Driving at night is one of the most hazardous situations commonly faced by the driver. It has been shown that low light levels encountered when driving at night,1,2 especially on roads that have no street lighting,3 can account for the disproportionately high observed nighttime accident rate. The effect of lighting on accidents involving pedestrians is even more pronounced, with pedestrians being three to seven times more vulnerable in the dark than in the daylight.4,5
A credible physiological explanation for the importance of good lighting when driving is based on the poor spatial and temporal characteristics of rod photoreceptors, which are largely responsible for visual perception at low light levels. As a result of these characteristics, many aspects of visual performance deteriorate under reduced lighting conditions (see Charman6 for a review), including spatial resolution,7–9 contrast discrimination,10,11 stereoscopic depth perception,12 accommodation response,13–15 and visual reaction time.3,16,17 This degradation of visual performance under low illumination is more pronounced in the elderly7,18–21 and has been linked to their involvement in motor vehicle collisions.22–25
The visual field of the driver is subject to continuous illumination changes. The driver is exposed to a remarkably wide range of luminance values, which necessitates variations in his state of adaptation, sometimes of a rapid and extensive nature. Obvious adaptation problems occur when moving from one level of ambient lighting to another, e.g., when driving from a well-lit tunnel (or a highly lit commercial area) into an adjacent unlit area. It is clear that, in these situations, it will take some time before the driver's vision has reached its maximal sensitivity.
A key factor in relation to the role of adaptation state in visual performance is the actual level of road lighting at night, which is typically found to be in the mesopic region, between 0.5 and 10 lux.15,16 The aim of the present study is to assess the state of retinal adaptation under conditions that simulate road lighting when driving at night, namely a roundabout associated with a high accident risk. This might reveal the underlying mechanisms that dictate visual perception. Moreover, because most of the visual field used while driving is peripheral26,27 (e.g., traffic signs, pedestrians, cyclists, other vehicles), and it is obvious that most unexpected objects must be detected in the periphery, a range of retinal eccentricities and target sizes is tested. A final point involves the estimation of the speed of recovery of visual sensitivity; this is essential because it is known that the adaptation level is subject to rapid changes as a result of the presence of other light sources such as the headlights of oncoming vehicles.
Three young subjects (aged 21, 25, and 26 years) participated in the study. They were optically corrected for any refractive error with glasses and their monocular visual acuity was better than 6/5. No ocular pathology was detected. Two of the subjects (LL and ES) were naive to the objectives of the experiment.
Dark adaptation curves were obtained using a slightly modified Goldmann-Weekers adaptometer (Haag-Streit AG, Ophthalmological Instruments, Bern, Switzerland) (see Fig. 1). The instrument contained an opal glass test target. The test field was illuminated by white light from an incandescent bulb having a correlated color temperature of 2250°K. Intermittent illumination of the test field was produced with the operation of a revolving diaphragm at a frequency of 0.5 Hz. The field luminance was controlled by neutral density filters (7 log units) and was calibrated with a digital spotmeter (Photo Research 1500; Micron Techniques Limited, London, U.K.) so the data from the raw records could be converted to candelas per square meter (cd/m2). Stimuli were viewed through natural pupils; thus, luminance rather than retinal illuminance units were used in our measurements. In all cases, pupil size was bigger than 5 mm.
Circular field stops were mounted on the opal glass, which allowed the angular subtense of the diameter of the target to be varied from 1° to 5° of visual angle. The test field was surrounded by a matt black cone (funnel) of low reflectance to eliminate reflections on the surface of the test field caused by ambient lighting.
Four small, dim, and red light-emitting diodes were used to control eccentric fixation. They were mounted on four adjacent housings at 10°, 20°, 30°, and 40° in the temporal visual field of the subject; only one fixation target was present for each threshold measurement. For central fixation, the subject was asked to fixate at the center of the test field. The fixation targets were placed at the same plane as the test field to avoid any accommodation differences.
Initial light adaptation was achieved using a hemisphere of 30 cm diameter. Its internal surface was coated with matt white paint that gave a uniform light adaptation of the entire field. The preadaptation exposure time was 1 min and the luminance level 3200 cd/m2, estimated to produce a retinal illuminance of 4.12 log trolands (assuming a pupil of 2.3 mm diameter).
