This is the third part of a three-part paper describing the Railway LED Lantern Test (RLLT).1,2 In the first part,1 the historical context of color vision testing on the railways was reviewed, the construction of the RLLT was described in the context of the recommendations3 of the Commission Internationale de l’Éclairage (CIE) regarding a practical test, and performance data were reported for color vision–normal (CVN) and color vision–deficient (CVD) subjects.1 In part 2, the RLLT was carried out at a 6-m working distance and compared with the CN Lantern used on the Canadian Railways.2 These two commercially available lanterns are based on railway practice, and their appropriateness as lanterns for Color Vision Standard 1 of the CIE was established. In this part, the RLLT, carried out at a 3-m working distance, will be compared with the Farnsworth Lantern (FaLant) and the OPTEC 900, which are recommended as appropriate lanterns for CIE Standard 2. The suitability of the RLLT, at a 3-m testing distance, as a lantern for CIE Standard 2 will be considered.
In addition to the CIE Color Vision Standard 1 (Normal Color Vision) referred to in part 2 of this paper,2 the CIE3 sets down two other standards of which only one is used in the recommendations for rail transport, CIE Color Vision Standard 2 (Defective Color Vision A). Under this standard, persons who have a sufficiently mild color vision deficiency will pass if they can demonstrate an ability to identify signal colors correctly. Protans, who have a very reduced ability to detect red signal lights, are excluded under this standard. The CIE advises3:
“recognition of colored signal lights and other color codes is important to safe operation, or red signal lights have to be seen, and significant social, economic or environmental costs may be associated with accidents or there is a community expectation of high standards of safety.”
CIE Color Vision Standard 2 is generally used for those who work trackside and who give or take colored signals and/or have to cross the track.
The CIE lists the following lanterns as examples of lanterns suitable for CIE Color Vision Standard 2: Holmes-Wright Type A, Farnsworth (New London), OPTEC 900 Color Vision Tester, and Beyne. Of these, the Holmes-Wright Type A and the FaLant are no longer commercially available and a supplier for the Beyne Lantern cannot be found. The OPTEC 900 is an accepted replacement for the FaLant4 although it is slightly more difficult to pass.4,5 The CIE3 set a fail criterion of two or more errors naming colors in two runs of nine pairs of colors, which is appropriate to the FaLant and OPTEC 900 but not the Holmes-Wright Type A. These lanterns will pass more subjects than the Holmes-Wright Type B, which is nominated for CIE Color Vision Standard 1.
This study compares the performance of the RLLT (presented at a 3-m test distance) with the FaLant and OPTEC 900 to confirm its appropriateness for use in testing for CIE Color Vision Standard 2.
Subjects were selected as in part 1 of this paper.1 They comprised 16 (7 male and 9 female) CVN and 37 (34 male and 3 female) CVD subjects who were not RailCorp employees and who did not have significant experience with identifying colored signals in commercial work and 30 (26 male and 4 female) CVN RailCorp employees who did have significant experience with identifying colored signals.
The procedure for color vision testing was generally the same as that described in part 2 of this paper.2 The exception was that the RLLT was administered at a 3-m viewing distance (see part 1 for the basis of this distance1) and the FaLant and OPTEC 900 were administered at 8 ft (2.4 m). The RLLT and OPTEC 900 automatically present the lights for 2 seconds. The FaLant timing is conducted manually; that is, the examiner was advised to think the words “one thousand, two thousand” in a normal speaking speed. Our experience is that this is about 2 seconds. In the case of the FaLant and OPTEC 900, the permitted color names were “red,” “green,” and “white,” and for the RLLT, “yellow” replaced “white.” The prototype RLLT used did not have the two modified red intensities referred to in part 1.1
It should also be noted that misnaming both colors of a pair is recorded as a single error on the FaLant and OPTEC 900, but counted as two errors on the RLLT.
The distribution of misnaming errors by CVD subjects is set out in Fig. 1. Note that the maximum possible misnamings for the FaLant and OPTEC 900 is 36 (the test is run twice), whereas in the RLLT, it is 48. In general, the three lanterns appear qualitatively to perform in essentially the same way. Table 1 shows a large difference in the misnamings of colors in the RLLT compared with the FaLant and OPTEC 900, but this is related to the colors used. By far, the highest proportion of FaLant/OPTEC 900 errors is misnaming white. Yellow is, similarly, the color most commonly misnamed in the RLLT, but the misnaming rate is almost half that of white. Because the color difference between red and green is about twice that between white/yellow (depending on lantern) and red and green, it is predictable that misnamings will involve white or yellow more often than red or green, assuming the luminous intensity clue has been removed or minimized.
