Vous êtes sur la page 1sur 9

Trafc Signal Color Recognition Is a Problem for Both Protan and Deutan Color-Vision Decients

David A. Atchison and Carol A. Pedersen, Queensland University of Technology, Brisbane, Australia, Stephen J. Dain, University of New South Wales, Sydney, Australia, and Joanne M. Wood, Queensland University of Technology, Brisbane, Australia
We investigated the effect of color-vision deciency on reaction times and accuracy of identication of trafc light signals. Participants were 20 color-normal and 49 color-decient males, the latter divided into subgroups of different severity and type. Participants performed a tracking task. At random intervals, stimuli simulating standard trafc light signals were presented against a white background at 5 to right or left. Participants identied stimulus color (red/yellow/green) by pressing an appropriate response button. Mean response times for color normals were 525, 410, and 450 ms for red, yellow, and green lights, respectively. For color decients, response times to red lights increased with increase in severity of color deciency, with deutans performing worse than protans of similar severity: response times of deuteranopes and protanopes were 53% and 35% longer than those of color normals. A similar pattern occurred for yellow lights, with deuteranopes and protanopes having increased response times of 85% and 53%, respectively. For green lights, response times of all groups were similar. Error rates showed patterns similar to those of response times. Contrary to previous studies, deutans performed much worse than protans of similar severity. Actual or potential applications of this research include trafc signal design and driver licensing. INTRODUCTION Approximately 8% of males and 0.5% of females in the population have congenital redgreen color-decient vision. There is little true color blindness (~0.0005%; Pokorny, Smith, Verriest, & Pinckers, 1979). These people have reduced ability to discriminate redness-greenness throughout the full gamut of colors. Most significantly, from a safety point of view, the problem includes the red, orange, yellow, and yellowgreen parts of the visible spectrum. Red-green color-vision deciencies are subdivided into a number of categories (after von Kries, 1899): People with dichromasy lack one of the three normal receptors. As a consequence, their ability to discriminate colors is two- rather than threedimensional. The two forms are protanopia and deuteranopia, lacking the long-wavelength (red) receptor or the middle-wavelength (green) receptor, respectively. Their ability to discriminate a red, yellow, and yellow green signal code on the basis of color is absent, and they must rely on the usual brightness hierarchy that yellow is brighter than yellow green, which is brighter than red. The use of a bluish green rather than yellowish green signal solves this problem for them. In addition, in protanopia, red signals are seen as substantially darker and are less alerting. A red traffic signal as seen in protanopia has approximately 25% the luminous intensity as it has for a color normal (Dain & King-Smith, 1981). Protanopia and deuteranopia each constitute about 1% of males. People with anomalous trichromasy have one receptor altered as compared with a color normal. As a consequence, their ability to discriminate colors is reduced rather than absent in the third dimension. The two forms are

Address correspondence to David A. Atchison, School of Optometry, Queensland University of Technology, Victoria Park Rd., Kelvin Grove Q 4059 Australia; d.atchison@qut.edu.au. HUMAN FACTORS, Vol. 45, No. 3, Fall 2003, pp. 495503. Copyright 2003, Human Factors and Ergonomics Society. All rights reserved.

496 protanomaly and deuteranomaly, in which there is an altered long-wavelength (red) receptor or middle-wavelength (green) receptor, respectively. Their ability to discriminate a red, yellow, and yellow green signal code on the basis of color is present but reduced. The loss of discrimination varies from person to person, ranging from close to normal to close to dichromatic. Those most mildly affected may not realize they have a problem until their color vision is tested, whereas those more affected realize this relatively early in life, as do the dichromats. The more affected they are, the more they must rely on brightness clues, as do dichromats. In protanomaly the problem with red signals is also present. Protanomaly and deuteranomaly constitute about 1% and 5% of the male population, respectively. Deuteranopia and deuteranomaly are collectively referred to as deutan color deficiencies, and protanopia and protanomaly are collectively referred to as protan color deciencies. Congenital blue-yellow color-vision deciencies are very rare (Wright, 1952). Acquired colorvision defects may occur in up to a further 15% of both males and females, with age being the major cause, but alcohol, drugs and disease are also factors (Whillans & Allen, 1992). As many aspects of driving depend in part on detection and recognition of colored objects, it is of considerable importance to determine which color deciencies adversely affect driving performance. Deficient color vision is rarely made a bar to general driving, but there are color-vision standards for professional drivers (e.g., drivers of buses, taxis, and heavy commercial vehicles) in several developed countries (Charman, 1985). Most of the emphasis has been placed on designing and selecting signs and signals for good recognition and visibility. For example, trafc light colors are chosen so that they have optimal recognition by color decients. However, no restrictions are placed on vehicle color other than for emergency service vehicles. Laboratory and field trials indicate conclusively that congenital color decients make more color confusions than do normal controls (see review by Vingrys & Cole, 1988). In addition, congenital color decients, particularly protans, have signicantly longer reaction times to red

