Night vision is an important factor in understanding the cause of accidents that occur under low visibility. Here, I briefly outline some basics, roughly what I would expect my students to know at the end of an introductory perception course. [See related articles The Invisible Pedestrian and Police Shootings.]
Photopic, Mesopic and Scotopic Vision
Human eye can see over a light intensity range of several million to one. However, this number is misleading because the eye cannot see over the entire range at any given time. Instead, the eye’s “dynamic range” when adapted to a given intensity is much smaller, perhaps only about 2000:1 (Fairchild, Jiang, and Xie, 2021). This limitation is necessary to allow for good contrast perception, which is the fundamental visual task. The eye uses a variety of visual mechanisms to slide this relatively narrow dynamic range up and down the much larger adapting intensity scale.
Every beginner's textbook discusses rods and cones, so amateurs pick up on these terms and focus too heavily on them. Photoreceptors alone are insufficient to explain night vision. Moreover, rod vision and night vision are not synonymous. The more important concept is "receptive field," which is fundamental to all visual processing. A thorough appreciation of receptive fields, their various types, how they operate, how they change with conditions and how they determine visual capability is neccessary for a deep understanding of visual processing. I won't go into a full explanation of receptive fields because it is too large a topic. However, I will mention two of their properties, inhibition and convergence.
Individual cones and rods have very similar sensitivity to light. Both respond to a single quantum of light, although rods produce a bigger response. A major difference between day and night vision is inhibition and convergence, the way the photoreceptors are wired together, and the amount of light-sensitive photopigment available. Moreover, most "night vision" occurs in a mixed rod/cone mode. The overall operation of the eye in diminishing light levels is better described in terms of three operating modes, photopic, mesopic and scotopic. Photopic vision occurs at high light levels and is characterized by 1) cone photoreceptors, 2) low light sensitivity, 3) high acuity and 4) color vision. Scotopic vision occurs at very low light levels and exhibits 1) use of rod photoreceptors, 2) high light sensitivity, 3) poor acuity and 4) no color vision.
The shift from photopic to scotopic represents a delicate balancing act. Lower light level changes the balance in contradictory demands of light sensitivity on the one hand and contrast sensitivity on the other. During the day when there is plenty of light, the visual system operates in 'photopic' mode which employs cone photoreceptors and which is optimized for seeing contrast. The real action occurs after the photoreceptors, where the visual system essentially throws light away by creating "lateral inhibition," edge-sharpening that actually decreases neural activity in response to light (see Green, 1984, for example). The effect is similar to the edge-sharpening function on a computer graphics program. The cost of this edge-sharpening is lower light sensitivity. During the day when there is plenty of light, the visual system gladly pays the price.
As illumination declines, the visual system starts conserving light in three ways. First, inhibitory responses weaken, and eventually stop. Second, inhibition is replaced by convergence, where the receptor outputs sum together to increase sensitivity but further reduce resolution. Third, there is more available photopigment as light declines. When light strikes a molecule in a photoreceptor, it "bleaches" the molecule, causing electrical activation that leads to a visual sensation. While in the bleached state, it is unresponsive to light. The more photopigment in a bleached state, the less available to respond to light and the lower the sensitivity. In dim light, very little of the photopigment is bleached, so the eye has greater light sensitivity. All of this occurs before and continues after the switch from cones to rods.
One effect of switching to rods, however, is the "Purkinje shift." During photopic (cone) vision, viewers are most sensitive to light that appears greenish-yellow. In scotopic vision, they are most sensitive to light which would appear greenish-blue during the day. (Of course, viewers can't actually see color in scotopic vision. It is incorrect to say that "people are most sensitive to blue light at night.") One main result of switching to rods is loss of most sensitivity to long wavelength colors (red).
A second difference between rods and cones is their distribution on the retina, the eyes' film. Cones are concentrated in the central visual field, with the very center of the fovea consisting of all cones packed tightly together. In contrast, rods are distributed through the visual field and dominate heavily in the periphery. Since there are few rods in the fovea, viewers see best in scotopic vision if they do not look directly at an object but rather look slightly off to the side. While maximum rod concentration lies 20 degrees from the fovea, the best scotopic vision lies only 4 degrees from the fovea. This reinforces the fact that night vision depends less the individual photoreceptors and more on the way that they are wired together.