Table 2 shows the correlation between lanterns for the numbers of errors made by the CVD subjects alone. For clarity, the confidence limits are not included in Tables 1 and 2. Where differences are referred to in the text, they are based on the binomial 95th percentile (one tail) confidence limits6 for the number of subjects in that category and the number of lights of that color. The relationship between errors on the FaLant and OPTEC 900, being nominally the same lantern design, is shown in Fig. 2.
For the FaLant and OPTEC 900, the pass criterion is less than or equal to two errors in two runs. Misnaming either or both of a pair counts as one error in determining a pass or fail. The pass criterion adopted for the RLLT was less than or equal to two misnamings, less than or equal to one green or yellow missed, and less than or equal to one blank named as a color. Misnamings were not grouped into pairs. These RLLT requirements represent the 95th percentile (one tail)6 of CVN subjects’ performance with a little leniency at 3 m so that the pass/fail criterion is, for simplicity of use, the same as that for a 6-m test distance.1 In the commercially available version of the RLLT, any missed reds constitute a fail, given the significance of this error in railway safety. The only CVN subjects who failed the RLLT at 3 m did so because they missed seeing one or two red lights, not because they misnamed any colors. Given the modification of two of the red intensities in the subsequent study in part 1,1 these two CVN subjects are not counted as failing for the analysis in this paper. Tables 3 and 4 show the comparative pass/fail performance for CVD subjects and CVN subjects on the lanterns as a pairwise comparison and as a number failed (out of the three), respectively. Because the function of these tests is to assess suitability of CVD subjects for the job, data for CVN subjects have not been included. Overall, the RLLT failed 67.6% (confidence interval [CI], 52.8 to 80.1%) of CVD subjects, the FaLant failed 73.0% (CI, 58.5 to 84.5%), and the OPTEC 900 failed 78.4% (CI, 64.4 to 88.8%). The measures of agreement, as indicated by the Cohen kappa,7 are shown in Table 5 with the rating in parentheses.8
It should be borne in mind that the RLLT is the only lantern of the three used in this study that is based in actual railway practice. The FaLant and OPTEC 900 use white not yellow as a color where railway fixed signaling uses yellow and the chromaticities of the red and green are not representative of any known railway trackside practice. The main interest in the FaLant and the OPTEC 900 is because they have been used in railway vision standards9 and are nominated by the CIE3 as suitable. To a great extent, any similarity of results from the FaLant and OPTEC 900 with the RLLT may simply be fortuitous and no more than retrospective justification of their use.
It must also be borne in mind that Farnsworth did not select signal colors for the three lights but aligned the three on the mean red-green confusion line.10 To a CVN subject, the white is visibly yellow, the red is pink, and the green is a yellow-green that would not be recommended in a signaling system.12 Farnsworth’s specification is, however, not necessarily followed in practice.5,11 This was also true in another study but the departure from specification was, in that case, with the red color.5 This is in contrast to the RLLT, which displays colors that are those used in at least one railway jurisdiction and which comply with the CIE recommendations.12
The FaLant and OPTEC 900 are designed to be the same; hence, it is expected that the results would be highly correlated. The correlation coefficient for errors made by the CVD subjects was 0.77 (Table 2) but increased to 0.90 when the CVN subjects were included. The line of best-fit intercept (Fig. 2) shows that, on average, a CVD subject making zero errors on the FaLant makes 1.5 errors on the OPTEC 900. Fig. 2 also shows that 8 CVD (21%; CI, 11.2 to 35.6%) subjects made more errors on the FaLant (points below the solid line) and 14 (37.8%; CI, 24.5 to 52.7%) made more errors (points above the line) on the OPTEC 900. This is also reflected in the slightly higher fail rate (Table 3) of the OPTEC 900. The increased difficulty of the OPTEC 900, compared with the nominally identical FaLant, found in this study is consistent with the comparisons previously reported.4,5 The number of errors on the RLLT does not correlate very highly with the FaLant and OPTEC 900 (Table 2), although it is higher if the CVN subjects are included. The level of agreement (Cohen kappa, Table 5) of the RLLT with the FaLant is only “fair” and that with the OPTEC 900 is “moderate” for CVD subjects. Table II of a previous comparison of the FaLant and OPTEC5 can be used to show κ = 0.68, which is comparable with the value in this study, κ = 0.71 (Table 5). The indifferent correlation and agreement between the RLLT and the FaLant/OPTEC 900 is not surprising given the different colors in use. This should be viewed as a failure of the FaLant and OPTEC 900 to represent the railway color vision task adequately rather than a failing in the performance of the RLLT.