Fall 2003 Human Factors signals than do color normals, both in the eld (e.g., Pokorny et al., 1979) and in the laboratory (e.g., Cole & Brown, 1966; Nathan, Henry, & Cole, 1964; Pun, Brown, & Lui, 1986). The relationship between color deciency and crash rates is not well dened because persons with color deficiencies are not recorded on crash records. Most studies of the issue of color vision in crashes have limitations: usually involving low numbers, drivers sex not being given (Hager, 1963; Sachsenweger & Nothaass, 1961), or not being applicable to the population of drivers (Norman, 1960). There is some weak evidence that protans have higher road crash rates than do color normals (Hager, 1963; Verriest, Neubauer, Marre, & Uvijls, 1980). This has been the subject of some unresolved debate (Cole, 2002a, 2002b, 2002c; Vingrys, 2002a, 2002b; Wolfe, 2002). Twenty percent of anomalous trichromats and 50% of dichromats admit to difculty recognizing signal colors, and approximately 14% of protans admit to difculty seeing red signal lights (Cole & Maddocks, 1997; Steward & Cole, 1989). The aim of this laboratory study was to assess the extent to which performance might be affected by the type and severity of color deficiency, using trafc signals that comply with a national standard (AS/NZS2144:2002: Standards Australia International, 2002) and Commission Internationale de lEclairage (CIE) guidelines (CIE, 1977, 1980). The investigation reported here formed part of a larger study investigating the effect of colored sunglasses on the ability of color-vision decient individuals to recognize trafc signal colors. Participants wore six pairs of glasses, one of which contained untinted lenses. As differences between various types and severity of color-vision deciency were more compelling in terms of performance than was the inuence of sunglasses, we report the results for the untinted lenses here. METHOD The participants were 69 young (1635 years) males who were free of systemic and ocular diseases and were not taking any medication that could affect color vision or driving performance. They consisted of 20 control (color-normal)

TRAFFIC LIGHTS AND COLOR VISION participants and 49 color-vision decient participants, divided into 25 deutans and 24 protans selected according to the criteria shown in Table 1. All participants had binocular visual acuity of 6/6 or better, but to achieve this 11 participants wore (untinted) ophthalmic corrections. The experimental setup is shown in Figure 1. The participant viewed a xation target placed in the center of a computer monitor at a distance of 4 m. Simulated single-aspect traffic signals were briey displayed at small angles to either side of the direction of xation. The participants task was to identify the color as quickly as possible. The single light presentation was adopted because the positional clues in a laboratorybased three-aspect lantern would be consistent and signicant. In the real driving task the positional clue is not always available. The two lights were located 5 away from the participants line of sight (10 apart). The angular subtense of the lights was equivalent to that of a 200-mm trafc signal lantern at a distance of 100 m (2 mrad or 6.9 arcmin). This is the standard Australian practice as detailed by Fisher and Cole (1974) and AS/NZS2144:2002 (Standards Australia International, 2002). The same specications appear in CIE publications (CIE, 1977, 1980). The lights were created using 20-W, 12-V tungsten halogen globes with appropriate lters to provide the appropriate trafc signal chromaticity coordinates specied in AS/NZS
TABLE 1: Color Vision Decient Groups Extent Mild anomalous trichromats Moderate anomalous trichromats Strong anomalous trichromats Dichromats Selection Criteria Pass Farnsworth lantern, pass Farnsworth-Munsell Panel D-15 Fail Farnsworth lantern, pass Farnsworth-Munsell Panel D-15 Fail Farnsworth-Munsell Panel D-15 (but not dichromats or extreme anomalous trichromatsa) Match whole red-green range on Nagel anomaloscope even after adaptation on Trendelenberg plate Deutan 5 5 5 100