Retinal thresholds were obtained under different levels of artificial lighting. Apart from the test field and the fixation target, the rest of the visual field during adaptation was a “yellow” diffusing surface (reflectance approximately 90%) uniformly lit to a controllable level of illuminance (ranging from 0.1 to 5.0 lux at the surface of the test field) with the aid of two lightboxes covered by broadband yellow filters (with these filters, 80% of the overall spectral energy of the source is contained between 540 and 650 nm). The level of ambient illuminance was calibrated with a digital lux meter (OM 210; Robin Electronics Ltd., U.K.). These spectral/illuminance characteristics were purposely chosen to simulate urban lighting conditions like found at a local roundabout, which was illuminated by low-pressure sodium lamps and was associated with a high accident risk. Under these conditions, horizontal illuminance ranged between 1.0 and 8.7 lux when standing on the pavement, which reduced to values between 0.2 and 1.4 lux from inside the vehicle (as a result of the transmission effects of the windscreen and the roof of the vehicle).
Each observer was seated in the experimental room. The normal lights in the room were turned off. After a lapse of approximately 1 min, the retinas of both eyes were bleached by exposure to the standard 3200 cd/m2 light source for 1 min.
The observer was seated at the dark adaptometer with his or her left eye covered by a patch. A constant head position was maintained by an adjustable chin and forehead rest placed 41 cm from the target. The subject was instructed to look at the red fixation target, which was placed at various eccentricities, while being aware of the flashing test field. The subject's task was very simple and involved determining the minimum luminance (threshold), by adjusting a knob, at which he or she could just see the test field. The threshold was measured using the method of ascending limits (from nonseeing to seeing). The first observation was made within the first 15 to 30 s after the termination of the bleaching light. Then, as many readings as possible were taken until the thresholds reached a steady level (fully adapted threshold).
Retinal Adaptation Under Mesopic Conditions
Figures 2A and 2B show results of retinal adaptation at different ambient illuminance levels (5.0, 1.0, 0.5, and 0.1 lux) as compared with the typical dark adaptation curve for subjects SP and LL. It is obvious that the retinal adaptation curve alters markedly with ambient lighting. In total darkness, there is a clear discontinuity in the curve, which is attributed to two distinct regions of recovery, dominated initially by cone and subsequently by rod photoreceptor function. However, at upper mesopic levels (5.0 lux), the curve consists of one portion undergoing a monotonic increase in sensitivity possibly attributed to cones only. Similarly, at 0.5 lux, no break is evident, suggesting that the rod recovery is desensitized by the cone system, which dominates at these levels. If ambient illuminance is decreased to 0.1 lux (low mesopic levels), there is a slight inflection followed by a second rod-dominated phase of adaptation. Presumably, the discontinuity between the first and second segments represents the transfer from cone to rod vision. Identifying the extent of this dichotomy is one of the main objectives of the present project.
To summarize, there is a suggestion that vision may be qualitatively different for low-mesopic (0.5-lux) and high-mesopic (5-lux) conditions, with the former being mainly rod-mediated and the latter mainly cone-mediated. Fully adapted sensitivity increases as ambient illumination decreases, and the maximal sensitivities under mesopic light levels are reduced at least 2 log units compared with complete darkness. Because rods are more common in the peripheral than central retina, we might expect this to be reflected in a sensitivity versus eccentricity function. The data for this experiment are described in the next section.
It is also evident in Figure 2 that for a background illuminance of 0.1 lux, the changeover point between the two parts of the adaptation curve is delayed by approximately 1 min compared with that for absolute darkness. This implies that the rate of adaptation for cones is slower under high-mesopic than low-mesopic conditions. To test this idea in subsequent experiments, an index of the rate of adaptation, i.e., of the speed of recovery of retinal sensitivity after the initial bleaching, is calculated. It is assumed that when the retinal adaptation curve consists of only one phase, it could be represented by the general equation:
where T is the absolute threshold, t is time, A is the initial level of adaptation, and B is the level of the asymptote of the curve. Taking natural logs give:
i.e., a plot of ln[logT-B] against time, t, should be a straight line of slope -k, which represents the rate of retinal adaptation. B is calculated by the mean of the last five values of the asymptote. Figure 3 tests this hypothesis. It should be noted that as a result of the rapidly changing threshold, there are not many data points for analysis.
