See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228372545

Establishing the statistical limits of “normal” chromatic sensitivity

Article

CITATIONS READS

23 1,003

3 authors, including:

John Barbur Marisa Rodriguez-Carmona

City, University of London City, University of London
223 PUBLICATIONS 5,459 CITATIONS 53 PUBLICATIONS 818 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Marisa Rodriguez-Carmona on 02 December 2014.

The user has requested enhancement of the downloaded file.

Barbur et al., Limits of “normal“ chromatic sensitivity

ESTABLISHING THE STATISTICAL LIMITS OF “NORMAL” CHROMATIC

SENSITIVITY
John L Barbur, Marisa Rodriguez-Carmona and Alister Harlow
Applied Vision Research Centre, The Henry Wellcome Laboratories for
Vision Sciences, City University, London UK

the wavelength tuning of photopigment

ABSTRACT
genes, other factors such as differences in
Measurement of colour discrimination the optical density of cone photoreceptors or
loss is particularly important in demanding / and variation in post-receptoral
visual tasks where colour is used to amplification of cone signals can also cause
enhance visual performance. significant changes in chromatic
sensitivity(3). Colour vision is currently
Ideally, colour vision assessment assessed using a wide range of tests. In the
requires a test that (I). provides true isolation absence of an internationally recognized
of colour signals, (II). is based on data that standard procedure for examination of
describe the statistical limits of colour colour vision, clinical assessment relies on
discrimination in “normal” trichromats, (III). the use of a battery of tests that can often
has enough sensitivity to detect “minimal” produce inconsistent results for a number of
deficiencies and to classify them, and (IV). different reasons. The principal aim of this
can be used to detect and monitor study was to extend the usefulness of the
“significant changes” in colour sensitivity. CAD test(7) by assessing the variability in
We describe a new Colour Assessment red-green (r-g) and yellow-blue (y-b)
and Diagnosis (CAD) test that exploits the chromatic discrimination sensitivity within
CIE-(x,y) 1931 system and extends the work normal trichromats and colour deficient
of MacAdam(6) using spatiotemporal observers.
luminance contrast masking techniques(4;
2. METHODS
5). The CAD test has been optimised to fulfil
the requirements stated above. The CAD was implemented on a
stable CRT display and employs direction-
Keywords: colour vision, noise masking,
specific, colour-defined, moving stimuli that
chromatic sensitivity, colour deficiency
are buried in a background of random,
dynamic, luminance contrast (LC) noise.
1. INTRODUCTION The subject’s task is to press an appropriate
Novel methods developed to assess button, following each stimulus presentation,
chromatic sensitivity often yield statistically to indicate the direction of motion of the
significant, inter-subject differences that can, colour-defined stimulus. An efficient, four-
in principle, be attributed to either congenital alternative, forced-choice procedure is used
or acquired colour deficiencies(1; 2). The to measure the subject’s chromatic
high sensitivity of such techniques can be displacement thresholds in a number of
used to monitor changes in colour vision in carefully selected directions in the CIE –
the same subject, a clear benefit when (x,y) chromaticity chart. This technique
monitoring the progress of disease or the isolates the use of colour signals(7) by
effects of treatment. The advantage of masking the detection of any LC signals that
improved test sensitivity is however less may be present in a colour stimulus
useful in detecting minimal colour vision (designed to be isoluminant for the CIE-
deficiencies, largely because of the large 1931 standard observer). When colour
variance within the “normal” population and signals are not available, the stimulus
the lack of test-specific, statistical data to conditions ensure that the subject remains
describe the parameters of the normal completely unaware of the moving stimulus.
population. The variation in chromatic
sensitivity in normal trichromats can be very Thresholds are measured along 16
large(8). In addition to small differences in interleaved directions in colour space. These

