People with deficient color vision have problems detecting and recognizing road traffic signals. The problems involved are increases in reaction time, relative to those of people with normal color vision, and incorrect recognition.1–7 These problems increase with severity, e.g., dichromats compared with anomalous trichromats. Nathan et al.4 found that protans performed worse than deutans, but Atchison et al.1 found the opposite effect, with the reasons for the difference explained by Cole.8 There is also some evidence that protans have higher road accident rates than color normals.7,9–11
It has long been recognized that colored sunglasses might impair recognition of signal lights, especially for drivers with defective color vision. Based on some experimental measurements in the literature, but mainly on his own theoretical determinations, Clark12,13 argued that sunglass coloration should not depart very much from neutral, and he proposed that this should be determined by signal factors. He recommended a red signal visibility factor (R) and a violet coloration factor (V) which are the ratios of the amounts of red and violet light, respectively, transmitted through a lens relative to the luminous transmittance through the lens. The former should be within certain limits and the latter should have a minimum value so that the already degraded color perception of color deficients across the visual spectrum should not be worsened by lenses that transmit little or no violet light.
Clark's ideas were taken up in the first14 and subsequent Australian sunglass standards (Table 1). At the time that this study was conducted, the then Australian sunglass standard AS 1067 to 199015 argued:
Lenses with an R < 1.0 can decrease the visibility and increase the reaction time for red signals, and color identification can also be adversely affected, especially for a person with defective color vision. Lenses with R > 1.0 can adversely affect the brightness cues used by color defectives in identifying red. Lenses which do not transmit sufficient violet light can seriously degrade color perception, especially for color defectives.
In AS 1067–1990 “general purpose” sunglasses (between 8% and 50% luminous transmittance) were required to have R and vs. >0.70 and 0.3, respectively. If they had R factors >1.40, they were to be labeled: “Not suitable for persons with defective color vision” or “Not suitable for persons with defective color vision. These lenses will further distort their color perception.” “Specific purpose sunglasses” had stricter ultraviolet absorption requirements than general purpose sunglasses and were classified as either type (a) or type (b). Type (a) had stricter R requirements than general purpose sunglasses (0.85 ≤ R ≤ 1.15), and a stricter V of ≥0.5. Type (b) has no coloration limits, but had to be labeled: “Not suitable for driving” or “Because they distort color perception, these lenses are inappropriate for driving” or, where the lenses did not meet the minimum coloration limits for the general purpose sunglasses, they had to be labeled “Not suitable for persons with defective color vision” or “Not suitable for persons with defective color vision. These lenses will further distort their color perception” if the lenses had R factors >1.40.
Sunglass standards around the world now impose coloration requirements on sunglasses to limit color distortions, especially those that might impede recognition of road traffic signals. The United States, European, and current Australian standards express their signal visibility or coloration factors differently from those of AS 1067:1990 (Table 1).15–20 The current U.S. standard ANSI Z80.3-200816 has minimum sunglass transmittances for daylight (D65 light source) and red signal lights of 8%, and minimum transmittances for green and yellow signal lights of 6%. These are absolute transmittances, whereas the other standards allow for adaptation in the visual system by specifying relative transmittances. The U.S. standard also specifies maximum color shift limits for these lights. It requires that the minimum transmittance across the wavelength range of 475 nm to 650 should be at least 20% of the luminous transmittance. The European Standard EN 1836:200517 defines “relative visual attenuation quotients” for sunglass lenses for each of blue, green, yellow, and red signal colors, each of which has different compliance values. It also requires that the minimum transmittance across the wavelength range of 500 nm to 650 should be at least 20% of the luminous transmittance. The U.S. standard is considerably more lenient than the European standard with regard to coloration, but many sunglasses sold in that country still fail its coloration limits.21 The European and U.S. Standards are substantially more lenient than AS 1067:1990 in the blue end of the spectrum.
