0042-6989/91 53.00 + 0.00
Copyright 0 1991 Pergamon Press plc
VisionRes. Vol. 31, No. IO, pp. 1797-1812, 1991
Printed in Great Britain. All rights reserved
COLOR VISION FOLLOWING INTENSE GREEN
EXPOSURE: DATA AND A MODEL
HARRY G.
LIGHT
SPERLING,*
ANTHONY
A. WRIGHTand STEPHEN
L. MILLS
The University of Texas Health Sciences Center at Houston, Graduate School of Biomedical Sciences,
Sensory Sciences Center, 6420 Lamar-Fleming Avenue, Houston, TX 77030, U.S.A.
(Received 22 January
1990; in revised form
17 January
1991)
Abstract-Hue discrimination, spectral sensitivity, and mathematical models of both are presented for a
rhesus monkey which was exposed to intense green light. One of the monkey’s eyes was blue-blinded in
a previous experimental procedure and the other was color normal. The results of green light exposure
showed a loss of sensitivity on both measures, with greater loss in the blue-blinded eye. Although there
was considerable loss of hue-discrimination in the blue-green spectral regions, hue-discrimination at the
point of best discrimination, 590 nm, remained unaffected. This pattern of results poses difficulties for
models of hue discrimination, and has resulted in the proposed model employing three opponent color
channels. The number of free-parameters are minimized and the integration between spectral sensitivity
and hue discrimination enhanced by deriving parameters used in modeling hue discrimination from
spectral sensitivity or vice versa.
Color blindness
Hue discrimination
Wavelength discrimination
Rhesus
Monkeys
Models of color vision
Color vision
INTRODUCTION
Repeated exposure of rhesus monkeys’ eyes to
intense blue light has been shown to produce
permanent loss of blue spectral sensitivity
resembling that of tritanopes (Harwerth &
Sperling, 1971, 1975). Histological evidence
supported this finding by providing a pattern
of damaged cones that resembled the pattern
of blue-sensitive cones (Sperling, Johnson &
Harwerth, 1980). Exposure to intense, narrowband, green light had been shown to produce a
loss of green spectral sensitivity resembling that
of deuteranopes, but the loss was temporary and
showed complete recovery to full trichromatic
function in the course of 1840 days (Harwerth
& Sperling, 1971, 1975). This pattern of results
opened the possibility of combining the effects
of these two exposures and effectively producing
an experimental monochromat
monkey, one
that has a permanent loss of blue response and
a temporary loss of green response. Recovery of
the green response would provide the possibility
of monitoring the color vision system from
monochromacy to dichromacy in the already
blue-blinded eye and from dichromacy to normal trichromacy in the normal eye. This would
‘To whom all correspondenceshouldbe addressed.
Deuteranopia
Tritanopia
provide a large new data surface for modeling
the color vision components and their interactions.
The psychophysical color vision measure in
these studies had been the increment-threshold
spectral sensitivity. This measure is somewhat
limited in sensitivity and power because
threshold sensitivity reflects only the upperenvelope of cone sensitivities. It measures only
the sensitivity of the most sensitive (opponent)
process at each wavelength, leaving less sensitive
processes unmeasured. One result of measuring
only the upper envelope was that one could not,
for example, discriminate the case where the
green response had been totally eliminated from
the case where it was made only somewhat less
sensitive than adjacent processes. In an attempt
to rectify this limitation and add more modeling
power, a hue discrimination procedure was
introduced in conjunction with the spectral
sensitivity procedure.
Hue or wavelength discrimination
does
appear to have some advantage over spectral
sensitivity as a phychophysical measure in that
it reflects the proportion or ratio of change in
two (or more) of the color-vision processes at
the wavelength being measured. In order to test
these potential advantages of hue discrimination
measures, one monkey that was blue-blinded in
1797
1798
HARRYG.
SPERLING et al.
the Harwerth and Sperling (1971) study was
trained to make hue discriminations
in a
Maxwellian view apparatus (Wright, Sperling &
Mills, 1987). This monkey was trained to discriminate which of two small spots of monochromatic light was different from a third
reference monochromatic spot. The resulting
hue discrimination functions added considerable information to that provided by the spectral sensitivity functions, and supported the
previous claim that the blue-blinding technique
produced experimental tritanopia. This hue
discrimination experiment demonstrated
the
feasibility of planning to measure both hue
discrimination and spectral sensitivity during
the course of recovery from exposure to intense
green light and the resulting temporary green
blindness.
The purpose of the research reported in this
article was to test spectral sensitivity and hue
discrimination during the course of recovery
from the effects of exposure to intense green
light. The results were surprising in that the hue
discrimination data did not bear out the earlier
conclusions from threshold spectral sensitivity
data that the prolonged green light exposures
produced temporary deuteranopia. The results
of these tests, nevertheless, provide understandable changes in both measures which when dealt
with together provide the basis for a model
which closely integrates spectral sensitivity and
hue discrimination, but in a way which is fundamentally different from other color discrimination models in that satisfying both sets of
data requires a second red-green opponent
channel.
MElTHODS
Subject
The subject was a 15 year old, 9 kg male
rhesus monkey (Mucucu muluttu) who had
extensive experience in several psychophysical
experiments and had been exposed to intense
blue light (Harwerth & Sperling, 1971, 1975;
Wright et al., 1987).
Apparatus
Hue discrimination.Three chromatic channels
produced stimuli of wavelengths ranging from
400 to 700nm of 2-11111
half-bandwidths; and
were formed by three sets of double monochromators with light from a single 1600 W xenon
arc lamp. These paths were manipulated with
standard shutters, apertures, mirrors, neutral
density wedges, and lenses to form an inverted
triangle of three 3000 phot. td circles that
emerged through 0.75 deg holes, bounded by a
2 deg circle, in a 10 deg white, circular, 3000 td
background field. This white background path
was illuminated by a tungsten-halide
lamp,
filtered to approximate equal energy through
the visible spectrum. All four paths were imaged
to form a Maxwellian view. Wavelengths, intensities, behavioral parameters and responses were
automatically
controlled or collected via a
Cromemco Z-2D microcomputer.
A drinking spout, used to deliver juice reinforcement, was positioned so as to provide a
cue for head position so that the trained animal
was able to maintain optimum head position for
alignment of the Maxwellian view. The alignment was frequently checked via a telescopic
arrangement, which allowed the experimenter to
view the refiection of the stimuli in the pupil (see
Wright et al., 1987 for complete details and a
diagram of the apparatus).
Increment-threshold spectral sensitivity. A
different three-channel Maxwellian-view system
was used to test spectral sensitivity. Light from
a 1600 W xenon arc lamp passed through two
10-nm half-bandwidth
monochromators
and
was regulated to produce 50 msec, 2 deg spectral
test stimuli. A tungsten-halide lamp formed an
18 deg, 3000 td, equal energy white background
field upon which the test flash was superimposed. A chromatic adaptation path of 18 deg
was formed by a second channel from the
1600 W xenon arc lamp via interference filters,
and provided 6-nm half-bandwidth
spectral
light. Intensity of the test flash was controlled by
a pair of counter-rotating
neutral density
wedges. The reciprocal of the energy at
threshold for each wavelength provided the
ordinate values of the spectral sensitivity
curves. (For complete details and a diagram of
the optical system see Harwerth & Sperling,
1975.)