1
 Barbur et al., Limits of “normal“ chromatic sensitivity

are grouped together so as to test red-green bands” based on data measured in protanopes,
and the yellow-blue colour sensitivity. The deuteranopes and tritanopes. The distribution of
distribution of these points indicates any the eight data points in the r-g direction changes
selective loss of chromatic sensitivity and significantly in minimal protanomaly and this
makes it possible to classify accurately the class of
provides enough information to classify even
deficiency involved.
minimal deficiencies. Threshold ellipses are
computed and plotted using the standard
CIE 1931 chromaticity chart. Chromatic
discrimination sensitivity has been
measured in 488 subjects (including 238
normal trichromats and 250 colour vision
deficient observers). The criteria for normal
trichromacy were based on the need to pass
all the principal occupational tests of colour
vision, including the Nagel anomaloscope.
3. RESULTS
The colour vision of 238 normal
trichromats and 250 colour deficient
observers has been studied using the CAD
test and a battery of conventional colour
vision tests. Fig. 2. The inserts show the appearance of the
moving colored stimulus during the test. The
subject’s task is to press one of four buttons
placed at the corners of a square box to
Deuteranope
Protanope indicate the direction of motion of the colored
Tritanope stimulus. The CAD test employs four-
2.5% < P < 97.5% alternative, interleaved staircases with a
"Standard Observer" chance probability of 1/16. The subject’s task
0.35 is to discriminate the direction of motion of a
colour-defined stimulus buried in dynamic LC
noise, a technique that isolates the use of
0.34
colour signals(5)
http://www.city.ac.uk/avrc/colourtest.html
0.33
y Analysis of the CAD data provided the
0.32 statistical limits needed to describe the
average ‘normal’ CAD observer. Ellipses
0.31
that describe the median and the 2.5% and
97% limits for the group of normal
trichromats are shown in Fig. 1. The data for
0.3
the normal population were used to produce
0.28 0.29 0.3 0.31 0.32 0.33
a template that allows immediate
x classification of normal and deficient colour
Fig. 1. The coloured discs show typical CAD test vision (see Fig. 1). 250 colour deficient
results for a subject with minimum deuteranomaly observers were also examined in order to
that passes the Ishihara test. The results are plotted assess the usefulness of the new template.
in the CIE –(x,y) 1931 chromaticity chart. The The median values for r-g and y-b
black cross at the centre of the diagram plots the
discrimination thresholds measured in
chromaticity of the white background i.e., 0.305,
0.323. The dotted black ellipse represents the normal trichromats can be used to express
median values computed from the distribution of r- all data in “standard normal” CAD units. The
g and y-b thresholds in 238 normal trichromats. data for the 238 normal trichromats and the
The corresponding 2.5% and the 97.5% statistical 250 colour deficient observers (plotted in
limits were used to plot the innermost and CAD units) are shown in Fig. 3. Sections B
outermost ellipses. Thresholds that fall within the &C show the usefulness of the CAD test in
grey region are taken to reflect “normal” separating unambiguously the subjects with
chromatic discrimination sensitivity. The red, minimal deficiency from the cluster of normal
green and blue lines denote “colour confusion trichromats.
2
CIE Expert Symposium, CIE Proceedings; 75 Years of the Standard Colorimetric Observer,
Ottawa, Ontario, May 2006.
 Barbur et al., Limits of “normal“ chromatic sensitivity

parameters we carried out a number of tests

that involved systematic changes in stimulus
Normal trichromats size and background luminance. The
3 Deuteranomalous
Protanomalous experiments required the use of spectrally
A calibrated neutral density filters in order to
2
achieve very low background luminance
y-b

levels with stable display operation.

1
Fig. 4 shows the effect background
luminance on r-g and y-b thresholds. Fig. 5
0 5 10 15 20 25 shows similar results, but for a range of
stimulus sizes.
6 3
B
C

r-g thresholds (CAD units)

4 2 30
A
y-b

25 Error bars: +/- 2SD

2 1
20

0 1.5 3 4.5 6 15
0 1 2 3
r-g 10

5
Fig. 3. The graph (section A) shows y-b
0
thresholds plotted against the corresponding -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
r-g thresholds in 238 normal trichromats
(black symbols) and 250 colour deficient
observers (green and red symbols). The
y-b thresholds (CAD units)

30
results are expressed in CAD units (i.e.,
median threshold values computed from 25
B
Error bars: +/- 2SD
measurements taken in 238 normal
trichromats). The distribution of r-g 20
thresholds is shown expanded in the range 0
15
to 6 units (section B) and 0 to 3 units (section
C) to illustrate the clear separation between 10
the cluster of points that define the “normal
trichromats” and subjects with minimal 5
deuteranomaly.
0
-1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
Congenital colour deficient subjects were
2
classified into different types of colour Luminance (cd / m )
deficiency and severity graded according to
their chromatic sensitivity loss expressed in
CAD units. Data from deutan and protan Fig. 4. Single subject thresholds for r-g
subjects form distinct patterns that can be (section A) and y-b (section B) colour
used to classify even minimal congenital discrimination measured for background
-2
deficiencies. The “minimal” luminances in the range 51 to 0.01cd m . The
deuteranomalous trichromats shown in thresholds are expressed in CAD units (i.e.,
Figure 3 often pass all the conventional tests median thresholds measured for 238 normal
with the exception of the Nagel trichromats). The arrows indicate the
anomaloscope. Any measured data point background luminance normally employed in
-2
that lies outside the grey shaded area in Fig. the CAD test (i.e., 24 cd m ).
1 indicate possible abnormal colour vision. Although both r-g and y-b thresholds
The two most important parameters that increase rapidly with decreasing background
affect chromatic sensitivity are the size of luminance, yellow-blue discrimination is
the stimulus and the luminance of significantly more affected at lower
background field. In order to establish how luminance levels. When expressed in
CAD thresholds vary with these two standard normal CAD units, stimulus size