497 2144:2002, and the intensity was controlled using neutral density filters. The chromaticity coordinates shown in Figure 2 were calculated according to CIE (1986). The lights were turned on for a maximum duration of 5 s. Rise time to full intensity was approximately 200 ms. The backgrounds to these lights were black backboards scaled in accord with the backgrounds around normal trafc lights (Fisher & Cole, 1974; Standards Australia International, 2002). Between the two lights was a computer monitor. Around the monitor and the black backboards was a white matte board. Two uorescent light tubes illuminated this so that it provided a luminance of 300 cd/m2 for the participant. The participant was shielded from a view of the light tubes. Trafc signals may operate over a range of intensity levels. The minimum and maximum for both red and green are 200 cd and 1000 cd, respectively, and the minimum for yellow is 600 cd. We presented lights of low and high intensity. The low-intensity lights were 0.32 cd for red and green and 0.96 cd for yellow, which are the 4-m equivalent of the 200-mm trafc signal at 100 m, complying with the minimum AS/NZS2144:2002 requirements. The high-intensity lights were twice the minimum requirement: 0.64 cd for red and green and 1.92 cd for yellow. The experiment was divided into three sections: button reaction time, practice, and the experiment proper.

Protan 5 5 5 9

Note. The Farnsworth lantern contains nine pairs of colored lights. Colors involved are green, red, and white. A pass is 2 or fewer identication errors on two runs. The Farnsworth-Munsell Panel D-15 test involves arranging 15 caps in order of color; color decients of sufcient severity make particular types of arrangement errors. The Nagel anomaloscope requires participants to match various red-green light mixtures with a yellow light. Extreme anomalous trichromats are dened on three criteria. They accept matches at one extreme of the Nagel anomaloscope range, they accept the normal match, and they demonstrate high variability in the apparent extent of their deciency. This has been described as tuning of their range after neutral adaptation (the Trendelenberg plate on most anomaloscopes; Pokorny et al., 1979). They were excluded from this study.


Fall 2003 Human Factors

Figure 1. Experimental setup (not to scale).

The rst section involved measuring reaction times for each nger on a three-button mouse that was adapted to measure responses in the experiment. This was necessary in order to take

Figure 2. Chromaticity coordinates of light source with different lters and the chromaticity coordinate range of trafc lights according to AS/NZS 2144:2002.

into account the different dexterities among participants and among fingers of any participant. If a participant was right handed, his index nger was on the left mouse button, his middle nger on the middle button, and his ring nger on the right button. If a participant was left handed, this sequence was reversed. Each participant was instructed to rest his hand comfortably on the mouse with the fingers in the correct position. To illustrate this, a drawing of a mouse with three buttons was presented on the computer screen and, randomly, one of the drawn buttons was highlighted. The participant pressed the button to which it corresponded. The participant was reminded of the need for both speed and accuracy. Two runs of 5 presentations of each button were made (i.e., 10 presentations for each button). Reaction times were averaged after discarding mistakes. For data analysis, response times for a button were adjusted by subtracting the reaction times for that button. The second section consisted of six practice runs of the experiment. Subsequent analysis showed that all participants achieved their peak response by the fourth run.