Procedure
Hue discrimination. A clicker stimulus signaled to the monkey that a trial could begin.
The monkey depressed a lever to signal readiness for delivery of the stimuli. Two seconds
after lever depression, the three monochromatic
circular spots appeared for one second. The two
upper chromatic spots were the comparison
stimuli. The lower chromatic spot was the reference or standard against which the comparison
stimuli were to be judged. One of the compari-
Intense green light exposure
son stimuli, randomly determined from trial to
trial, always matched the reference stimulus.
The monkey was required to release the lever
within the period, and to make a right-left
choice within the next 4 set, thereby signaling
which of the two upper stimuli appeared different from the lower stimulus. Orange-flavored
Tang was delivered on half of the correct
choices (on a pseudo-random basis). Intertrial
intervals were 7 set following correct responses
and 10 set following incorrect responses or
aborts (trials with no response or responses at
inappropriate times).
The disparity between the reference stimulus
and the nonmatching comparison stimulus was
the independent variable. If the nonmatching
stimulus was of longer wavelength than the
reference and matching comparison stimulus,
this procedure was referred to as a cue-larger
trial; if the nonmatching stimulus was of shorter
wavelength, the trial was referred to as a cuesmaller trial. Thresholds were determined by a
descending method of limits, where the experimenter arranged the initial disparity of the
matching and nonmatching wavelengths to be
easily discriminable. As the disparity was decreased, large steps were initially taken. Smaller
steps were taken as discrimination became more
difficult. The nonmatching chromatic light spot
varied pseudorandomly from the right to the left
side on successive trials. The wavelength difference was reduced after the animal had correctly
identified the nonmatching stimulus as it appeared on each side. A miss on each of the trials
where the stimulus appeared on different sides
resulted in that wavelength difference being
defined as the threshold and threshold deterrnination was then begun for the next wavelength.
All other combinations of events led to a case of
“suspended judgment”, where the same wavelength difference was repeated until one of the
other criteria was met.
Typically, a session consisted of trials in
which the nonmatching comparison stimulus
was of a longer wevelength than the two matching stimuli for the entire session, or occasionally
it was of shorter wavelength. The procedure was
tailored so that an entire function could be
obtained in a single session, thereby reducing
day-to-day variability within individual functions. The dynamic nature of the changes induced by our intense light exposures demanded
this rapid method, as opposed to the more
precise, but time consuming frequency-of-seeing
.
_.
(constant stimulus) method.
1799
sensitivity
Spectral
sensitivity.
Spectral
involved a similar descending method-of-limits
(see Sidney & Sperling, 1967; Herwerth & Sperling, 1975). The 2 deg test stimulus was flashed
for 50msec in a randomly determined time
interval OS-10 set (mean 5 set) after a trial was
started by the monkey depressing the lever. A
correct detection was signaled by release of the
lever within 0.4 set of the flash onset. Test
intensity was reduced by 0.05 log units after
each trial until two successive misses had been
recorded. The intensity of the first miss was
taken as the threshold and the trials for the next
wavelength began. Thresholds were obtained
twice for each 10 nm from 410 to 690 nm and
back (alternating initial directions) in each
session. During color-blinding sessions, where a
large number of trials must be reserved for the
actual exposure to the intense light, a skeleton
function was measured prior to addition of the
intense background field.
Intense green-light exposure. The intense
spectal light exposures, referred to as “greenblinding” were performed using the spectral
sensitivity paradigm. An exposure session began
with determination of a few thresholds for some
control wavelengths, requiring about 100 trials.
Following completion of this portion of the
session, the test flash wavelength was set at
610 nm and an 18 deg, 530 nm, green field was
added coincident with the 18 deg, 3000 td white
background. This field was initially only moderately intense, but its intensity (radiance) was
gradually increased. As the monkey light
adapted, intensity was added to the green field,
maintaining a task that was neither uncomfortable nor unusually difficult for the animal to
perform. An infrared camera enabled the experimenter to monitor the amount of time actually
spent looking at the stimuli as well as any
apparent reluctance of the monkey to view the
increasingly intense stimulus.
As the exposed eye of the monkey adapted to
the green field, threshold sensitivity declined. To
facilitate continued detection at maximal intensity of the green background, the test duration
was lengthened and the wavelength of the test
flash was changed gradually to 580 nm, near
peak sensitivity of the red receptors. Furthermore, the interval in which the test flash might
appear was lengthened to 200 msec to promote
a greater time of exposure to the intense stimulus per trial. An intensity of the background was
always reached, nevertheless, beyond which discrimination of the increment flash was not
1800
HARRYG. SPERJANG
et al.
possible, or beyond which the detection rate was
too infrequent to support the operant behavior.
At this point, the background and stimulation
conditions were maintained for the duration of
the session. This level for the normal eye was
1.2 x 10-l W/sterad of 530 nm, 6-nm half-bandwidth light. For the blue-blind eye, this level was
reached between 1.5 and 2.0 x 10-l Wjsterad.
In the experiment on the normal eye, the
animal was exposed to the intense green light
regime for 10 consecutive days, of several hours
per day terminated when the animal would stop
working because his thirst was slaked. Hue
discrimination measurements, which were
always taken in the afternoon, commenced on
the day of the final intense green exposure. Hue
discrimination sessions were preceded by increment-th~hold spectral sensitivity sessions, and
the spectral-sensitivity values were used to
equate the intensity of the hue discrimination
stimuli.
In the experiment on the previously blueblinded eye, the animal was exposed to the
intense green light regime for 30 consecutive
days, and hue discrimination measurements
were also taken on most green light exposure
days.
chromatic spot of a longer wavelength than the
reference. This is the “cue-larger” relationship.
The functions in Fig. 1 are 5-day means (day 4
of recovery is not included because a cuesmaller relationship was tested on that- day). The
results form a continuous gradient of deficit in
the green region of the fun&on. Loss was
greatest at about 530 nm, with the limen increasing from a normal level of about 6 nm to over
27 nm. Discrimination loss was restricted to the
500-570 nm interval, except for recovery days
6-10 which showed some loss over the blue
region. With this one exception, thresholds outside this region were at least as small as normal.
Figure 2 shows the mean increment-threshold
spectral sensitivity curves taken over the same
time period shown in Fig. 1. The change in
spectral sensitivity was very similar to those
previously shown under similar conditions by
Harwerth and Sperling (1971, 1975). There was
a change from a three-peaked sensitivity curve
to a broadening of the long-wave peak immediately after exposure, a concomitant near-disappearance of the mid-spectral peak, and a return
to the pre-exposure three-peaked curve with
recovery.