The current Australian and New Zealand Standard on sunglasses AS/NZS 1067:200318 is technically equivalent to the European Standard,17 but the compliance requirements vary in some measures. AS/NZ 1067:2003 has a higher compliance value than EN 1836 for the blue signal color (0.7 compared with 0.4) and the range across which the minimum spectral transmittance should be compared with the luminous transmittance is extended to 450 nm (Table 1).
The research reported in this article forms part of a larger laboratory study investigating the effect of color vision deficiency on signal detection response times and on the accuracy of recognition of the signals. The aims of this component of the study were (a) to assess the extent to which sunglasses of a range of colorations impede the detection and recognition of traffic signal lights, and (b) to establish whether Australian and other standards are adequate in specifying sunglass coloration limits.
Sunglasses were developed by dyeing plano plastic CR39 lenses to achieve the appropriate tint characteristics required for the study. Each pair of lenses was measured spectrophotometrically (Varian Cary 5000 UV-VIS-NIR spectro-photometer) at the Optics and Radiometry Laboratory of the University of New South Wales. An untinted pair (clear) and a neutral density pair (gray) acted as controls. The lenses were fitted into spectacle frames that required minor adjustment to fit each observer's head comfortably. Fig. 1 shows their spectral transmittances. Table 2 lists the tints, together with the ways in which their coloration factors fall outside the limits for the General Purpose sunglass category of AS 1067:1990. The tints had luminous transmittances of 21% to 31%. Three pairs of lenses (yellow-green, yellow-brown, and red-brown) had R factors >1.4 (high), one pair (green) had an R factor <0.70 (low), and three pairs had V factors <0.3 (green, yellow-green, and yellow-brown). The tints lie within the spread of colors typically seen in the sunglasses that have been tested by Optics and Radiometry Laboratory in recent years (about 2000 pairs a year).
Table 2 indicates also where the tints were outside the specified limits for AS/NZS 1067:2003, EN 1836:2005, and ANSI Z80.3:2008. The green tint failed each standard on one or more criteria, the yellow-green tint failed the AS/NZS 1067:2003 and ANSI Z80.3:2008 standards, and the yellow-brown and red-brown lenses passed all three standards.
Observer and experimental information have been reported in detail previously.1 Observers were 69 young, healthy (16–35 years) males, consisting of 20 color normals, 15 deuteranomals, 10 deuteranopes, 15 protanomals, and 9 protanopes. Selection criteria are given in Table 3. All observers had binocular visual acuity of 6/6 or better, with 11 observers wearing their (untinted) ophthalmic corrections behind the sunglass tints to achieve this visual acuity.
Observers viewed a fixation target in the center of a computer monitor at a 4 m working distance (Fig. 2). Simulated single aspect traffic signals were displayed for a maximum of 5 s at 5° either side of fixation and observers were instructed to identify the color as quickly as possible.
Signal size was equivalent to that of 200 mm traffic signal lantern at 100 m distance (2 mrad), which is the standard Australian practice.22,23 Signals were created with 20 W 12 V tungsten halogen globes and filters to provide the appropriate traffic signal chromaticity coordinates. Intensity was controlled using neutral density filters. The chromaticity coordinates of the signals, from spectral radiance measurements made with a Topcon SR-3 telespectroradiometer, are represented in Fig. 3(a to c) by open circles. Also shown are the color requirements of the signals. The international standard on traffic signals24 references the IS0/CIE S 004 standard on colors of signal lights.25 The red signals lie within the class A1 limits which are specified “when persons with defective color vision are included in the user group.” The yellow signals lie within the permitted limits for yellow (there are no classes) and the green signals lie within the class A requirements which are for the same application as class A1 red. In other words, the colors were, in the context of CIE S 004, optimized for use by people with defective color vision. The chromaticity coordinates of the signals viewed through the tinted lenses were calculated using the CIE 2° standard observed and are represented by the other symbols in Fig. 3 (a to c).