It should also be noted that when lanterns based on railway signaling practices are compared, although the practices are different in detail, the agreement is “almost perfect” (κ = 0.81).2
Whereas the numbers and nature of the errors vary between the tests, the performance in accepting and rejecting applicants for a job is the important occupational issue. Tables 3 and 4 show that the RLLT is the easiest for CVD subjects to pass, but this is anticipated because the color differences are greater in the RLLT and the green light does not lie on a confusion line with the red or white, as is the case for the FaLant and OPTEC 900.10,11
The failure rates in this study of 73.0 and 78.4%, respectively, are consistent with previous studies, 76 and 79%5 and 81 and 83%4 (corrected for sample demographics5), and are within the confidence limits of the current data. The values reported by the CIE3 (66 and 70%, respectively) are lower than these other studies. Variations could be explained by the sample population, as these data are dependent on the distribution of the extent of CVD in the sample and that, in turn, depends on how they are selected. The two lanterns reported as acceptable by the CIE, but not studied in this paper, have similar (77% for the Holmes-Wright Type A Lantern) and lower (60% for the Beyne Lantern) CVD fail rates. The data reported in this paper indicate that the RLLT performs within the span of lanterns endorsed by the CIE3 and is an appropriate addition to the list of suggested lantern tests for CIE Color Vision Standard 2. It is the only lantern for CIE Standard 2 explicitly modeled on a railway practice and has an innate legitimacy that the FaLant and OPTEC 900 lack.
The RLLT performance is consistent with the range of pass/fail performances of the lanterns recommended by the CIE3 for Color Vision Standard 2. The relative similarity of pass/fail performance of the RLLT with the FaLant and OPTEC 900 indicates that personnel who have passed the FaLant or OPTEC 900 should still be considered as meeting the requirements of the standard. Given the direct link to actual railway signaling practice, the RLLT has a legitimacy that the other recommended lanterns do not have because they do not, directly, represent railway practice. The ability to support employment decisions when challenged is improved by the use of a lantern that simulates the actual visual task.13
This study was supported, in part, by a grant from RailCorp NSW (now Sydney Trains). The assistance of Clair Taylor in data collection is appreciated.
The RLLT is available from ART Electronics (www.rllt.com.au). Sydney Trains own the rights to the RLLT. The OPTEC 900 is available from Stereo Optical (http://www.stereooptical.com/products/vision-testers/optec-900).
Received January 31, 2014; accepted August 19, 2014.
1. Dain SJ, Casolin A, Long J, Hilmi MR. Color vision
and the railways: Part 1. The Railway LED lantern test. Optom Vis Sci 2015; 92: 138–46.
2. Dain SJ, Casolin A, Long J. Color vision
and the railways: Part 2. Comparison of the CN Lantern used on the Canadian Railways and Railway LED lantern tests
. Optom Vis Sci 2015; 92: 147–51.
3. Commission Internationale de l’Éclairage. International Recommendations for Colour Vision Requirements for Transport
. Vienna, Austria: Commission Internationale de l’Éclairage; 2001.
4. Laxar KV, Wagner SL, Cotton TC. Evaluation of the Stereo Optical Co. Farnsworth Lantern (FALANT) color perception test: a specification and performance comparison with the original FALANT. Groton, CT: Naval Submarine Medical Research Laboratory; 1998.
5. Cole BL, Lian KY, Lakkis C. Color vision
assessment: fail rates of two versions of the Farnsworth lantern test. Aviat Space Environ Med 2006; 77: 624–30.
6. Clopper CJ, Pearson ES. The use of confidence or fiducial limits illustrated in the case of the binomial. Biometrika 1934; 26: 404–13.
7. Carletta J. Assessing agreement on classification tasks: the kappa statistic. Comput Linguist 1996; 22: 249–54.
8. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: 159–74.
9. National Transport
Commission. National Standard for Health Assessment of Rail Safety Workers. Melbourne, Australia: National Transport
10. Farnsworth D, Foreman P. Development and Trial of New London Navy Lantern as a Selection Test for Serviceable Color Vision
, Report No. 2. Groton, CT: Naval Submarine Medical Research Laboratory; 1946.
11. Dain SJ. The Farnsworth Flashlight is not equivalent to the Farnsworth Lantern. J Opt Soc Am (A) 2012; 29: A377.
12. Commission Internationale de l’Éclairage. Colours of signal lights. Vienna, Austria: CIE; 2001.
13. Raffaelli F. Abberton v Rail Corporation of New South Wales (U2008/4982). Sydney, Australia: Australian Industrial Relations Commission; 2008.
Keywords:© 2015 American Academy of Optometry
accident prevention; visual function; vision standard; medical standard; color vision; color-vision loss; color vision standards; railroad worker; transport; occupational risk; safety standard; lantern tests; practical test