Blue -blinded eye
Figure 3 shows hue discrimination results
from a series of intense green-light exposures
to the experimentally blue-blinded eye. The
RESULTS
Normal eye
Figure 1 shows average hue discrimination
curves obtained over 18 days of recovery following exposure of the monkey’s normal eye to
intense green light. The first day of recovery
followed the tenth day of green exposure. Most
sessions were conducted with the nonmat~hing
30 -
20 z
3
10 I
I
500
600
\\I
$700
Wavelength (nm)
500
600
Wavelength (nm)
Fig. 1. Mean hue eon
functions obtained on the
rhesus nomA eye before (&shod line) and svemgadover
days: l-5 (A), 6-10 (V), 11-15 (O), 16-20 (+) follOwin
the 10 day green light exposure series.
Fig. 2. lncmt-tmhold
spectral sensitivity functions
obtained with a 2&g testIi@ ag&ist a 3000 td 18 deg
neutral white background in the rhesus’ tiorma~ eye before
(darker line) and averagad over days: 1-5 (A), 6-10 (7%
11-15 (Cl), 16-20 (+) following the 10 day green light
exposure series.
1801
Intense green light exposure
60
E-
E. 40
$
20
0
400
500
Wavelength
600
700
(nm)
Fig. 3. Hue discrimination functions for the blue-blind eye
before exposure (dashed line) and on days: 2 (V), 4 (A) and
6 (0) of intense green light exposure.
tain motivation for hue-discrimination testing).
It is of some importance that green light
exposure completely eliminates the notch at
580 nm which formerly delineated the red and
green region peaks prior to green light exposure.
Two subsequent exposure experiments replicated the important features of the results
shown in Figs l-4. Although the deficits were
not as long-lasting, they showed similar spectral
sensitivity, hue discrimination, and stability of
the 590 nm minimum in the hue-discrimination
function.
DISCUSSION
AND MODELING
The results of these experiments show the
effects of intense, prolonged
green light
exposure on the spectral sensitivity and hue
discrimination of the monkey’s normal and
blue-blind eye. As shown in Figs 1 and 2, the
prolonged exposure of the normal eye to intense
green light resulted in changes in both hue
discrimination and increment-threshold spectral
sensitivity. At first, there was change at the
580 nm notch of the increment-threshold spectral sensitivity function which was replaced by a
large, broad peak near MO-570nm. This was
accompanied by a roughly three-fold loss of
sensitivity in the hue discrimination function in
the green region between 490 and 590 nm. Notably, there was no discrimination loss at the two
hue discrimination minima at 490 and 590 nm.
Then, as the 580 nm notch in the incrementthreshold function gradually recovered, over a
period of days, the 490-590 nm hue discrimination loss gradually recovered as well.
In the blue-blinded eye, as seen in Figs 3
and 4, with prolonged intense green light exposure there was a loss of the 580 nm increment-7 r
threshold notch, accompanied by a large loss of
hue discrimination over the blue-green region
for wavelengths less than 590 nm. In this bluegreen region, the hue-discrimination sensitivity
decreased from about 30nm to about 80nm.
By contrast, hue discrimination sensitivity at
590 nm, the minimum of the function, remained
unchanged.
Before considering implications of these
green-light
exposure results for models of color
I
I
\\
\ ’
vision, evidence will be presented and it will be
600
L700
500
argued that the monkey based its discrimiWavelength
(nm)
nations
on color or hue differences and not any
Fig. 4. Mean increment-threshold spectral sensitivity functions obtained with a 2 deg test light against a 3000 td neu- possible brightness differences. The results will
tral white backgroundon the rhesus’blue-blindeye before then be compared to human hue discrimination
(solid curve) and on day 4 of green light exposure (A).
results that have shown, for comparable
exposures were continued for a total of 30 days
in an attempt to maintain a peak level of color
deficiency. Recovery began to manifest itself by
day 7 of exposure, however, and peak levels
could not be maintained. Although elevations of
the limens were not maintained for nearly as
long a time period as those to the normal eye
(Figs 1 and 2), the maximum limens attained
were in the 60-80 nm range, as opposed to a
maximum of 30 nm in the normal eye. This
much larger deficit is not surprising in view of
the absence of blue response. In effect, wavelength discrimination in this eye is dependent
upon red-green opponency. With the green
response greatly reduced, it would seem that
there would have to be poorer discrimination
than in the normal eye where blue response was
present.
Spectral sensitivity functions are shown in
Fig. 4 for the pre-exposure and maximum green
light effect conditions (but only down to wavelengths of 490 nm to conserve trials and main-
1802
HARRYG. SPERLINGet al.
reasons, elevated thresholds. These comparisons
will serve to “set the stage” for the hue-discrimination model to be presented in the final section.
Control by hue di$eences not brightness direrences
Spectral sensitivity measurements are critical
to performing valid tests of hue discrimination.
Spectral sensitivity provides the means to equate
the brightness of the stimuli so that subjects will
not and cannot use brightness difference as the
basis for their discrimination performance. Animal subjects underscore this potential problem
because animals, unlike humans, cannot be
brought under instructional control to attend
only to color differences. Animals tend to use
any and all cues available in order to maximize
reward. Therefore, in the hue discrimination
task of these experiments, intensities of chromatic stimuli were adjusted according to the
most recent increment-threshold spectal sensitivity measurements and linear interpolations
were made for itermediate wavelengths. Furthermore, since spectral sensitivity coefficients
which may produce equal brightness at one
intensity level will not necessarily produce equal
brightness at other intensity levels, spectral sensitivity and hue discrimination were conducted
at the same overall intensity level-a 3000 td
background. In addition, as the wavelengths
were made more similar, which occurred each
time the hue discrimination threshold was approached, the intensities of the different wavelengths and hence their brightnesses were forced
to similar values. For example, with a wavelength difference of only 5 nm, there would not
be enough brightness difference, for virtually
any reasonable set of spectral sensitivity coefficients, to provide any reliable brightnessdiscrimination cue.
Finally, a test was conducted to assess any
possible brightness control (Wright, et al.,
1987). The CIE photopic luminosity coefficients
for the standard human observer were used,
with 560 nm set to 3000 td. The difference
between the stimuli adjusted according to the
CIE coefficients as opposed to the monkey’s
increment-threshold
spectral sensitivity values
was as large as 2.4 log units in some cases with
an average difference across the spectrum of
0.67 log units. These differences are much larger
than any day-to-day variability from our increment-threshold spectral sensitivity procedure.
The main reason for this substantial difference
between these two coefficient sets is that flicker
photometry
and hetrochromatic
brightness
matching results, on which the CIE values are
based, reflect the response of the nonopponent
“luminance”
channel, while the incrementthresholds spectral sensitivity values for 50 msec
test flashes of large (2 deg) diameter against high
photopic neutral backgrounds are based upon
opponent processes. The resulting hue discrimination function for the CIE brightness corrections, however, was very similar to the average
hue discrimination function based on the increment-threshold corrections (see Fig. 4 of Wright
et al., 1987). This test provided evidence that the
hue discrimination procedure used in these experiments trained the monkey to rely upon hue
differences not brightness ones.