We presented signals of low intensity—0.32 cd for red and green and 0.96 cd for yellow and high intensity—0.64 cd for red and green and 1.92 cd for yellow which are the 4 m equivalent of the 200 mm traffic signal at 100 m complying with the maximum and minimum AS/NZS 2144 requirements.23 The signals were surrounded by black backboards in scale with the backgrounds around normal traffic lights.22,23 Around the computer monitor and the black backboards was a white matt board illuminated by two fluorescent light tubes to provide 300 cd/m2 luminance.
The experiment was divided into three sections—button reaction time, practice, and the experiment proper. The reaction times for the first section were used to adjust response times for data analysis. The first two sections are described in our previous article. In the experiment proper, we simulated driving using a divided attention task. The fixation target was a 1.5 cm diameter circle which moved in straight lines at random speed and direction on the computer monitor. The observers were asked to place the fixation target inside a 1.5 × 2 cm rectangle by moving the computer mouse. They received feedback by the circle changing into a cross when they were successful. At random intervals of between 6 and 12 s, either the left or the right light was turned on. The observer abandoned the tracking task, identified the color as quickly as possible and indicated this by pressing one of three buttons on the computer mouse: left button for red, middle button for yellow, and right button for green. Failure to respond within 3 s was regarded as a detection failure. After the response (or after 5 s if no response) the next sequence began.
The observers informed the experimenter immediately if they had made a mistake in responding to a light. This was later correlated with the computer's record of responses and these “mistakes” were not used in analysis (mean ± SD = 2% ± 1%). Observers were not given feedback about which lights were correctly or incorrectly identified.
Target presentation and response recording were under computer control. Each run consisted of 12 presentations, with 1 presentation on each side of low and high luminance red, yellow, and green lights. These presentations were randomized within each run. There were four runs per signal color, so each color was presented 16 times for each sunglass tint. At the completion of a set of runs, the next sunglass was selected. The order of sunglass wear was randomized between observers, using an incomplete Latin square design.
Repeated measures analysis of variance were conducted for both response time and response accuracy with two within group factors [lens tint (six levels) and signal color (three levels)] and one between subjects group factor (normals, deuteranomals, deuteranopes, protanomals, and protanopes). In the previous article,1 we used the term “mean adjusted response time” to allow for the button reaction time, but here we simply use the term “response time.” Sphericity assumptions for some of the analyses were violated. Analyses are therefore reported as significant only if they were also significant with adjustment for sphericity via the Greenhouse-Geisser epsilon. Given the large number of conditions in the experiment, it was not considered valid to conduct all possible follow-up comparisons, so visual inspection was used as the basis for the interpretation of any significant interactions.
There were significant main effects of signal color (F2,128 = 83.72; p < 0.001) and lens tint (F5,320 = 33.35; p < 0.001) on response times and the group effect was also significant (F4,64 = 14.18; p < 0.001), where the deuteranopes had significantly longer response times than all the other groups except for the protanopes. Fig. 4 shows response times for the different color vision groups, collapsed across all signals, for each sunglass tint. It clearly shows that the response times for all color deficient groups were longer than those of the color normals for the clear and gray lenses. There were also significant two-way interactions between signal color and group (F8,128 = 8.15; p < 0.001), lens color and group (F20,320 = 3.66; p < 0.001), as well as a three-way interaction between signal color, lens color and group (F40,640 = 2.96; p < 0.001). As seen in Fig. 4, the green and yellow-green lenses exacerbated the increase in response times for all of the color deficient groups relative to the color normal group.
Given the significant three-way interaction, the data were broken down and analyzed for each of the five participant groups. Results for the normal group are shown in Fig. 5a. There was a significant main effect of signal color (F2,38 = 6.67; p = 0.003) but not lens tint (F5,15 = 0.15; p = 0.98) on response times. There was also a significant interaction effect between signal color and lens tint (F10,190 = 5.97; p = 0.045), where response times were increased when viewing the Y signal through the yellow-green lens relative to the other lens tints, and responses times were decreased when viewing the R signal through the yellow-green lens.