Effect of Stimulus Location on Retinal Adaptation
Figure 4 depicts the fully adapted threshold of the three subjects for different eccentricities in the temporal retina at the two different ambient lighting levels (0.5 and 5.0 lux). It is clear that overall, as the stimulus is presented more temporally (at eccentricities higher than 10°), the eye requires more stimulus intensity for detection. For two subjects, there is an approximate linear decrease in sensitivity with eccentricity and the slope of the graph is approximately the same for low and high illumination. For subject LL, the two functions converge so that at 40°, they are approximately the same. Crucially, at 0.5 lux, the maximum sensitivity shifts to approximately 10° of retinal eccentricity for subjects SP and ES, indicating rod-dominated perception under these conditions. This is seen as a peak in the ratio of the sensitivities for the 0.5 versus 5 lux conditions.
Figure 5 shows plots of the rate of adaptation for the same stimulus conditions and light levels. There is a decline in the rate of adaptation with increasing retinal eccentricity; this is more evident at the higher mesopic levels (5.0 lux). This is not surprising because it is known that cones adapt at faster rates than rods and that spatial density of cones decreases outside the fovea.
These data further emphasize the rod–cone dichotomy between the low and high illuminations. They suggest that although there may be little or no difference in sensitivity of the two detecting mechanisms, it is likely they will have very different temporal characteristics, and as outlined in the “Discussion,” this is crucial when driving.
Effect of Stimulus Size on Retinal Adaptation
In Figure 6A, fully adapted thresholds and rates of adaptation are plotted as a function of stimulus size for 0.5 lux. Data for only one subject are presented, but they are representative of observations made with the other two. For central fixation, no significant difference in absolute threshold with stimulus size is observed above a critical size of approximately 2°. However, for the peripheral part of the retina (20° and 30° temporal eccentricities), the critical sizes above which there are no improvements in sensitivity are approximately 3° at 20° eccentricity and above 5° for 30° eccentricity. Note that for 30° eccentricity, there is a roughly linear relationship between stimulus area and retinal sensitivity.
Although larger stimuli produce summation of signals in periphery, this does not necessarily improve the recovery rate of sensitivity, probably as a result of the fact that detection is still rod-dominated. This is confirmed in Figure 6B, which illustrates that there is little change in the rate of adaptation with increasing stimulus size for both central and peripheral parts of the retina.
DISCUSSION: IMPLICATIONS FOR DRIVING
Level of Adaptation
By studying retinal adaptation levels under light levels that simulated those encountered when driving at night, it was found that absolute thresholds ranged between 5.2 × 10−3 (at 5 lux) and 4.7 × 10−4 cd/m2 (at 0.1 lux). These values are slightly higher than those described by Davey and Sheridan,28 who reported dark adaptation thresholds of 1.2 × 10−4 cd/m2 when driving in the city center compared with 2.3 × 10−5 cd/m2 on unlit roads. This denotes lower levels of adaptation in the present study, probably resulting from improvements in road lighting over the years, which have led to higher levels of ambient illuminance. Moreover, our experiments took place in a uniformly lit environment, which simulated lighting levels encountered at night but cannot apply to a real driving scene, where the visual field of the driver is usually subject to continuous illumination changes. Also, it has to be stressed that these values apply for drivers; pedestrians are expected to show higher sensitivity,29 which means that the pedestrian's visual system is better primed to detect targets than the driver and that therefore pedestrians may overestimate their own visibility and thus take undue risks.30
It was also shown that the critical level of illuminance, where the rod-dominated second phase appears, lay in the region between 0.1 and 0.5 lux for young observers. As a result, it is expected that in the absence of road lighting, the rod system will tend to dominate perception, improving detection of objects of interest, but also leading to increased response times. However, it is almost certain that as a result of rod–cone interactions, this level may be varied depending on the spatiotemporal characteristics of the stimulus.31–33 It is expected that older subjects would show an increased critical level of illuminance and a slower recovery of sensitivity18,34 because of the reduced amount of light reaching their retina (e.g., pupillary miosis, increased ocular media density), changes in retinoid cycle, and degeneration of visual pathways.20,21,35
A point that is particularly relevant to the observations described in this article is that usually when driving in urban/residential roads, most of the field is of similar luminance. On rural roads, however, most of the field will be dark and only a part of the field will be illuminated by vehicle headlights, which causes further problems for the adaptation level of the driver.