The results themselves have the form one
would expect with discriminations based upon
hue, which is additional evidence that color
differences, not brightness difference, were the
basis for the discriminations. Figure 5 shows
comparisons with comparable human hue discrimination data. The results from the monkey
are shown as solid symbols; those from humans
are shown as open symbols. The results from the
monkey’s normal eye (solid circles) are similar
in form and magnitude to those from normal
humans (Wright, 1946, Fig. 95, p. 313). Those
from the monkey’s blue-blind eye (solid
squares) are similar in form and magnitude to
0
400
500
600
700
Wavelength (nm)
Fig. 5. Hue discrimination data from rhesus normal eye
befure exposure series (e), from blue-blind eye before
exposure series (a), from bhw-Mind eye after green exposure (A) compamd with Wright’s (1965) data on 5
normal humans (O), Fiir
et al’s (1951) data on a
congenital human tritanopa (a), Willmcr and Wright’s
(1945) data averaged fkom two nomml hwmn w tated
with a 2Omin diameter comparison Add (A).
. . All data are
displayed on a true A1 scale, without dbptram~nts.
Intense green light exposure
those from a congenital tritanope (Fischer, Bouman & ten Doeschate, 1951). Finally, those
from the monkey’s blue-blinded eye during
green-light exposures (solid triangles) are similar in form and magnitude to those from humans tested with a small field, a 20 min diameter
target in the central fovea from two normal
humans (Willmer & Wright, 1945).
It is of some interest that the form of the
hue-discrimination
function for the blueblinded eye after green light exposure is similar
to that taken under conditions of small-field
tritanopia. Support for this small-field comparison was shown in some spectral-sensitivity
results on a different rhesus monkey when the
test stimulus was made progressively smaller.
Figure 6 shows that as the stimulus gets smaller,
the 580 nm notch becomes shallower, completely disappearing below 15 min diameter.
Recall that, in increment-threshold spectral sensitivity, green exposure of the blue-blinded eye
results in a loss of the notch at 580 nm (Fig. 4).
If one were to speculate about underlying mechanisms, it would appear that with intense, repeated green exposure there is a selective loss of
the inhibitory interaction underlying color opponency, signified by a loss of the notch. Some
of the same processes may occur with small field
-7
-6
m
E
:
u
:
-9
$
-I
-11 I
400
500
Wavelength
600
700
(nm)
Fig. 6. Increment-threshold
spectral sensitivity functions
obtained on a normal rhesus eye against an 18 deg 3000 td
neutral background with decreasing stimulus diameters of:
2 deg (top), 30 min (center) and 15 min (bottom).
1803
stimulation, since there is good evidence that
red-green opponent ganglion cells are spatially
organized. When this loss of red-green opponency is combined with an absence (or near
absence) of blue receptor response (in blue-blind
monkey and/or central fovea, then the loss
of discrimination in the blue-green regions is
profound.
It seems clear from the close correspondence
between the monkey’s hue discriminations
under the different conditions and the human
data, in Fig. 5, that the results reported here are
hue, not brightness discrimination, data, since
the human subjects, in each case, were judging
brightness equated stimuli.
Other studies showing elevated hue-discrimination thresholds
Previously it seemed that prolonged greenlight exposure produced temporary deuteranopia (Harwerth & Sperling, 1971, 1975); but
the only measure in those studies was increment-threshold
spectral sensitivity. Deuteranopes, however, show no hue discrimination at
longer wavelengths than about 550 nm (e.g. Pitt,
1935, 1944). But in this region, where hue
discrimination is unmeasurably large for deuteranopes, the monkey of this article discriminates
best-at
the 590 nm minimum. This 590 nm
minimum remained stable despite an apparent
loss of G-cone response sufficient to eliminate a
peak response in the increment-threshold spectral sensitivity function, and to cause a three
fold rise in the mid-spectrum hue discrimination
thresholds.
Deuteranomaly, a color-vision defect which
appears to be due to shifted cone sensitivities
rather than the deuteranope’s reduced sensitivity (Rushton, Powell & White, 1973; Piantanida & Sperling, 1973a, b; Alpern, 1981),
like deuteranopia
shows hue discrimination
threshold elevations at the 590 nm minimum (Wright, 1946, Fig. 190). It is, therefore,
necessary to conclude that the effects of intense
green-light exposure are different from both
deuteranopia and deuteranomaly.
Other experiments which for one reason or
another show elevated hue discrimination
thresholds also show elevated thresholds at the
590 minimum. Figure 7 compares results from
three different experiments (panels A-C) to
the results from the monkey’s green-exposed
normal eye (panel D). The upper two panels
(7A and B) show hue discrimination curves
taken from fovea1 and more eccentric retinal
1804
fiAttRYG.
%ttLINO et d.
Wavelength (nm)
0
400
500
660
Wavelength (nm)
700
500
600
700
Wavelength (nm)
Wavelength (nm)
Fig. 7. Hue discrimination functions obtained (A) by Weaie (1951) in the fovea and at IO@ in the
parafovea, (B) by Mark and Mar& (1978) in the fovea and at 4deg in the parafovas,(Cl by Hurvich
and Jamwon (l%I) in the fovea with two lumimwc kv& of a red annuhw. (D) is our thaws data on
the normal eye obtained in the fovea before and immedintely aRer the series of intcnac gmen light
exposures.
locations. Figure 7A, from Weale (1951, 1953)
compares curves taken in the fovea and at
10 deg eccentricity. The mid-range limen rises to
50 nm, as compared to 30 nm for our monkey’s
limen, and the 49Omn minimum is elevated
relative to the 590 nm minimum. Weale’s study
held the stimulus size constant and made no
special attempts to control possibk rod intrusion at the eccentric location. Unlike the green
exposure results from the monkey’s normal eye
(panel D), there is appreciable loss of discrimination at all wavelengths, including the two
minima, as compared with the fovea1 curve.
Figure 7B shows a similar elevation at 4deg
eccentricity (Ma& & Mar&, 1978). Again, the
entire 4deg curve shows poorer discrimination
than the fovea1 curve. Hue carnation
curves taken with stimuli scaled by cone densities show much less dramatic e&c&s of eccentricity (van Esch, Koldenhof, van Doom &
Koenderink, 1984). Also, hue nonagon
curves taken at 7.5deg temporahy during the
cone plateau period following exposure to
bright light demonstrate that much of Weak’s
effects were due to rod intrusion (Stabell &
Stabell, 1977). Figure 7C shows two curves
taken with a red annutus surrounding the comparison stimuli (Hurvich & Jameaon, 1961).The
lower curve was taken with the luminance of the
test patch set at twice that of the annulus.