TRAFFIC LIGHTS AND COLOR VISION In the experiment we simulated driving using a divided-attention task (Moskowitz, 1974). The xation target was a 1.5-cm diameter circle on the computer monitor. The participant was asked to place this inside a 1.5 2 cm rectangle, which moved in straight lines at random speeds and directions, by moving the computer mouse held in the preferred hand. For the periods in which the participant succeeded in this task, he received feedback by the circle changing into a cross. At random intervals of between 6 and 12 s, one of the lights, either to the left or the right, was turned on. The participant was required to abandon the tracking task, identify the color as quickly as possible, and indicate this by pressing one of three buttons of the computer mouse: left button for red, middle button for yellow, and right button for green. This coding was the same regardless of which hand was used. The participant was given no feedback about which lights were correctly or incorrectly identified. Failure to respond within 3 s was regarded as a failure of detection. After the response, or upon failure to respond within 5 s, the next sequence started. The participant was instructed to immediately inform the experimenter if he made a mistake in responding to a light. For example, if he knew a light was green but had pressed the red button accidentally, he told the experimenter, and this was noted in the appropriate location in the sequence on a manual recording sheet. This was later correlated with the computers record of responses, and these mistakes were not used in any analysis (the mean mistake rate per sunglass was 1 to 4 across the participants). Presentation of targets and recording of responses was under computer control. A 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, so each color was presented 16 times. A short break was given between runs. The experiment reported here, of trafc signal recognition with an untinted lens, was randomized within a set of experiments with the same procedure but using tinted lenses. The experiments using the tinted lenses will be reported elsewhere. Approximately 2 hr of experimental

499 time were required for each participant, including rest breaks. Analysis of Results Separate analyses were done for mean adjusted response times and for response accuracy. The participants were divided into five groups: normals, deuteranomals, deuteranopes, protanomals, and protanopes. For the response times, assumption of normality was assessed by the Shapiro-Wilk test for each of the nine participant subgroups (four additional subgroups were created by placing anomalous trichromats in severity categories) and found to hold for all signals. Parametric analyses of variance (ANOVAs) were then used. If the signal color was called incorrectly, its time was still included in the response time analysis, as we found that the mean times for incorrect and correct responses were not significantly different when using paired sample t tests. For the error analysis, assumption of normality was again tested. Although some participant group-signal color combinations having few errors gave nonnormal distributions, we proceeded with parametric ANOVAs. For all ANOVAs giving signicant results, post hoc pair-wise comparisons were made with the Bonferroni adjustment for multiple comparisons. A 5% level of signicance was used for all tests. RESULTS Response Times Figure 3 shows the mean adjusted response times for each of the ve participant groups for the red (R), yellow (Y), and green (G) signals. Figure 4 is a more detailed analysis in which the deuteranomals and protanomals are divided into mild, moderate, and severe subgroups. A mixed ANOVA with one within-subjects factor (signal color) and one between-subjects factor (color deciency) demonstrated that the main effects of signal color, F(2, 128) = 64.21, p < .001, and color-deciency group, F(4, 64) = 9.45, p < .001, were signicant, and there was also a significant interaction between signal color and group, F(8, 128) = 5.36, p < .001. Accordingly, we performed a series of oneway ANOVAs for each signal color. There were signicant between-group variations for R and Y


Fall 2003 Human Factors signals, F(4, 64) = 9.816, p < .001, and F(4, 64) = 11.561, p < .001, respectively, but not for G signals, F(4, 64) = 1.898, p = .122. Post hoc analysis for R signals shows that normals have signicantly shorter response times than do all color-decient groups except for protanomals, and that protanomals have signicantly shorter reaction times than do deuteranopes. A similar pattern occurs with Y signals, except that both deuteranomals and protanomals have signicantly shorter response times than do deuteranopes. No statistical analysis was done using the deuteranomalous and protanomalous subgroups because of the small number (5) in each subgroup, but a clear trend of increasing response time with increasing severity is apparent for the deuteranomalous participants for both R and Y signals (Figure 4). Repeated group ANOVAs for each participant group were used to compare the response times to the different signal colors. For each participant group, there were signicant differences among the response times (p .002). For each group except for the normal group, shorter response times occurred for G signals than for Y signals. For each group, shorter response times occurred for Y signals than for R signals. The differences are statistically significant, except for normal participants between G and Y signals, for deuteranomals between G and Y signals, for deuteranopes between R and Y signals, for protanomals for Y signals with both R and G signals, and for protanopes for Y signals with both R and G signals. The main features of Figures 3 and 4 indicate the following:
1. The response times of color decients increased compared with those for the color normals for both the R and Y signals. 2. Deutans performed worse than protans (of the same severity) for R and Y signals. The mild and moderate protanomals performed a little worse than the normals. The deuteranopes showed the greatest increases in response time, 53% for R and 85% for Y, relative to those of normals. However, protanopes showed increases in response times of 35% for R and 53% for Y, relative to those of normals. 3. Response times to G signals were not affected signicantly by category of defect.