The common equation of scotopic/rod vision with night is an error. In urban environments, there is enough ambient light at night to prevent real scotopic vision. Instead, the eye operates in mesopic vision, a mixed state, where the bottom of cone and the top of the rod operating levels overlap. Most night accidents occur when the viewer is operating in the mesopic mixed rod/cone mode rather than in scotopic visual conditions.
Mesopic vision is more complicated than photopic or scotopic vision. Visual performance will depend greatly on whether objects lie in the sightline and cast images on the cone-dominated fovea or in rod-dominated peripheral vision. The fovea contains mostly cones, so viewers will continue to act as if they are using photopic vision for small, fixated objects. The result is relatively good acuity and photopic color vision. The spectral sensitivity will be photopic with no Purkinje shift to low wavelengths. In contrast, objects viewed in peripheral vision will be seen in a mixed photopic/scotopic vision. As distance from the fovea increases, the balance shifts increasingly toward scotopic vision. Acuity is poor, color vision absent and spectral sensitivity shifted to lower wavelengths. The balance also shifts more toward scotopic vision as light level declines from the top of the mesopic range toward the bottom.
Figure 1 Luminance and visual mode at various sky conditions.
Figure 1 shows typical situations where the eye would be in each operating mode. The ambient light level created by the sun level is almost independent of position until the sun falls to 5-10 degrees above the horizon. The eye's contrast sensitivity is roughly constant (Weber's Law holds) when the sun is much above the horizon. Once the sun is over the horizon, twilight begins and the change from photopic to mesopic and eventually to scotopic vision begins. Pure scotopic operation occurs only when there is no significant light source. Even good moonlight can prevent full scotopic operation.
Visual Adaptation
The eye adapts to changing light level, altering light sensitivity like a camera modifies exposure. Adaptation is an important, but frequently overlooked issue in many cases. A person's perceptual ability is determined, not only by the scene viewed at the time of the accident, but also by what he had previously viewed. In one case, for example, a man walked out of house at night, took two steps to cross a porch and fell on the front steps. The porch light was broken, making the steps difficult to see. However, the broken porch light had a second effect. The man had come out of a brightly illuminated house, so the broken porch put him in a highly misadapted state and impaired his vision.
There are four types of adaptation. The eye can dark adapt, going from bright to dark environment, or light adapt, going from dark to light. Each of these adaptations comes in two varieties, a slow and a transient phase. Most people are familiar with the slower phase process that takes about 45 minutes to complete but there is also a transient adaptation effect that has a time course of about a second or a bit more.
Slow Phase
The classic slow phase adaptation curve (Figure 2) shows the change in sensitivity as a function of time. It is a common error to believe that adaptation is merely the switch from cones to rods. For a viewer using photopic vision when the lights go out, the initial dark adaptation occurs entirely within the cones (green), as explained above, by convergence and loss of inhibition. However, the lower bound of sensitivity (blue) is still set by cones. As time proceeds, the cone sensitivity increases (threshold falls) but then levels off. Simultaneously, the rods (red) also begin to adapt. At higher light levels they are inactive and cones determine threshold. As time passes, they increase sensitivity. After roughly 5-7 minutes the photopic and scotopic curves cross and the scotopic system becomes the more sensitive and sets the lower sensitivity bound. Depending on the initial adaptation level, complete dark adaptation may require 45 minutes.
Figure 2 Classic slow phase dark adaptation curves.
Note that even when the viewer is in scotopic mode, the cones do not simply disappear. The rods set the lower sensitivity bound, but the cones will still respond to a sufficiently intense light. The belief that viewers can't see color at night is another common fallacy. A driver on a dark rural road, for example, may be in scotopic mode but could still see the color of a warning light well ahead if it were intense enough to stimulate cones.
The exact timing of dark adaptation depends on two factors: 1) the higher the level of initial adaptation, the longer the time and 2) the lower the final resting adaptation, the longer the time. For example, a person going into a movie theater during the day will take longer to adaptation than a person going into the theater at night. Rate of adaptation also depends on other factors, including wavelength, viewing time in the brighter light, area of the retina illuminated and viewer age.
Light adaptation is the inverse of dark adaptation. Viewers switching from dark to light also suffer a temporary visual loss. Light adaptation, however, happens in a few minutes, far faster than the same transition would occur going from light to dark.
