.

Accidents at Rail-Highway Crossings

Marc Green


Nearly two-thirds of all crossing accidents occur during daylight hours. Because two-thirds of all accidents occur at crossings equipped with automatic warning devices, driver inattention is clearly the major cause.- Conrail.

This statement may be accurate but it is untrue. It places the blame on factors inside the drivers' heads rather than where it belongs - on inadequately designed warnings. Below, I explain why.

There is no doubt that railroad-highway crossing accidents are common. They kill or seriously injure approximately 2400 a year, a number which has not significantly changed over the last 15 years. Accidents may occur either because the driver races the train through the crossing or because the driver simply fails to see the train coming.

Drivers Lose The Race

A driver approaching a crossing must correctly decide whether his/her vehicle will arrive at the crossing before the train. This requires correct perceptual speed judgment of both the driver's own vehicle and the train. Accidents frequently happen because the driver underestimates the approaching train's speed.

Collision Course Objects Don't Make Moving Retinal Images

Humans, in part, see motion by registering the movement of an object image projected on the retina, the light-sensing portion of the eye. When a driver looking ahead at the road is on a collision course with an object approaching from the side, there is no retinal image motion.

The figure illustrates the point by showing the driver approaching a crossing while a train is coming. The driver notices the train at time T, and the train's image falls in the visual periphery at angle, theta. As the driver moves forward, the train moves leftward. Three things can happen:


  1. The driver arrives before the train. In this case, the angle increases, and the retinal image moves farther to the periphery. (This becomes obvious if you visualize the train as remaining stationary.)

  2. The train arrives before the driver. In this case, the angle decreases, and the retinal image moves toward the fovea. (Interestingly, people are better at judging motion toward the fovea than away from it.)

  3. The driver and object arrive at the same moment. In this case, theta remains constant, as shown at time T+t. The car and train movements cancel out, and the angle remains constant. There is no retinal image motion, so speed is more difficult to accurately judge. Of course, the same principle applies for any type of collision course, whether it is car-train, car-bicycle or car-motorcycle, and for angles of approach other than 90 degrees.

    1. There is one caveat to add. When the driver is very close to the tracks and the train's motion is almost directly toward the driver, the retinal image expands almost equally in all directions. In this "looming" situation, the driver can use the image's expanding edges for retinal image motion cues to judge time-to-collision (TTC). However, by the time that the driver is close enough to the tracks to use looming as a motion cue, it is usually too late to stop. Moreover, there is evidence that people tend to underestimate speed in such situations, so the driver would likely expect the train to arrive later than it actually would. TTC judgment, however, is critical for someone moving parallel to or on the tracks.

      Large Objects Appear To Move Slower

      People judge large objects as moving slower than smaller ones. Because trains are large objects, drivers underestimate train speeds. This general bias is further compounded by "object familiarity." When drivers see the train, they base speed judgment on their more common experience of judging motion of automobiles, much smaller objects. Speed underestimation is then reinforced.

      Lastly, drivers may race across the tracks because the flashing signal has low credibility. Flashing warnings usually begin far in advance of the train, so people learn that an activated signal does necessarily not mean that a train will arrive soon. This creates a "cry wolf" situation; if the lights begin to flash too far in advance of the train, then ironically the danger signal is transformed into a safety signal - it communicates the message that there definitely won't be a train coming for awhile. The message communicated by warnings is not always the intended one.

      Drivers Fail To See The Warning

      It might seem unlikely that a driver would miss an object as big as a train or as conspicuous as a flashing light, yet this occurs frequently. One of the reasons is that warnings at railroad-highway crossings are far less effective than those used at normal highway intersections.

      Railroad-highway warnings may be passive, such as the crossbuck sign, or active. The major active warning is a set of alternating red lights, a standard based on tradition rather than on good human factors design. Before the introduction of automatic electric signals, railroad employees warned of trains by swinging a red lantern back and forth. The flashing light design simply emulates the old tradition.