The Importance of Peripheral Visual Sensitivity
The driver depends on peripheral vision to rapidly identify hazardous situations. Therefore, data based solely on human foveal visual detection capabilities may be inadequate to predict practical performance, especially when objects first appear in the periphery of a driver's visual field. Our findings (see Fig. 4) confirm that the luminance of targets, when placed in the far periphery (approximately 40°) of the subject's visual field, must be increased by a factor of 2 in comparison with foveal targets to assure visual detection. Furthermore, the decline in the rate of retinal adaptation with increasing eccentricity shows a slower recovery of sensitivity for peripheral targets.
These results are in agreement with previous studies,36,37 which reported that the ability of subjects to detect and identify suprathreshold targets (e.g., road signs) decreases considerably as visual angle increases.
The Importance of Stimulus Size
Obviously, it is visual size (angular subtense) rather than physical size, which is a basic parameter in determining the visibility of an object to a driver. Also, the conspicuity of an object, among other factors, increases for larger sizes. The visual size of road traffic objects (vehicles moving in the same or opposite direction, cyclists, pedestrians, road signs, signal lights, and so on) varies with distance. Hills38 reported how visibility distance increases with the visual size of an object. Similar calculations based on our data are displayed in Table 1.
As expected, the results presented show that retinal sensitivity increases as the size of the target increases (Fig. 6). Moreover, it seems that there is a critical size of the target that gives maximal sensitivity, which varies with eccentricity (see Table 2). For central targets, there is no increase in sensitivity for sizes larger than 2° (at 0.5 lux illuminance), whereas at 20°, the critical size is 3° and at 30° larger than 5°. Therefore, unless the luminance (or contrast) of an object is adequate for its size and position in the visual field, it may not be sufficiently conspicuous to be detected by drivers.
It should also be mentioned that when a pedestrian or a traffic sign are relatively near to the driver (e.g., located on the right-/left-hand side of the road), they are frequently detected by peripheral vision. Therefore, although their visual size increases, the retinal sensitivity may not improve, because sensitivity is relatively low in the periphery (see Fig. 4).
This study investigated the state of retinal adaptation under typical nighttime ambient illuminances. It was shown that at levels close to 0.1 lux, the rods inhibit perception, and although retinal sensitivity increases, the adaptation rate of the system slows down notably. We have previously shown that this leads to increased reaction times.3,39 Furthermore, both retinal sensitivity and speed of recovery are considerably decreased in the near periphery, and this is of critical importance if we consider that most of the visual field used while driving is peripheral. The interaction between size and eccentricity is also important. Although sensitivity increases with object size in the peripheral field, for central targets, there seems to be a "critical" size, which achieves maximal sensitivity beyond which it does not increase.
The experiments in this work involved stationary targets. However, objects to be perceived by the driver are rarely stationary but generally move with differing speeds and directions relative to the driver. Second, the driver is in motion while driving and is constantly making head and eye movements, thus displacing retinal image locations. This may or may not increase object visibility depending on the spatial and spectral characteristics of the targets.
Finally, our measurements refer to a driver who is alert, is looking in the right place, and is not distracted by nearby objects competing for his attention. However, there is strong evidence that peripheral sensitivity to a simple light is reduced when combined with a foveal, cognitive task.40,41 This means that some features of interest falling on a driver's peripheral retina may go unnoticed under conditions of excessive driver visual workload or fatigue.42 Such distractors may be in the road environment (e.g., a complex road network in an unfamiliar city) or an in-vehicle device (e.g., a mobile phone,43 an advanced information display44), and they undoubtedly contribute to a reduction in the driver's functional field of view.45–47 This may be more pronounced in conditions of reduced visibility, encountered when driving at night, when degradation of "focal" vision is noticeable.48,49
This work is part of a LINK Transport Infrastructure and Operations program. The authors gratefully acknowledge the support of the Engineering and Physical Sciences Research Council (EPSRC) and Urbis Lighting Limited. Preparation of this article was partially supported by The British Council and the Greek Secretariat for Research & Technology. The authors also thank Dr. Kamlesh Chauhan for many helpful comments during the course of this research.