Transient
Both light and dark transient adaptation occur in about a second, although exact timing depends on size of the light/dark change. Crawford's (1947) classic study investigated how light adaptation would proceed for a driver entering and leaving a tunnel. He measured sensitivity to a target at various times just before, during and after a sudden light level change. There is a large loss of sensitivity for a second or less light goes on (light adaptation) or off (dark adaptation). This transient adaptation (or "masking by light," Green, 1981; Green, 1984) occurs on a much shorter time scale, lasting a few seconds or less in most cases. This effect has major implications in many situations. Any sudden transition of lighting conditions will greatly impair vision. For example, light flashes, such as from a gun or strobe or headlamp glare, will have two effects. They will adapt the viewer to a higher level of illumination, requiring the gradual slow-phase reacquisition of dark adaptation over several minutes. But they will cause a strong short-term adaptation effect that lasts a second or two.
Lastly, adaptation effects have large safety implications. Whenever a person transitions from a brightly lit or very dark environment to one of very different luminance, there will be a large visual loss. Elderly people exhibit impaired abilities and suffer the largest visual loss due to abrupt light changes. Proper design should create smooth transitions from light to dark. For example, building entrances and tunnels now often incorporate a zone of middle light level between the dark night outside and the bright interior lighting. This smoother transition minimizes adaptation effects and permits better vision.
Eye Adaptation Vs. Camera Exposure
Some people draw an analogy between human adaptation and exposure changes of a camera. In both cases, the goal is the same: extend the dynamic range of good contrast by changing light sensitivity. However, cameras and eyes work in very different ways, making it impossible for photographs to capture what the eye records. The camera sets the exposure globally while the eye adapts locally. That is, the camera sets one exposure level for the entire scene, but each portion of the retina (the eye's "film") sets adaptation level independently. For example, suppose the scene has a small light at night. Figure 3 shows how a typical camera determines exposure using an average of the light meter to sample points from the entire image. The 9 point grid shown by the array of red squares is common, although more sophisticated cameras may use other sampling grids and point weightings. Notice that the camera is sampling the dark areas in the image, so the average reading will be low. The relatively bright lights will then be overexposed and appear far brighter than they would appear to a real viewer. It is easy to make a relatively dim and obscure object look highly conspicuous and to give the false impression that it should easily have been seen. Conversely, a scene with many bright lights will underexpose dark areas and make low contrast objects less visible than they would be to the eye. Of course, cameras using different sampling grids will produce different images of the same scene.
Figure 3 Typical 9 point camera exposure grid.
Some "experts" attempt to remedy this problem by taking a series of photographs at different exposures, assuming that one will accurately represent the scene. After the photographs are developed or shown on a display, the "expert" chooses the one that accurately represents the scene. However, there are several problems with this approach. First, memory for brightness is poor, so it is highly unlikely that a person could remember a previously viewed scene with much precision. Any attempted match between a photograph and a memory is highly questionable. Moreover, the photograph/display is being viewed in different lighting under a different state of adaptation. There are other issues, such as the effect of changing aperture on depth of field, which creates blur and distortions in perceived distances. Second, the eye simply does not work like a camera when determining "exposure." Each area of the eye adapts to the local illumination, so there isn't a single "exposure" set for the entire image. Afterimages prove the point. If you look at a black dot on a white background and then at a uniform gray surface, you see the white afterimage of the dot. The area of retina being stimulated by the dot was adapting to black while the rest of the retina was adapting to white. The gray would then look bright to the dark adapted area and dark to the white adapted area. This is only one of many ways the eyes and cameras differ.
Figure 4 Local retinal adaptation.
Lastly, there have been attempts to model local retinal adaptation so that images can be corrected for local adaptation effects. The method sets separate rod and cone response functions for each pixel. The methods are complex and their accuracy unclear, although some examples are highly compelling. As already noted, for example, adaptation is a convergence and not just the response of individual photoreceptors. Object size, viewing time and many other factors may be important.
Environmental Lighting Condition
Any visibility assessment starts by examining lighting conditions. Outdoors, natural illumination is often the main light source and sun and moon positions determine visibility. During the day, however, there is no moon and sun position has little effect on visibility. As the sun approaches the horizon, and especially for about a half hour after sunset, visibility undergoes a rapid decline. Moonlight is a significant factor only in very dark rural locations.