Figure 3. Mean adjusted response times as a function of color-vision deficiency groups for R, Y, and G signals. N = normals, DA = deuteranomals, D = deuteranopes, PA = protanomals, P = protanopes. Error bars indicate standard errors.

Figure 4. Mean adjusted response times as a function of color-vision deficiency subgroups for R, Y, and G signals. Error bars have been omitted for the sake of clarity. N = normals, DA = deuteranomals, D = deuteranopes, PA = protanomals, P = protanopes.

The purpose of using the two intensity levels for each signal color was to represent the

TRAFFIC LIGHTS AND COLOR VISION variation of intensities of trafc signals and, in the context of the experiment, to minimize participants learning to use intensity as a cue to color. However, we did some analysis of how luminance affected response times. Overall, the color defectives were quicker for the dimmer than for the brighter R signal. All subgroups (including normals) were quicker for the brighter than for the dimmer Y signal, with the difference increasing with severity and being greater for deutans than for protans of the same severity (except dichromats). There were no notable differences between the response times for bright and dim G signals for any of the subgroups. Errors Figure 5 shows mean percentage errors for each of the ve participant groups for the R, Y, and G signals. Figure 6 shows a more detailed analysis, in which both the deuteranomals and protanomals are divided into mild, moderate, and severe subgroups. Again, a mixed ANOVA with one withinsubject factor (signal color) and one betweensubjects factor (color deciency) demonstrated that the main effects of signal color, F(2, 128) = 17.38, p < .001, and color-deficiency group, F(4, 64) = 40.34, p < .001, were signicant, and


Figure 6. Mean errors (%) as a function of color-vision deciency subgroups for R, Y and G signals. Error bars have been omitted for the sake of clarity. N = normals, DA = deuteranomals, D = deuteranopes, PA = protanomals, P = protanopes.

Figure 5. Mean errors (%) as a function of color-vision deciency groups for R, Y, and G signals. Error bars indicate standard errors. N = normals, DA = deuteranomals, D = deuteranopes, PA = protanomals, P = protanopes.

there was also a signicant interaction between signal color and group, F(8, 128) = 5.39, p < .001. Accordingly, we performed one-way ANOVAs for each signal color. There are significant between-subject group variations for R, F(4, 64) = 15.876, p < .001, and Y signals, F(4, 64) = 9.343, p < .001. No errors occur for the G signal for any participants. The errors for R and Y signals show a trend similar to that for response times in Figures 3 and 4. Generally, errors increased as the severity of the color vision deficiency became more marked. Except for the mild subcategory, deutans, overall, performed worse than protans of the same severity category (Figure 6). Protanomals actually performed about as well as normals. Deuteranopes made signicantly more errors than did any other group (p .001), with 30% and 23% errors for R and Y signals, respectively. For all participant groups, the main errors were in calling a R signal as Y and vice versa. This is demonstrated for deuteranopes in Figure 7.