The upper curve shows hue discrimination
when the test stimuli were half the luminance of
the annulus. By the well known process of
simultaneous color induction, the red annulus
would be expected to induce a green sensation
(the complement of red) in the area where the
test stimuli are located, and therefore reduce the
ability to discriminate &rely between green stimuli. Although these data were taken in the fovea,
it is important to note that (unlike our monkey’s
post-green-exposure curve) they also show elevthmsh&s at all waveated hue lotion
lengths, including the 590 nm minimum.
Comparisons to these other results help to
exchde the possibility that the monkey in the
expcritrtent of this article shifted his tixation
during testing to a retinal position outside the
area exposed to intense green light. Two points
argue against of&center 8xation as the expianation for the performance changes after green
Intense green
1805
light exposure
light exposure. First, the test stimuli remained
centered in the 3000 td white background
field regardless of eccentricity of viewing and
would thus be expected to maintain rod s&uration. Second, in all cases where a stimulus
of constant size is viewed in the periphery,
discrimination is lost throughout the visible
spectrum. In the results presented for the
monkey following green exposure, however,
there is no loss except in the mid-spectral region
(Fig. 5D), and the minima near 490 and 590 nm
show no loss or wavelength shift. A similar
argument can be made that these results from
the green-light exposed monkey are not due to
some prolongs,
induced green sensation like
those of Fig. 7C, because, there too, all wavelengths showed a loss, unlike our monkey’s
data.
(4
30
20
S
7
Q
10
0
400
500
700
600
Wavelength (nmf
(W
30
20
Models of hue discrimination and spectral sensi-
tiuify
Models of hue discrimination typically begin
with the basic premise that hue discrimination is
more acute in spectral regions where chromatic
response functions change rapidly than where
they change slowly. This is directly related to
cone response spectra in the case of the trichromatic models, or incorporated in opponent
functions in the case of the zone models. Chromatic processes are generally computed under
conditions of constant brightness, and so functions normalized for brightness must therefore
be considered, rather than spectral sensitivities
of the cone or opponent processes. The rate of
change of the chromatic channels is usually
expressed relative to the total amount of
ongoing activity. This is expressed in the use
of Weber fractions for some models (e.g.
Helmholtz, 1891; Stiles, 1946; Boynton, 1979)
and estimation of quanta1 noise fluctuations for
others (Massof, 1977; Vos & Walraven, 1972).
Finally, the relative contributions of individual
channels or components must be combined in
some fashion. These include the use of simple
summation, vector combination, or the use of
only the mechanism most sensitive over a wavelength region.
*The Guth model can predict some loss of mid-range
sensitivity with stability of the 590 nm minimum with an
increase(rather than a decrease) in the + R - G channel
muftipiier.This approach, however, predicts a threshold
spectral sensitivitycurve with a large increase in the
mid-spectral rangethat is incompatible with the spectral
sensitivityresults shown in Figs 2 and 4.
IT
E.
S-z
* 10
i
400
0'
I
I
I
1
500
600
700
Wavelength fnm)
Fig. 8. Comparison of mean hue discrimination data obtained on the normal eye before green light exposure (0)
and immediately after green light exposure (A) with predicted hue discrimination as a function of increasing attenuation of the G cone input to the model from lower to upper
curves in the model of (A) Stiles (1946) and of (B) Guth
et al. (1980).
Figure 8 models the green-exposure results
from our monkey’s normal eye by reducing the
G-cone contribution using the Stiles (1946) and
Guth, Massof and Benzschawel(l980) models*
of hue discrimination, representative of the two
broad categories of single stage trichromatic
and zone models in panels A and B, respectively. As the relative input of the G-cone class
is attenuated, the curves rise primarily at wavelengths longer than 500 nm and will eventually
show a complete elimination of the 590~nm dip
in the function when the U-shaped deuteranopic
endpoint is reached (Pitt, 1935). In zone models,
with attenuation of G-cone response, the elevation of the 590~nm minimum is due to the
broadening of the + R - G opponent function
from one with a peak beyond 600 nm and a
rapid decrease around 590 nm to one closer to
the R-cone function, with a peak nearer to
570 nm and less change around 580-590 nm.
Other zone models (Hurvich & Jameson, 1955;
1806
HARRY G. SPEULINGet al.
Vos & Walraven, 1972; Ingling & Tsou, 1977)
produced similar elevations of the 590 nm minimum, making them inappropriate for modeling
the results of this article.
Proposed model of hue discrimination and spectral sensitivity
A comprehensive model is proposed that
integrates spectral sensitivity and hue-discrimination results by using the parameters derived
from one to account for the other, and is shown
to be compatible with appropriate visual physiology. The problem that previously discussed
models encounter is that reducing the G-term of
the + R - G opponent channel reduces the rate
of change near 590nm and thus falsely predicts
poorer discrimination in that region. A way to
get around this problem is to introduce a
+ R - G term, which then compensates for the
loss of -G in this region with an opponent
mechanism whose rate of change is mod&d
differently with loss of G response. The use of
such a term is defensible on several grounds.
First, neural units with +G - R characteristics
have been found with high frequency in all
single unit studies of retina, LGN and cortex
(e.g. de Monastario & Gouras, 1975; De Valois,
Abramov &Jacobs, 1966; Michael, 1978;Wiesel
& Hubel, 1966). Second, De Valois et al. succeeded in modeling his hue discrimination data
with the inclusion of such a channel based on his
single unit findings in the LGN. Third, in their
treatment of increment-threshold spectral sensitivity data obtained with different intensities
and wavelengths of spectral light backgrounds,
Sperling and Harwerth (1971) found that the
middle- and long-wave peaks, near 530 and
610 nm, required opponent interaction between
R and G cones, but with different rates of
change of the shape of the peaks, with different
chromatic adaptation series, such that the same
R and G inputs could not symmetrically model
both peaks with a change of signs, as proposed
in the models of Guth et al. (1980) and Inghng
and Tsou (1977). Sperling and Harwerth (1971)
therefore proposed a model with different
weightings of the R and G cone inputs to two
R/G channels. Figure 9 shows recently gathered
increment-threshold spectral sensitivity data
obtained on the two eyes of the same rhesus
which illustrates this kind of asymmetry in a
more exaggerated form. The left eye had been
injected with a neurotransmitter blocking agent.
Several things are made clear by these data.
Very nearly the same amounts of +G and -R
m
5
?
r-
.
\.
-9
.
t?
-1
I
I
500
600
\,
700
Wavelength (nm)
Fii. 9. Mean increment-threshold spectral sensitivity functionsobtainedon the r&Ineye (above) and kA eye (below)
ofarhesusmonlteyovertheumepriod,withaZdcsttst
flash against an 18 deg, 3000 td equal energy, white background. The two eyes had been injected with different
chemical agents to modify inhiiory
ncuro-transmitter activity. This data requires the same red-pan
ratio to fit the
center peak in both eyea and a very d3ferent red-pen
ratio
to fit the long wave peak in the two eyes. If there were only
one R/G channel this result would be impossible.
are required to fit the center peak in each eye.