      Beams Are Narrow And Require Careful Adjustment

      The standard flashing lights have severe limitations. The fundamental deficiency originates with the need to operate on batteries in case of power failure. Lamp power consumption must be low, so the bulbs have far less intensity than those of normal traffic signals. The low-wattage bulb is then covered with a red glass roundel, which filters and reduces light by as much as 80-90%. The red cover means that the spectral output is mostly long wavelength, where the eye is not very sensitive. Lastly, the lights are relatively small. Newer installations are 12-inches in diameter, but there are still many of the older 8 3/8 inch models in service.

      In order to compensate for the weak output, reflectors mounted behind the bulbs concentrate the light into a narrow beam. The beam is about 30 degrees wide, but is more intense in the center than in the periphery. The driver looking down the center axis will see much of the bulb's light output so little is wasted. However, the beam will not appear bright for viewers who are looking even slightly off axis. For example, one study of found that intensity drops by half for a viewer who is only 3 degrees off axis. That's about twice the width of your thumb held at arm's distance. If the light is not aimed directly at the driver, with very little error, its brightness degrades significantly.

      Location of the flashing lights off to the side of the road further compounds problems. In order to ensure that the approaching driver sees the bright beam center, the light must be aimed to intersect the road at some distance. Before and after that distance, the driver's line of light will only be along the dimmer beam periphery. The exact intersection distance varies with the exact light configuration, but 100 to 400 feet are the typical values.

      The beam is even narrower vertically than horizontally. The beam strength falls off greatly with deviations of only 1-2 degrees above and below the center. The normal adjustment sets beam height to reach the eye of automobile drivers. For a truck driver, who sits higher, the beam center will fall below the line of sight and the beam will be far weaker.

      The narrowness of the beam also means that proper adjustment is critical. The lights have many finicky pieces to adjust, and must operate in an environment where frequent vibration causes misalignment. Further, the environment is prone to cover the roundel with dust and dirt that filter and disperse the beam. Anytime there is an accident where a driver may have missed the warning lights, maintenance records should immediately be inspected. Crews should frequently realign beams and clean off dust and dirt from the glass housing to maintain even minimal signal brightness.

      Sunlight Obscures The Signals

      The problems created by the warning light's low intensity and narrowness of focus are compounded by reflections off the roundel. In order to detect that a train is coming, the driver must see the light flash. However, the light does not black out when the bulb is off. Light from the sun and from the atmosphere reflect off the surface back to the driver's eye. This is usually called "veiling light" or 'veiling glare" because it lowers contrast as when looking through a veil. The task is then to detect the brightness increment created by the bulb from this background reflection. Increased veiling light decreases the flash amplitude and lowers conspicuity. The amount of veiling depends on the cloud cover, atmospheric clarity, and the sun position.

      Reflection will be greatest on a clear day when the sun is behind the driver and the sun angle is low - early or late in the daylight period, depending on which direction the lights face. The hood over the light blocks direct light when the sun elevation is more than 45 degrees above the horizon. As the sun declines in elevation, direct sunlight begins to reflect off the signal light. At 30 degrees elevation, half the light receives direct sunlight. At 15 degrees, the reflection is greatest because sun intensity begins to fall at lower elevations. Regardless of sun elevation, however, there is always some reflected light from atmospheric scatter and from reflection off the ground.

      Black Backgrounds Can Make Flashing Lights Harder To See

      In an attempt to increase visibility, the lights are mounted in front of a black metal backdrop. The designers doubtless thought that they were maximizing brightness of the light when it flashes. While this may be true at night, it has exactly the opposite effect when there is significant reflection.

      In order to understand why, I must explain how people see flashing lights. The intuitive notion is that we see flashing because there is no light coming from the bulb when off and then a pulse of light coming from the bulb when on. The perceived magnitude of the flash is presumably determined by the amount of light produced when the bulb is on.

      This is not quite correct. When viewing a flashing light, the eye actually only judges the change in "temporal" contrast" between the light and its surround, in this case the black backdrop. The fundamental factor of perception is that humans see contrast at edges, not absolute amounts of light. This cannot be stressed too often because many find it an unintuitive notion.

      However, black surrounds impair detection of light flashes because they mask contrast. A full explanation would require a description of Weber's Law, contrast gain control and some other details of contrast detection. In any event, there are many studies that have shown that the presence of black surrounds impairs flash detection.