Sotiris Plainis, PhD
Vardinoyiannion Eye Institute of Crete (VEIC)
Department of Ophthalmology, School of Medicine
University of Crete
P.O. Box 2208, 71003
Heraklion Crete, Greece
1. Owens DA, Sivak M. Differentiation of visibility and alcohol as contributors to twilight road fatalities. Hum Factors 1996;38:680–9.
2. Andre J, Owens DA. The twilight envelope: a user-centered approach to describing roadway illumination at night. Hum Factors 2001;43:620–30.
3. Plainis S, Murray IJ. Reaction times as an index of visual conspicuity when driving at night. Ophthal Physiol Opt 2002;22:409–15.
4. Owens DA, Sivak M. The role of visibility in nighttime traffic accidents. Proceedings of the 2nd International Symposium on Visibility and Luminance in Roadway Lighting; Orlando, Florida; 1993. New York: Lighting Research Institute; 1993:133-47.
5. Sullivan JM, Flannagan MJ. The role of ambient light level in fatal crashes: inferences from daylight saving time transitions. Accid Anal Prev 2002;34:487–98.
6. Charman WN. Night myopia and driving. Ophthal Physiol Opt 1996;16:474–85.
7. Sloane ME, Owsley C, Alvarez SL. Aging, senile miosis and spatial contrast sensitivity at low luminance. Vision Res 1988;28:1235–46.
8. Arumi P, Chauhan K, Charman WN. Accommodation and acuity under night-driving illumination levels. Ophthal Physiol Opt 1997;17:291–9.
9. Howard CM, Tregear SJ, Werner JS. Time course of early mesopic adaptation to luminance decrements and recovery of spatial resolution. Vision Res 2000;40:3059–64.
10. Peli E, Arend L, Labianca AT. Contrast perception across changes in luminance and spatial frequency. J Opt Soc Am (A) 1996;13:1953–9.
11. Arend LE. Mesopic lightness, brightness, and brightness contrast. Percept Psychophys 1993;54:469–76.
12. Allen MJ. Vision and driving. Traffic Safety 1969; September 8-9:38-40.
13. Jiang BC, Gish KW, Leibowitz HW. Effect of luminance on the relation between accommodation and convergence. Optom Vis Sci 1991;68:220–5.
14. Johnson CA. Effects of luminance and stimulus distance on accommodation and visual resolution. J Opt Soc Am 1976;66:138–42.
15. Charman WN. Night myopia and driving. Ophthal Physiol Opt 1996;16:474–85.
16. He Y, Rea M, Bierman A, Bullough J. Evaluating light source efficacy under mesopic conditions using reaction times. J Illumin Engin Soc 1997;26:125–38.
17. Mansfield RJ. Latency functions in human vision. Vision Res 1973;13:2219–34.
18. Howard CM, Tregear SJ, Werner JS. Time course of early mesopic adaptation to luminance decrements and recovery of spatial resolution. Vision Res 2000;40:3059–64.
19. Jackson GR, Owsley C. Scotopic sensitivity during adulthood. Vision Res 2000;40:2467–73.
20. Eisner A, Fleming SA, Klein ML, Mauldin WM. Sensitivities in older eyes with good acuity: cross-sectional norms. Invest Ophthalmol Vis Sci 1987;28:1824–31.
21. Elliott DB. Contrast sensitivity decline with ageing: a neural or optical phenomenon? Ophthal Physiol Opt 1987;7:415–9.
22. Owsley C. Vision and driving in the elderly. Optom Vis Sci 1994;71:727–35.
23. Owsley C, Ball K, McGwin G, Sloane ME, Roenker DL, White MF, Overley ET. Visual processing impairment and risk of motor vehicle crash among older adults. JAMA 1998;279:1083–8.
24. Keltner JL, Johnson CA. Visual function, driving safety, and the elderly. Ophthalmology 1987;94:1180–8.
25. Wood JM. Age and visual impairment decrease driving performance as measured on a closed-road circuit. Hum Factors 2002;44:482–94.
26. Wood JM, Troutbeck R. Effect of restriction of the binocular visual field on driving performance. Ophthal Physiol Opt 1992;12:291–8.
27. Johnson CA, Keltner JL. Incidence of visual field loss in 20,000 eyes and its relationship to driving performance. Arch Ophthalmol 1983;101:371–5.