Yet, to fit the long-wave peak in the two eyes
requires very different amounts of both + R and
-G. It follows that the middle and long wave
peaks cannot be represented by the same
+ R - G channel. Also, the center peak cannot
be represented by + R + G, as though it were
produced by a luminance channel, because it is
too narrow. Thus, the conclusion of Sperling
and Harwerth that there are two R/G-opponent
channels with different inputs from the two
classes of cones seems inescapable.
Having two R/G channels raises an interesting theoretical issue. If both are activated at
once, which of the two determines perception?
This is not a problem in the Sperling-Harwerth
model of spectral sensitivity or the huediscrimination model proposed below because they are
upper-envelope models. The most sensitive
channel determines the threshold detection or
discrimination over a spectraI region.
The proposed model of hue discrimination
will be shown to account for the huediscrimination results of this art& and is generalized to
other hue discrimination results as well. The
1807
Intense green light exposure
basic tenets of this model are: (1) spectral
sensitivity and hue discrimination are different
functions of the same three opponent channels;
(2) spectral sensitivity is a direct reflection of the
sensitivity of the three channels; (3) hue discrimination reflects the rate of change as a
function of wavelength in the three channels,
which takes the form of the first derivative with
respect to wavelength; and (4) the opponent
channels are weighted differently in spectral
sensitivity and hue discrimination. These four
tenets are expressed mathematically in the following expression:
spectral sensitivity data and then carried over
unchanged to predict hue discrimination obtained on the same day and under similar
conditions. It was found, empirically, that the
long-wave channel must be weighted 0.1, the
middle 0.35 and the short-wave channel 0.6.
These weightings are not parameters free to vary
for each condition, but are fixed weights
throughout the application of this model for
going from threshold detection to suprathreshold hue discrimination over all conditions
studied. The Sperling-Harwerth model for spectral sensitivity (1971) is:
1
A
Al = 1.3/M
A
A
O.l(K,& - K,G,)
M[O.l(K, R,- &GA),0.35(K,Gi. - K4Ri),
0.6(&B;. - K,R,)]
0.35(K, G1 - &R,)
(1)
M[O.l(K, R;.
- K2G,),
0.35(&G, - K4R,),
0.6(&B;. - KSR,)]
0.6(&B, - KSR,)
M[O.l(K, Rj.-K,G,),0.35(K,Gj.-~Rj.),0.6(K,Bj.-K,Rj.)].
where A, is the predicted hue discrimination
value in nanometers at each measured wavelength; M signifies the supremum, that the most
sensitive term over a wavelength range will
determine the function. A signifies the first
derivative with respect to wavelength. K,-&
weight the cone response functions. The denominator is the same as equation (2) for modeling spectral sensitivity, except that the same
weightings to go from threshold to suprathreshold values are used as in the numerator.
Weighting by the denominator replaces the
usual conversion to luminosity units because, in
this study, the stimuli for hue discrimination
were set equal based on the animal’s incrementthreshold spectral sensitivity measured just
before each hue discrimination session. A formal derivation of the equations will be found in
the Appendix.
The addition of the + G - R opponent channel allows hue discrimination to be modeled as
the first derivative of the three opponent channels of the Sperling-Harwerth model, with the
same constants representing the amount of R,G
and B cone responses. Thus, hue discrimination
and spectral sensitivity can be modeled using
comparable conditions of adaptation or recovery from intense light exposure. In the tests of
the model for hue discrimination (equation 1),
the values of K,-K, were determined by a leastsquares procedure on the increment-threshold
1
KI Rj.- KZGj.
K,G,- K4Rj,
SSI= M
[
1
(2)
KsB~.-K~R~.
where SS represents increment-threshold spectral sensitivity at each measured wavelength; M
signifies the supremum, that each term determines sensitivity completely over the spectral
region where it is most sensitive. R, G and B
represent the cone response functions and K,-K6
weight the cone-response functions for degree of
response.
Table 1 presents the K values which minimized the least squared deviations between the
model of equation (2) and the incrementthreshold spectral sensitivity data. Spectral
sensitivity was measured just prior to each
hue-discrimination session and these data were
used to set the luminosity of the spectral stimuli
in the hue discrimination task.
Table 1. Coefficients (K) for modeling hue-discrimination
(equation 1) and spectral sensitivity (equation 2)
Eye
Coefficients
4
K,
K,
Kz
Normal
pre-exposure
post-exposure
36
1.5
51
0
16
3
Blue-blinded
pre-exposure
post-exposure
14
6
16
2.4
10
4.5
KT
11
5
1.6 1
8
0.5
0.052
0.025
&
3.5
0
2
0
1808
HARRYG. SPERLINGet al.
Figure IOA shows the model applied to the
increment-threshold
spectral sensitivity data
obtained on the normal (kft) eye prior to green
light exposure (Fig. 2, bold line), using the K
values from row 1, Table 1. Figure 10B shows
the model applied to the hue-discrimination
data of Fig. 1 (bold line) using the same K
values. Figure 1lA and B show the model
applied to the normal eye following green light
exposure. The K values, used in both equations,
are those shown in row 2, Table 1. Lilce most
models of hue discrimination, the fit is somewhat less good at the extremes of the visible
spectrum, but our model clearly accounts for
the rise in delta-lambda with green light exposure while holding the 490 and 590 nm minima down, which the models of Fig. 8, and
others, were unable to do.
Figure 12 shows the modeling of the increment-threshold
and hue discrimination
performance before green light exposure of the
monkey’s blue-blinded (right) eye using the K
values shown in Table 1 (row 3). Figure 13
-8
m
5
ii
= _g
in?
-1°40Ao
Wavelength (nm)
(8)
80,
.
iz
5 40
$
20
l-l
I
(4
Wavelength (nm)
Fig. Il. (A) Increment-threshold spectral sensitivity of
rhesus normal eye immediitely after the green light aposure
series as modeled by equation (2); and (B) hue discrimination data obtained on the same days as modeled by
equation (1). both using the same values of K,-&.
0
-:0-o
Wavelength (nm)
8or
W
I
60 -
E
40-
$
20 0
400
500
800
I
700
Wavetength (nm)
Fig. 10. (A) Increment-thrwhold spectral sensitivity of
rhesus normal eye just before the green light e%pcw~re1MitJ
as modeled by equation (2); and (e) hue dircriminatiou
dptp
obtained on the same days as modeled by equation (1). both
using the same values of K,-& . ‘lb dashedline ti@iti the
intrusion of the luminosity channel.
shows the modeling of increment-threshold and
hue discrimination performances for the blueblinded eye immediately after the green light
exposures using the K values of Table 1 (row 4).
The modeling of the blue-blind eye, before and
after green-exposure, is very good and holds the
predicted values at the 590 nm minimum down
well in contrast to the other models with which
we have attempted to fit these data.