3
Barbur et al., Limits of “normal“ chromatic sensitivity

appears to affect both r-g and y-b thresholds recognized colour system. When expressed
in a very similar way (Fig. 5). As the stimulus in standard normal units the results are easy
size is increased from 0.16 to 1.96 square to understand and provide an immediate
degrees, a 2.5 fold increase is observed in indication of the severity of colour vision
both r-g and y-b thresholds. The stimulus loss.
size and background luminance employed in
The test has proved particularly useful in
the CAD test have been selected to
assessing changes in chromatic sensitivity
correspond to the asymptoted region of
in subjects with diseases of the retina and
these response curves. An increase in either
the optic nerve and in specifying minimum
background luminance or stimulus size is
colour vision requirements in occupational
not therefore likely to cause significant
environments. The studies carried out so far
suggest that the new test and the
establishment of the standard normal CAD
r-g thresholds (CAD units)

4
observer provide an accurate means of
detecting and classifying deficiency, of
3 assessing the severity of r-g and y-b colour
vision loss (whether congenital or acquired)
2 and of monitoring small changes in colour
vision either in disease or treatment.

1 ACKNOWLEDGEMENTS
We acknowledge the United Kingdom Civil
0 Aviation Authority for financial support with
0 0.5 1 1.5 2 2.5 this study.

REFERENCES
y-b thresholds (CAD units)

1. Alpern, M. (1979). Lack of uniformity

3 in colour matching. Journal of
Physiology (London) 288: 85-105.
2
2. Alpern, M. & Pugh, E. N., Jr. (1977).
Variation in the action spectrum of
1 erythrolabe among deuteranopes.
Journal of Physiology (London) 266:
0 613-646.
0 0.5 1 1.5 2 2.5
3. Barbur, J. L. (2003). Understanding
Stimulus size (square de g)
colour. Trends Cogn Sci. 2003. Oct.
changes in colour thresholds. 7: 434-436.

4. Barbur, J. L. (2004). 'Double-

Fig. 5. Single subject thresholds for r-g blindsight' revealed through the
(section A) and y-b (section B) colour processing of color and luminance
discrimination measured for as a function of contrast defined motion signals.
stimulus size in the range 0.16 to 1.96 square
Prog. Brain Res. 144: 243-259.
degrees. The arrows indicate the stimulus
size normally employed in the CAD test (i.e.,
0.8 square degrees). 5. Barbur, J. L., Harlow, A. J., & Plant,
G. T. (1994). Insights into the
3. DISCUSSION & CONCLUSIONS different exploits of colour in the
visual cortex. Proc. R. Soc. Lond. B.
The CAD test detects minimum Biol. Sci. 258: 327-334.
deficiencies and quantifies the severity of
colour vision loss by evaluating both r-g and 6. MacAdam, D. L. (1942). Visual
y-b thresholds in an internationally sensitivities to color differences in
4
CIE Expert Symposium, CIE Proceedings; 75 Years of the Standard Colorimetric Observer,
Ottawa, Ontario, May 2006.
 Barbur et al., Limits of “normal“ chromatic sensitivity

daylight. J. Opt. Soc. Am. A. 32:

247-274.

7. Rodriguez-Carmona, M. L., Harlow,

J. A., Walker, G., & Barbur, J. L.
The variability of normal trichromatic
vision and the establishment of the
"normal" range. In Proceedings of
10th Congress of the International
Colour Association.
pp 979-982. Granada, Spain.

8. Squire, T. J., Rodriguez-Carmona,

M., Evans, A. D. B., & Barbur, J. L.
(2005). Color vision tests for
aviation: comparison of the
anomaloscope and three lantern
types. Aviat. Space Environ. Med.
76: 422-429.

Corresponding author:

John L Barbur
Applied Vision Research Centre, City University,
Northampton Square, London EC1V 0HB.
Phone: +44 20 70405060
Fax: +44 20 70408355
e-mail: [email protected]

View publication stats