28. Davey JB, Sheridan M. Levels of dark adaptation when driving at night. Br J Physiol Opt 1957;14:183–9.
29. Olson PL, Aoki T. The Measurement of Dark Adaptation Level in the Presence of Glare. Report UMTRI-89-34. Ann Arbor, MI: The University of Michigan Transportation Research Institute; 1989.
30. Shinar D, Schieber F. Visual requirements for safety and mobility of older drivers. Hum Factors 1991;33:507–19.
31. Goldberg SH, Frumkes TE, Nygaard RW. Inhibitory influence of unstimulated rods in the human retina: evidence provided by examining cone flicker. Science 1983;221:180–2.
32. Hess RF, Mullen KT, Nordby K. Mutual rod–cone suppression within the central visual field. Ophthal Physiol Opt 1992;12:183–8.
33. Naarendorp F, Denny N, Frumkes TE. Rod light and dark adaptation influence cone-mediated spatial acuity. Vision Res 1988;28:67–74.
34. Werner JS, Bieber ML, Schefrin BE. Senescence of foveal and parafoveal cone sensitivities and their relations to macular pigment density. J Opt Soc Am (A) 2000;17:1918–32.
35. Weale RA. Senile ocular changes, cell death, and vision. In: Sekuler R, Kline D, Dismukes K, eds. Aging and Human Visual Function. New York: A R Liss; 1982:161-71.
36. Zwahlen HT. Traffic sign reading distances and times during night driving. In: Transportation Research Record 1495. Washington, DC: Transportation Research Board, National Research Council; 1995:140-6.
37. Zwahlen HT. Conspicuity of suprathreshold reflective targets in a driver's peripheral visual field at night. In: Transportation Research Record 1213. Washington, DC: Transportation Research Board, National Research Council; 1989:35-46.
38. Hills BL. Vision, visibility, and perception in driving. Perception 1980;9:183–216.
39. Plainis S, Murray IJ. Neurophysiological interpretation of human visual reaction times: effect of contrast, spatial frequency and luminance. Neuropsychologia 2000;38:1555–64.
40. Wall M, Woodward KR, Brito CF. The effect of attention on conventional automated perimetry and luminance size threshold perimetry. Invest Ophthalmol Vis Sci 2004;45:342–50.
41. Plainis S, Murray IJ, Chauhan K. Raised visual detection thresholds depend on the level of complexity of cognitive foveal loading. Perception 2001;30:1203–12.
42. Lansdown TC, Brook-Carter N, Kersloot T. Distraction from multiple in-vehicle secondary tasks: vehicle performance and mental workload implications. Ergonomics 2004;47:91–104.
43. Lamble D, Kauranen T, Laakso M, Summala H. Cognitive load and detection thresholds in car following situations: safety implications for using mobile (cellular) telephones while driving. Accid Anal Prev 1999;31:617–23.
44. Lansdown T, Fowkes M. An investigation into the utility of various metrics for the evaluation of driver information systems. In: Gale AG, Brown ID, Haslegrave CM, Taylor SP, eds. Vision in Vehicles: VI. Proceedings of the 6th International Conference on Vision in Vehicles, Derby, UK, 13-16 September 1995. Amsterdam: North Holland; 1998:215-24.
45. Crundall D, Underwood G, Chapman P. Driving experience and the functional field of view. Perception 1999;28:1075–87.
46. Pauzie A, Gabaude C. Effect of dynamic central task on the usual field of view: investigation of visual and attentional abilities of elderly drivers. In: Gale AG, Brown ID, Haslegrave CM, Taylor SP, eds. Vision in Vehicles: VI. Proceedings of the 6th International Conference on Vision in Vehicles, Derby, UK, 13-16 September 1995. Amsterdam: North Holland; 1998:325-32.
47. Roge J, Pebayle T, Lambilliotte E, Spitzenstetter F, Giselbrecht D, Muzet A. Influence of age, speed and duration of monotonous driving task in traffic on the driver's useful visual field. Vision Res 2004;44:2737–44.
48. Leibowitz HW, Owens DA, Post RB. Nighttime Driving and Visual Degradation: SAE Technical Paper No. 820414. Warrendale, MI: Society of Automotive Engineers; 1982.
49. Wilkie RM, Wann JP. Driving as night falls: the contribution of retinal flow and visual direction to the control of steering. Curr Biol 2002;12:2014–7.