The power of this model is best appreciated
by noting that the K values of Table 1 which fit
the cone response functions to the incrementthreshold data are not arbitrarily scakd. The
increment-threshold units are re@~ocal quanta
at threshold and the K values which fit them are
carried over, without any adjustments, to model
the hue disuimiaation
results which show
thresholds that vary from 5 to 76nm. Indeed,
the fits of the model to the data are strong
evidence that the same opponent ctranacls determine both the spectal &&ion
and hue discrimination functions, and that the model of
wuations (1) and (2) embodies the correct met-
1809
Mense green light exposure
-’r (A)
400
Or (4
I
I
I
500
600
700
Wavelength
Wavelength (nm)
W
(5)
80 1-r
80
0
r
60
E6
(rim)
z
5
40
ai
I
-
40-
2
20 -
20
0
0
400
600
500
Wavelength
700
(nm)
Fig. 12. (A) Increment-threshold spectral sensitivity of
rhesus blue-blind eye just before the green light exposure
series as modeled by equation (2); and (B) hue discrimination data obtained on the same days as modeled by
equation (l), both using the same values of RI-&,. The
dashed lines represent the intrusion of reds and of the
luminosity channel neither of which entered the hue discrimination calculation.
ric for relating spectral detection and discrimination.
yodeling tritan~pic spectral sensitivity and hue
discrimination
In a classic experiment on human congenital
tritanopia, Wright (1952) measured both spectral sensitivity and hue discrimination. These
results have been used previously in the evaluation of color discrimination models and are
examined in relation to the model of this article.
In modeling Wright’s data, the six K values were
determined to best fit his hue discrimination
data, rather than deriving them from the fit to
spectral sensitivity as was done in modeling our
own data. These K values were then entered in
equation (2) and compared with the mean results of Wright’s tritanopic spectral-sensitivity
measurements. Wright used flicker photometry
to obtain his spectral sensitivity values, This
technique gives no weight to the response of
0
400
500
600
I
700
Wavelength (nm)
Fig. 13. (A) Increment-threshold spectral sensitivity of
rhesus blue-blind eye taken on the fourth day of the green
light exposure series as modeled by equation (2); and (B) hue
discrimination data obtained on the same day as modeled
by equation (I), both using the same values of FL,-&.
blue receptors, but yields a curve which is well
fit by the sum of the red and green channels. We
have applied these modifications to our model.
The three terms in equation (2) were summed
with & (the coefficient for the B-cones) set to
zero. The same changes were made in the denominator of equation (1). As shown in Figs
14A and B, the fit is quite good, further demonstrating that the model of equations (1) and (2)
is an efficient way to predict both hue discrimination and spectral sensitivity under a variety of
conditions.
The phy~iologic~
exposures
nature
of the green
light
The results presented in this article raise
questions about the physiological effects of the
green light exposure. It is clear that blue light
exposures (Harwerth & Sperling, 197 1, 1975;
Wright, Sperling & Mills, 1987) resulted in
irreversible damage to the B cones, as attested
by the lack of recovery over many years and the
histological evidence of degeneration of cones
which were distributed across the retina in the
1810
HARRYG. SPERLING
et al.
Implications and concluding remarks
(4
60
60
P
5
40
$
20
0 L
400
500
Wavelength
400
500
Wavelength
600
700
(nm)
600
700
(nm)
Fig. 14. (A) Mean hue diination
function from four
tritanopes measured by Wright (1952) as modeled by
equation (1); (B) mean Bicker photometric spectral sensitivity function of t&mopes from Wright (1952) modeled by
summingthe + R - G and + G - R channelsof equation(2)
with the same values of K,-K,.
same way as the distribution of blue receptors
labeled by histochemical means (Spetling et al.,
19gO). There are several reasons that green-light
exposures are of a different nature. First, they
appear to be entirely reversible. Second, the
effects of green-light exposures seem to be as
great to the R cones as to the G cones (see
Table 1 and Figs lO-13), unlike the blue
light experiments which affected only the B
cones. Third, the post-exposure green light data
could not be modeled without eliminating or
greatly reducing the inhibitory -R and -G
terms of both threshold and hue discririaination
equations while the + R and +G t@ms,
although reduced, rentained appracipby large.
Since inhibition ocours at higher-order sUges
of visual proce&n& these facts point to the
green-light exposures primarily aECCt& postreceptoral processes, and not the receptors
themselves.
Several implications can be drawn from the
results and the model presented in this article.
First, the simple upper-envelope model that
fits increment-threshold
spectral sensitivity
(equation
2) also fits hue discrimination
(equation 1). That is, the opponent channel that
is most sensitive over a region of the visible
spectrum determines the hue discrimination
function over that region. Second, the model fits
hue discrimination functions obtained over a
variety of conditions for both of the monkey’s
eyes, before and after the green light exposure
series, without changing the channel weights.
This single set of channel weights successfully relates constants determined from spectral
sensitivity to hue discrimination data over
a widely changing set of sensitivities, without
need for additional parameters. It follows
that the same opponent mechanisms serve
both detection and discrimination. Third, these
processes are adequately modeled by linear subtractive interaction between cone response functions. The cone response functions that we have
used are those of Smith and Pokorny (1975). but
they do not differ appreciably from the more
directly measured data of Baylor, Nunn and
Schnapf (1987) when pre-retinal absorptions
are taken into account. Fourth, the fact that
the same changes in the K values predict the
effects of green light exposure on both detection
and hue discrimination is further evidence that
the same mechanisms are involved in both
functions and also that the first-derivative or
rate of change of the paired linear opponent
mechanisms of equation (2) is an appropriate
for predicting
hue discrimination
metric
(equation 1).
Although it may be argued that six parameters to fit 22 data points is not very powerful
modeling, we consider it very powerful that
under four widely differing conditions the same
values of the six parameters predict both detertion and discrimination.
Acknowledgemetits-This rcs~arch was supported in Pati
by the National Eye Institute, grant EYa23gg. National
&i~nce Foundation grant BPlS 8414977, and U.S. Am~y
Surgeon CjeneralContract DAMDl7-C3003. WCp@fUllY
acknowledge
the~aretul
arsirtana
of DAd
Gnmtt
and
James Mathcny during portions of this n%ear&.
REFERENCES
A&m, M. (1981). The coior vision of the color
Japanese Journal of Ophthdtdogy,
25, l-17.
blind.
1811
Intense green light exposure
Baylor, D. A., Nunn, B. J. & Schnapf, J. (1987). Spectral
sensitivity of cones of the monkey Macaca fascicularis.
Journal of Physiology, London, 390, 145-160.
Rushton, W. A. H., Powell, D. S. & White, K. D. (1973).
Pigments in anomalous trichromats. Vision Research, 13,
2017-2031.
Boynton, R. M. (1979). Human color vision. New York:
Holt, Rinehart & Winston.
van Esch, J. A., Koldenhof, E. E., van Doom, A. J. &
Koenderinkm, J. J. (1984). Spectral sensitivity and wavelength discrimination of the human peripheral visual field.
Journal of the Optical Society of America, Series 2, 1,
Sidley, N. A. & Sperling, H. G. (1967). Photopic spectral
sensitivity in the rhesus monkey. Journal of the Optical
443450.
