated signals (that is, where there are walk signals and pedestrian buttons for both streets). In these situations, if the main street has the green light (for crossing the secondary street) and no vehicle or pedestrian is waiting to cross the main street, the walk signal to cross the secondary street does come on as soon as you push the pedestrian button, contrary to what we've been teaching. This fact isnít well known, even among traffic engineers -- in 1999, at a meeting of the Metropolitan Washington O&M Association with traffic engineers from state and county jurisdictions, most of the engineers didnít realize this was true. The problem was solved at this intersection when APS were installed a few weeks after we requested them.

    Meanwhile, I developed a strategy which I thought would deal with situations where we cannot get an APS. The strategy is to first push the button to cross the main street, then push the other button to cross the secondary street. Because you put in a request to cross the main street first, it will not give you the walk signal to cross the secondary street until it has responded to the request to stop traffic on the main street and allow a pedestrian to cross it. So you can be assured that you have the walk signal to cross the secondary street when the main street traffic begins again.

    However an O&M specialist at the Northeast AER conference told me that he noticed that this strategy isnít reliable either. Main streets are usually given a minimum of time before their traffic will be stopped to respond to a request to cross them, and if the main street has enough time left to allow a pedestrian to cross before its minimum time is up when you push the button, it WILL give you the walk signal to cross the other street immediately, even after youíve pushed the button to cross the main street!

    This revelation -- that a strategy, which I had assumed took everything into consideration, wasnít reliable -- convinces me that itís impossible to comprehend and take into consideration all the possible mechanisms which could affect the traffic patterns and timing of signals. For example, one mechanism which makes an exception even to our traditional rules (and which curls our toes to think about!) is a system that exists where actuated signals are coordinated with other signals along the road. In that situation, if you push the pedestrian button to cross the main road, the walk signal may not come on the next time the parallel traffic gets the green signal! This happens whenever there is time to allow a few cars to enter the intersection from the secondary street without messing up the system, but not enough to allow a pedestrian to cross. So the vehicles get a green light for a few seconds but the pedestrians do NOT get a walk signal, nor enough time to cross the street!

    How can we teach our consumers to use timing or patterns to predict when their walk signal will begin when the system is so complex, and we continue to discover exceptions to the rules? The possibility for error is too great because of exceptions of which we were not aware, and which even the traffic engineers hadnít necessarily considered. As Barlow, Bentzen and Bond wrote (2005, p. 597):

    "The lack of awareness of laws and signal-timing issues puts blind pedestrians at risk of injury and O&M instructors at risk of being considered liable for giving clients incorrect information. Updated techniques for evaluating intersections, using pedestrian pushbuttons, aligning to cross, and determining the appropriate crossing time are needed. However, at many intersections, strategies and techniques will not resolve the difficulties or provide enough information for crossing safely without access to the signal information."


    Barlow, J., Bentzen, B., and Bond, T. (2005). Crossing Strategy for Modern Signalized Intersections. Journal of Visual Impairment and Blindness, 99, pp. 587-598.

    Frieswyk, J. (2005, Winter). Crossing strategies for modern signalized intersections. Newsletter of the Orientation and Mobility Division [Alexandria, VA: Association for Education and Rehabilitation of the Blind and Visually Impaired], 11 (2), pp. 16-17

    A wealth of resources and information about APS and other accessibility issues can be found on these websites:
    U.S. Access board
    Accessible Design for the Blind
    Accessibility page of the Institute of Transportation Engineers

    Back to Traffic Signals
    Back to Crossing Streets
    Back to home page