Results for the deuteranomals are shown in Fig. 5b. The response times were significantly affected by signal color (F2,28 = 12.0; p < 0.001) and lens tint (F5,70 = 10.08; p < 0 0.001), and there was also a significant interaction between signal color and lens tint (F10,140 = 11.61; p < 0 0.001). Examining the two-way interaction (Fig. 5b) it is evident that response times were slower to the G signal when viewed through the green lens and to the Y signal when viewed through the yellow-green lens compared with the other lens tints.
Results for the deuteranopes are shown in Fig. 5c. The response times were significantly affected by signal color (F2,18 = 21.55; p < 0.001) and lens tint (F5,45 = 3.65; p = 0.007). The observers responded slower to the R signal, across all lenses, than to either Y or G signals, and they responded slower to the Y than to the G signal. The differences were considerable at 59% (R vs. G) and 47% (Y vs. G). There was also a significant two-way interaction between signal color and sunglass lens tint (F10,90 = 3.01, p = 0.003), where response times to the G signal were longer when viewed through the green and yellow-green lenses compared with the other sunglass lens tints (Fig. 5c).
Results for the protanomals are shown in Fig. 5d. The response times were significantly affected by signal color (F2,28 = 28.89; p < 0 0.001) and lens tint (F5,70 = 9.33; p < 0 0.001), and there was also a significant interaction between signal color and lens tint (F10,140 = 7.12; p < 0 0.001). The significant interaction effect reflects the increase in response times to the G signal viewed through the green lens and to the R signal viewed through the red-brown lens.
Results for the protanopes are shown in Fig. 5e. The response times were significantly affected by signal color (F2,16 = 25.95; p < 0.001) and lens tint (F5,40 = 11.32; p < 0.001) and there was also a significant two-way interaction (F10,80 = 7.0; p < 0.001). This interaction effect reflects the increase in response times to the G and Y signals for the green and yellow-green lenses; interestingly, the effects of signal color were greatest for the red-brown lens, where responses were clearly slower to the R and Y signals compared with the G signal.
From considering Fig. 5a to e, it is clear that the response times of the color normals were less affected by either signal color or sunglass tint than were the color deficient groups. Response times for the color deficient groups were considerably slower than the color normals for both R and Y signals at all sunglass colors, but for the G signals they were only noticeably slower with the green and yellow-green lenses.
There were significant main effects of signal color (F2,128 = 42.41; p < 0.001), and sunglass tint (F5,320 = 9.58; p < 0.001) on error rates. The group effect was also significant (F4,64 = 37.33; p < 0.001), with the deuteranopes making significantly more errors than any other group, as shown in Fig. 6, which represents error rate as a function of sunglass tint and group collapsed across all signal colors. There were also significant two-way interactions between signal color and group (F8,128 = 12.99; p < 0.001), between lens color and group (F20,320 = 1.88; p = 0.013) and a three-way interaction between signal color, lens color, and group (F40,640 = 2.77; p < 0.001). Given the three-way significant interaction effect, the data were broken down and analyzed for each of the five participant groups.
Results for the normal group are shown in Fig. 7a. There were no significant effects of signal color (F2,38 = 2.38; p = 0.106) or sunglass lens tint (F5,95 = 2.14; p = 0.068) on error rates, but there was a significant two-way interaction effect (F10,190 = 4.29; p < 0.001). Only the Y signal viewed through the yellow-green lens produced higher error rates relative to the other conditions.
Results for the deuteranomals are shown in Fig. 7b. The error rates were significantly affected by signal color (F2,28 = 8.31; p = 0.001) and by lens tint (F5,70 = 5.29; p < 0.001), and there was also a significant interaction between signal color and lens tint (F10,140 = 7.18; p < 0.001). The interaction effect reflects the stronger detrimental effect of the yellow-green lens for Y signals, such that the yellow-green lens resulted in an error rate of 18%, 3.5× higher than that with the gray lens.