Fall 2003 Human Factors had increased response times of 48% and 86%, respectively. It is possible to partially reconcile the differences in our results with those of previous studies. We used single R, Y, and G colors, rather than the several variations used in the Nathan et al. (1964) study, which made the task simpler for the color decients in our study. Our ndings are more in line with those of Cole and Brown (1966), who found that the response time of protanopes to an R signal was never more than 50% greater than that for normals at any luminance as they used only an R signal, there was no time lost in making decisions about the correct color. No errors were made to the G signal in our study, which indicates that this green chromaticity gamut is well chosen. Judging by the low missed-signals rate (e.g., Figure 7), our signals are well above threshold for all color decients, and under these circumstances protans are no longer at a disadvantage relative to deutans. Protans perform worse when lights are closer to threshold, such as in experiments measuring the closest distance at which a signal can be recognized (Verriest et al., 1980; Cole & Vingrys, 1983). With highly visible signals, protans may be able to use intensity cues to distinguish between colors. Having two levels may have reduced intensity cues in our study, but the factor of two between them may not have been enough to make much difference. Also, some differences could be attributable to modern familiarity with a blue green rather than yellow green signal color. In addition, any study that looks at detection rather than recognition as the participants task will be more likely to demonstrate the loss of red sensitivity of the protan. The difficulties in distinguishing red and yellow occurred despite the yellow signal having three times the intensity of the red signal. Given that the higher intensities required in yellow signals are less easily attainable in signals composed of LEDs, there may be pressures to reduce the required intensities, but the present study should be a discouragement to that move. It should also be noted that modern practices using red LEDs will exacerbate the problems of detection of red signals, because the LED signals are redder than the conventional signal.

Figure 7. Mean errors (%) as a function of signal color for deuteranopes.

Across all participant subgroups, most R signal errors were made with the brighter signal and most Y signal errors were made with the dimmer signal. This is consistent with differences in response time to the two luminance levels of these signals. DISCUSSION This investigation agrees with earlier studies in that color deficients have longer response times and make more mistakes than do color normals when responding to signals. However, contrary to previous investigations (Nathan et al., 1964; Pun et al., 1986; Cole & Vingrys, 1983), we have found that deutans perform worse than protans. Furthermore, our increases in response times and error rates of color decients are less than those found in the Nathan et al. study. For example, our deuteranopes showed the greatest increases in response time 53% for R signals and 85% for Y signals, relative to those of normals whereas protanopes showed increases in response times of 35% for R signals and 53% for Y signals. In the Nathan et al. study, the respective response time increases were 88%, 95%, 105%, and 118%. Another difference between this and previous work is that we found no effect of color-vision deciency on response times to G signals, whereas Nathan et al. found that deuteranopes and protanopes

TRAFFIC LIGHTS AND COLOR VISION Given the multifactorial nature of road crash causation and the failure to collect the necessary driver data, it is not surprising that few data linking color-vision deciencies are available. Despite this, a link with protan color-vision deciencies has yet to be eliminated as an issue (Cole, 2002a, 2002b; Vingrys, 2002b; Wolfe, 2002). At present, the only color-vision requirements in driving usually relate to protans (given their demonstrated problem in detecting red signals) in high-exposure jobs where the consequence of error may be greater (heavy goods vehicles, taxis, and buses). The results of the study reported here indicate that deutans in particular, deuteranopes also merit close scrutiny in driving performance. ACKNOWLEDGMENTS A Queensland University of Technology Research Encouragement Award supported this study. Eddie Matejowsky wrote the computer program and built the electronic circuitry for the experiment. Nancy Spencer provided statistical advice. REFERENCES
Charman, W. N. (1985). Visual standards for driving. Ophthalmic and Physiological Optics, 5, 211220. Commission Internationale de lEclairage. (1977). Road lighting lantern and installation data Photometrics, classication and performance (CIE Publication 34). Vienna: Author. Commission Internationale de lEclairage. (1980). Light signals for road trafc control (CIE Publication 48). Vienna: Author. Commission Internationale de lEclairage. (1986). Colorimetry (CIE Publication 15.2). Vienna: Author. Cole, B. L. (2002a). Protan colour vision deciency and road accidents. Clinical and Experimental Optometry, 85, 246253. Cole, B. L. (2002b). Protans and driving safety. Clinical and Experimental Optometry, 85, 401402. Cole, B. L. (2002c). Whos responsible for safe vision on the roads? Clinical and Experimental Optometry, 85, 207209. Cole, B. L., & Brown, B. (1966). Optimum intensity of red road traffic signal lights for normal and protanopic observers. Journal of the Optical Society of America, 56, 516522. Cole, B. L., & Maddocks, J. D. (1997). Defective colour vision is a risk factor in driving. In C. R. Cavonius (Ed.), Colour vision deficiencies XIII (pp. 471481). Dordrecht, Netherlands: Kluwer Academic. Cole, B. L., & Vingrys, A. J. (1983). Do protanomals have difculty seeing red lights? In Proceedings of the 20th Session of the CIE (CIE Publication 56, E04, pp. 13). Paris: Commission Internationale de lEclairage. Dain, S. J., & King-Smith, P. E. (1981). Visual thresholds in dichromats and normals: The importance of post-receptoral processes. Vision Research, 21, 573580. Fisher, A. J., & Cole, B. L. (1974). The photometric requirements of vehicular traffic signal lanterns. In Proceedings of the Seventh Conference of the Australian Road Research Board (pp. 246265). Melbourne: Australian Road Research Board.