Daylight
The sun creates three main illumination components. The direct component is light that travels on a straight line from the sun to the surface. The amount of illumination falling on an outdoor surface is then related to sun position, which is specified by elevation (angle above the horizon) and azimuth (clockwise horizontal angle from true north), and the relative orientation of the surface. On cloudy days there is no direct component because the sun is hidden behind clouds. The sun still creates a second, diffuse component by means of atmospheric scatter. This lessens the importance of azimuth variation. Similarly, as the sun goes below the horizon, the effect of azimuth variation becomes vanishingly small. Lastly, the sun can illuminate objects indirectly through reflection off other surfaces, such as the ground.
IESNA report Recommended Practice For the Calculation of Daylight Availability (1983) provides a more detailed method for calculating both direct and ambient solar illumination as well as a model for sky luminance. The formula contains some errors that are corrected by supplements.
Twilight And Night
Several landmark astronomical events occur during the change from photopic to scotopic vision. One is sunrise/sunset, the time when the upper edge of the sun's disk is on the horizon. The second is twilight, the time when the sun is below the horizon but still providing illumination. Finally night begins at the end of twilight in the evening and at the beginning of twilight in the morning.
Twilight is a common time for outdoor accidents. Contrast sensitivity declines rapidly as the eye moves to mesopic operation. The sun is still providing both foreground and background illumination, so artificial light sources, such as car headlamps are less effective (see Green, 2002c). In addition, visual skills, such as reaction time, fall rapidly (Campbell, Rothwell and Perry, 1987).
The US Naval Observatory defines the three phases of twilight as:
Civil twilight: "the center of the Sun is geometrically 6 degrees below the horizon...illumination is sufficient, under good weather conditions, for terrestrial objects to be clearly distinguished.";
Nautical twilight: "the center of the sun is geometrically 12 degrees below the horizon. General outlines of ground objects may be distinguishable, but detailed outdoor operations are not possible, and the horizon is indistinct"; and
Astronomical twilight: the center of the Sun is geometrically 18 degrees below the horizon.
There are many sources for equations that will estimate solar position and illumination. The US
Naval Observatory site (http://aa.usno.navy.mil/) provides data on sun and moon position for any location on earth at any time of day in the recent past or future. The observatory's tables, however, assume a perfectly flat horizon at sea level. Buildings, hills and other irregularities on the horizon can change estimates. In addition, local weather conditions cause refraction changes to the atmosphere and can produce errors for sun locations near the horizon. Once the sun position is known, ground illumination may be obtained from the Explanatory Supplement To The Astronomical Ephemeris And The American Ephemeris And Nautical Almanac (Seidelman, 1992) or from Naval Observatory Circular 171, Computer Programs for Sun and Moon Illuminance With Contingent Tables and Diagrams," which has both an illuminance/sun altitude diagram and a hopelessly outdated BASIC computer program.
Lastly, an expert will usually be making light measurements several years after the original event. According to the US Naval Observatory, for example, on Sept 15 2001 at 8:08 PM EDT in Pittsburgh, the sun was 8.3 degrees below the horizon at an azimuth of 280.8 degrees (clockwise from north). On the same day and time in 2004, the sun's position was -8.4 degrees altitude and 280.7 degrees azimuth. Obviously, this variance is too small to have any practical significance.
Conclusion
I have introduced some basics of night vision. The main tasks are 1) determination of lighting conditions through examination of sun and moon position data, 2) calculation of available contrast and 3) assessment of visibility through an understanding of visual function. This is only the start, however, because visibility only says whether a viewer's sensory apparatus has capability to sense an object, not whether a viewer is likely to see it. Conscious perception is also determined by attention and by other cognitive processes.
References
Crawford, B. (1947). Visual adaptation in relation to brief conditioning stimuli, Proceedings of the Royal Society of London B, 134, 283-302.
Fairchild, M. D., Jiang, F., & Xie, H. (2021). Perception in HDR and displays. Information Display, 37(3), 15-19.
Green, M. (1981). Spatial frequency effects in masking by light. Vision Research, 21, 801-806.
Green, M., (1984). Masking by light and the sustained-transient dichotomy. Perception and Psychophysics, 34, 617-635.
IESNA (1983). Recommended Practice For the Calculation of Daylight Availability.
Seidelman, P. (1992) Explanatory Supplement To The Astronomical Ephemeris And The American Ephemeris And Nautical Almanac. (Herndon, Va.: University Science Press).