Results for the deuteranopes are shown in Fig. 7c. The error rates were significantly affected by signal color (F2,18 = 16.02; p < 0.001) but not lens tint (F5,45 = 1.19; p = 0.331). There was also a significant two-way interaction between the factors (F10,90 = 4.05, p < 0.001) representing a complex series of effects. The error rates for R and Y signals were 21× higher overall than that for G signals, but the pattern of error rates was quite different, with the error rates for R signals exacerbated for the yellow-brown and red-brown lenses and with the error rates for Y signals exacerbated for the green and yellow-green lenses.
Results for the protanomals are shown in Fig. 7d. The error rates were not significantly affected by either signal color (F2,28 = 2.17; p = 0.133) or lens tint (F5,70 = 2.07 p = 0.79), nor was there a significant interaction between signal color and lens tint (F10,140 = 2.403; p = 0.103).
Results for the protanopes are shown in Fig. 7e. Error rates were significantly affected by signal color (F2,16 = 4.47; p = 0.03) and lens tint (F5,40 = 3.83; p = 0.006). There was also a significant two-way interaction effect (F10,80 = 4.09, p < 0.001), such that error rates were elevated for the Y signal when viewed through the yellow-green lenses.
From considering Fig. 7a to e, it is apparent that for most of the color deficient groups there were particular problems for Y signals combined with the green and yellow-green lenses. In addition to this, deuteranopes had particular problems for R signals combined with the yellow-brown and red-brown lenses. Although the patterns were somewhat different from those seen for response times, in general combinations of signals and sunglasses of similar colors were of particular concern.
For the clear lens component of our study, we found that the individuals with a color deficiency had longer response times and made more recognition errors than color normals in response to signals simulating traffic signals.1 Deutans performed noticeably worse than protans.
In the current article, we considered the effect of different sunglass tints on response times and errors and found that the poorer performance of those with a color deficiency was exacerbated by non-neutral sunglass tints, with combinations of signals and sunglasses of similar colors being of particular concern. Two of these tints (yellow-brown and red-brown) passed the current European, U.S., and Australian standards, whereas the yellow-green tint failed the Australian standard and the U.S. standard (marginally for the latter), and only the green tint unequivocally failed all three standards (Table 2).
Anything that limits their ability to design products is generally disliked by the sunglass industry and the restriction placed on them by the color limits of sunglass standards is no exception. It is often pointed out that most analyses, especially those in the detailed articles of Clark, are theoretical. However, given the immense variability of on-road conditions and the range of color vision deficiencies that exist, it is unlikely that adequate on-road studies will ever be funded and the data collected in accident cases have never included documentation of any tinted media worn by the driver(s) involved. So “real-life” studies are not a viable option in exploring the important question of whether sunglass tints compromise the road safety of color deficient individuals; reasonable representations in the laboratory are the only feasible option.
What has been carried out in the current study is an evaluation of tints that are representative of those that are permitted or used in everyday life. The study used contemporary traffic signal design standards in the design of the stimuli and a task that represented the dual task components of driving performance. Therefore the on-road situation was replicated, as best as is possible, in the laboratory.
The study has shown that some sunglass tints, currently permitted for wear by drivers and riders, cause a measurable decrement in the ability of color deficient observers to detect and recognize traffic signals. This is prima facie evidence of a risk in the use of these lenses. What this study has not addressed is how that risk might translate into road accidents nor what might be an acceptable risk (because driving is already a risky business), nor how the magnitude of that risk might be considered relative to other avoidable or controllable factors. However, the study has illustrated that the issues that authors such as Clark have raised over many years are real.
This study was supported by a Queensland University of Technology Research Encouragement Award. Eddie Matejowsky wrote the computer program and built the electronic circuitry for the experiment, Carol Pedersen assisted with experiments, and Nancy Spencer and Phillipe Lacherez provided statistical advice. Barry Cole provided comments on a draft.
Institute of Health and Biomedical Innovation
Queensland University of Technology
60 Musk Avenue
Kelvin Grove Q 4059, Australia
e-mail: [email protected]
1.Atchison DA, Pedersen CA, Dain SJ, Wood JM. Traffic signal color recognition is a problem for both protan and deutan color-vision deficients. Hum Factors 2003;45:495–503.