De Valois, R. L., Abramov, I. & Jacobs, G. H. (1966).
Analysis of response patterns of LGN cells. Journal of the
Optical Society of America, 56, 966-977.
Sperling, H. G. & Harwerth, R. S. (1971). Red-green cone
interations in the increment-threshold spectral sensitivity
of primates. Science, 172, 180-184.
Sperling, H. G., Johnson, C. & Harwerth, R. S. (1980).
Differential spectral photic damage to primate cones.
Vision Research, 20, 1117-l 125.
Stabell, U. & Stabell, B. (1977). Wavelength discrimination
of peripheral cones and its change with rod intrusion.
Fischer, F. P., Bouman, M. A. & ten Doeschate, J. (1951).
A case of tritanopy. Documenta Ophthalmologica, 5-6,
73-87.
Guth, S. L., Massof, R. W. & Benzschawel, T. (1980).
Vector model for normal and dichromatic
color
vision. Journal of the Optical Society of America, 70,
197-212.
Harwerth, R. S. & Sperling, H. G. (1971). Prolonged color
blindness induced by intense spectral lights in rhesus
monkeys. Science, 174, 520-523.
Harwerth, R. S. & Sperling, H. G. (1975). Effects of intense
visible radiation on the increment-threshold spectral sensitivity of the rhesus monkey eye. Vision Research, 15,
1193-1204.
Helmholtz, H. von (1891). Versuch einer erweiterten Anwendung des Fechnerschen Gezetzes im Farbensystem.
Zeitschrtft fur Psychologie und Physiologie Sinnesorgen, 2,
I-30.
Hurvich, L. & Jameson, D. (1955). Some quantitative
aspects of an opponent-colors theory-II.
Brightness,
saturation and hue in normal and dichromatic vision.
Journal of the Optical Society of America, 45, 606616.
Hurvich, L. & Jameson, D. (1961). Opponent chromatic
induction and wavelength discrimination. In Jung &
Komhuber (Eds), The visual system: Neurophysiology,
biology and psychophysics. Berlin: Springer.
Ingling, C. R. Jr & Tsou, B. H. (1977). Orthogonal combinations of three visual channels. Vision Research, 17,
1075-1082.
Marre, M. & Marre, E. (1978). Different types of acquired
colour deficiencies on the base of CVM patterns in
dependence upon the fixation mode of the diseased eye.
Modern Problems of Ophthalmology, 19, 248.
Massof, R. W. (1977). A quantum fluctuation model for
fovea1 color thresholds. Vision Research, 17, 565-570.
Michael, C. R. (1978). Color vision mechanisms in monkey
striate cortex: Dual-opponent cells with concentric receptive fields. Journal of Neurophysiology, 41, 572-588.
de Monastario, F. M. & Gouras, P. (1975). Functional
Properties of ganglion cells of the rhesus monkey. Journal
Society of America, 57, 8 16.
Smith, V. C. & Pokomy, J. (1975). Spectral sensitivity of the
fovea1 cone photopigments between 400 and 5OOnm.
Vision Research, 15, 161-171.
Vision Research,
17, 423-426.
Stiles, W. S. (1946). A modified Helmholtz line-element
in brightness-colour space. Proceedings of the Physical
Society, 58, 41-65.
Vos, J. J. & Walraven, P. L. (1972). An analytical descrip
tion of the line element in the zone-fluctuation model of
colour vision-II.
The derivation of the line-element.
Vision Research, 12, 1345-I 365.
Weale, R. A. (1951). Hue-discrimination in pars-central
parts of the human retina measured at different luminance
levels. Journal of Physiology, London, 119, 170-190.
Weale, R. A. (1953). Spectral sensitivity and wavelength
discrimination of the peripheral retina. Journal of Physiology, London, 119, 170-190.
Wiesel, T. N. & Hubel, D. H. (1966). Spatial and chromatic
interactions in the lateral geniculate body of the rhesus
monkey. Journal of Neurophysiology, 29, 1115-l 156.
Willmer, E. N. & Wright, W. D. (1945). Colour sensitivity
of the fovea centralis. Nature, London, 156, 119.
Wright, A. A., Sperling, H. G. & Mills, S. L. (1987).
Researches on a unilaterally blue-blinded rhesus monkey.
Vision Research, 27, 1551-1564.
Wright, W. D. (1946). Researches on normal and defective
color vision. London: Kimpton.
Wright, W. D. (1952). Characteristics of tritanipia. Journal
of the Optical Society of America, 42, 509-52 1.
APPENDIX
Both spectral sensitivity and hue discrimination are assumed
to arise from activities in three channels that are transforms
of cone responses as follows:
(Al)
of Physiology, London, 251, 167-196.
Piantanida, T. P. & Sperling, H. G. (1973a). Isolation of a
third chromatic mechanism in the protanomalous observer. Vision Research, 13, 2033-2047.
fiuntankia, T. P. & Sperling, H. G. (1973b). Isolation of a
third chromatic mechanism in the dueuteranomalous
observer. Vision Research, 13, 2049-2058.
Pitt, F. H. G. (1935). Characteristics of dicromatic vision.
Medical Reseach Council, Report of Committee on Physiology of Vision @. 14). London: HMSO.
Pitt, F. H. G. (1944). The nature of normal trichromatic and
dichromatic vision. Proceedings of the Royal Society,
London, B, 132, 104.
where Ri, G, and Bi are the cone response functions. The
KS are determined by least squares fit to the increment
threshold spectral sensitivity function as follows:
WI
1812
HARRYG. SPEJWNGet al.
Hue dc3dmimtion
’
involves the same three channels,
reweighed because they am suprathreahold with rape
to
threshold m
sensitivity and because of procedural
diiereMea,
-so:
(‘46)
when
6
~~;=lrg;+,,-~g;l
8r;
for cue larger and
by;
The equal brightness constraint of the hue discrimination paradigm necessitates a weighting by the br$&neas
tcfnl:
(A4)
This is exactly analogous to SS;, but with the weights
of equation (A3). The channel activities ,after bri#ttncas
Arg; = 1rg; - rg;_,,1
for cue smaller.
So hue discrimination
above matsure as:
is inwxsely proportional
M = l.3/AHuei
to the
647)
Combining the above and entering the weightiltgs:
O.l(K,R, - K,G,)
“max[O.l(K,Ri-KzGi),0.35(K3GA-K&),0.6(K,Ri-K&)]
AA = 1.3/max
A
0.35(K,G, - K,RJ
max [O.l(K, RA- K,G,),
0.35(K,G, - K,Ri).0.6&Bi
- &RJ
O.B(K,B; - K,Ri)
A
_ max[0.l(K,R~-K,G~),0.3S(K,G,-K,R1),0.6(K,B~-&R~)]
adjustment are then:
-
bd
Discrimination of hue is proportional
activity in these three channels:
Lbv:j
to rate of change of
which is the working cqtion
whii the hue diskmination
calculated.
@qualiOn t) of the text, with
pmdktions of Figs 8-M were