Hager, G. (1963). Das Sehorgan und das Unfallgeschehen im Strassenverkehr [The eye and accident occurrences in street traffic]. Klinische Monatsbltter fr Augenheilkunde, 142, 427455. Moskowitz, H. (1974). Effects of alcohol on peripheral vision as a function of attention. Human Factors, 16, 174180. Nathan, J., Henry, G. H., & Cole, B. L. (1964). Recognition of colored road trafc light signals by normal and color-vision-defective observers. Journal of the Optical Society of America, 54, 10411045. Norman, L. G. (1960). Medical aspects of road safety. Lancet, 276, 1039-1045. Pokorny, J., Smith, V. C., Verriest, G., & Pinckers, A. J. L. G. (1979). Congenital and acquired color vision defects. New York: Grune & Stratton. Pun, H. W., Brown, B., & Lui, R. (1986). Tinted contact lenses slow reaction time in colour defective observers. Clinical and Experimental Optometry, 69, 213218. Sachsenweger, R., & Nothaass, E. (1961). Eine Analyse von 4011 Verkehrsunfllen aus augenrzlicher Sicht [An analysis of 4011 traffic accidents from an ophthalmologists point of view]. Deutsche Gesundheitwesen, 12, 868872. Standards Australia International. (2002). Traffic signal lanterns (AS/NZS 2144:2002). Sydney: Author. Steward, J. M., & Cole, B. L. (1989). What do color vision defectives say about everyday tasks? Optometry and Vision Science, 66, 288295. Verriest, G., Neubauer, O., Marre, M., & Uvijls, A. (1980). New investigations concerning the relationships between congenital colour vision defects and road traffic security. International Ophthalmology, l2, 8799. Vingrys, A. J. (2002a). The case against protan drivers holding professional driving licences. Clinical and Experimental Optometry, 85, 4648. Vingrys, A. J. (2002b). Protans and driving safety. Clinical and Experimental Optometry, 85, 400401. Vingrys, A. J., & Cole, B. L. (1988). Are colour vision standards justified for the transport industry? Ophthalmic and Physiological Optics, 8, 257274. von Kries, J. (1899). ber die anomalen trichromatisch Farbensystem [Concerning the anomalous trichromatic color system]. Zeitschrift fr Sinnesphysiologie, 19, 6369. Whillans, M. G., & Allen, M. J. (1992). Color defective drivers and safety. Optometry and Vision Science, 69, 463466. Wolfe, R. J. B. (2002). Protans and driving safety. Clinical and Experimental Optometry, 85, 399400. Wright, W. D. (1952). The characteristics of tritanopia. Journal of the Optical Society of America, 42, 509521.

David A. Atchison received his Ph.D. in optometry in 1984 from the University of Melbourne. He is an associate professor in the School of Optometry, Queensland University of Technology. Carol A. Pedersen received her BAppSc(Optom) (Hons) in optometry in 1992 from the Queensland University of Technology. She is an optometrist and senior research assistant in the School of Optometry, Queensland University of Technology. Stephen J. Dain received his Ph.D. in optometry in 1972 from City University (London). He is head of the School of Optometry and Vision Science, University of New South Wales. Joanne M. Wood received her Ph.D. in optometry in 1987 from the University of Aston. She is an associate professor in the School of Optometry, Queensland University of Technology. Date received: October 17, 2001 Date accepted: June 2, 2003

Vous aimerez peut-être aussi