2.Cole BL, Brown B. Optimum intensity of red road-traffic signal lights for normal and protanopic observers. J Opt Soc Am 1966;56:516–22.
3.Freedman M, Davit PS, Staplin LK, Breton ME. Traffic signal brightness. An examination of nightime dimming. Publication No. FHWA-RD-85-005. Washington, DC: Federal Highway Administration; 1985.
4.Nathan J, Henry GH, Cole BL. Recognition of colored road traffic light signals by normal and color-vision-defective observers. J Opt Soc Am 1964;54:1041–5.
5.Pokorny J, Smith VC, Verriest G, Pinckers AJ. Congenital and Acquired Color Vision Defects. New York: Grune & Stratton; 1979.
6.Pun HW, Brown B, Lui R. Tinted contact lenses slow reaction time in colour defective observers. Clin Exp Optom 1986;69:213–8.
7.Vingrys AJ, Cole BL. Are colour vision standards justified for the transport industry? Ophthalmic Physiol Opt 1988;8:257–74.
8.Cole BL. The handicap of abnormal colour vision. Clin Exp Optom 2004;87:258–75.
9.Hager G. Das Sehorgan und das Unfallgeschehen im Strassenverkehr. Klin Monatsbl Augenheilkd 1963;142:427–33.
10.Verriest G, Neubauer O, Marre M, Uvijls A. New investigations concerning the relationships between congenital colour vision defects and road traffic security. Int Ophthalmol 1980;2:87–99.
11.Cole BL. Protan colour vision deficiency and road accidents. Clin Exp Optom 2002;85:246–53.
12.Clark BA. Color in sunglass lenses. Am J Optom Arch Am Acad Optom 1969;46:825–40.
13.Clark BA. Coloration limits for sunglass lenses. Aust J Optom 1970;53:307–15.
14.Australian Standards Association. Australian Standard AS 1067-1971. Sunglasses
. Sydney: Australian Standards Association; 1971.
15.Australian Standards Association. Australian Standard AS 1067-1990. Sunglasses
and Fashion Spectacles. Part 1: Safety Requirements. Sydney: Australian Standards Association; 1990.
16.American National Standards Institute. Ophthalmics: Nonprescription Sunglasses
and Fashion Eyewear: Requirements: Z80.3-2001. New York: American National Standards Institute; 2001.
17.European Standards Organisation (CEN). Personal Eye-Equipment: Sunglasses
and Sunglare Filters for General Use and Filters for Direct Observation of the Sun: EN 1836:2005. Brussels: CEN; 2005.
18.Standards Australia/Standards New Zealand. Australian/New Zealand Standard Sunglasses
and Fashion Spectacles: AS/NZS 1067:2003. Sydney: Standards Australia Ltd.; 2003.
19.International Standards Organisation. CIE Standard Colorimetric Observers: ISO 10527:2007. Geneva: International Standards Organisation; 2007.
20.International Standards Organisation. CIE Standard Illuminants for Colorimetry: ISO 10526:2007. Geneva: International Standards Organisation; 2007.
21.Hovis JK, Cranton D, Chou BR. Tinted lenses and the ANSI standards for traffic signal transmittances. Optom Vis Sci 1991;68:750–5.
22.Fisher AJ, Cole BL. The photometric requirements of vehicular traffic signal lanterns. In: Proceedings of the Australian Road Research Board, vol 7. Melbourne: Australian Road Research Board; 1974:246–65.
23.Standards Australia/Standards New Zealand. Australian/New Zealand Standard Traffic Signal Lanterns AS/NZS 2144:2002. Sydney: Standards Australia Ltd.; 2002.
24.International Standards Organisation/Commission Internationale de L'Éclairage. Road Traffic Lights—Photometric Properties of 200 mm Roundel Signals. ISO/CIE 16508. Vienna: CIE; 1999.
25.Commission Internationale de L'Éclairage. Colours of Light Signals. CIE Standard S 004/E. Vienna: CIE; 2001.