VisionRrs.Vol. IS.pp, 1205-1212. Pergamon
Press 1975. Punted in Great Bntain.
THE SPECTRAL SENS~~~IES OF ISOLATED HUMAN COLOR ~E~ANIS~S DETER~I~D FROM CONTRAST EVOKED POTENTIAL MEASURELESS 0. EST&QZZ,H. SPEKREIJSE,T. J. T. P. VAN
DEN
BERG and C. R. CAVONIUS
Laboratory of Medical Physics, University of Amsterdam, and The Netherlands Ophthalmic Research Institute, Wilhemina Gasthuis, Amsterdam, The Netherlands (Received 14 October 1974)
Abstract-Previous attempts to measure human spectral sensitivity curves by physiological methods have not yielded results with the precision and over a sufficiently wide intensity range to enable useful predictions to be made about the fine details of color vision. In the present experiment we have used a contrast evoked response technique to measure a set of spectral sensitivities whose quality is comparable to that of psychophysically obtained data. For this purpose a visual stimulus was prepared that would presumably be seen as a pattern by only one color mechanism. The EP data obtained from one rod-monochromat correspond well with the CIE 1951 scotopic relative luminous efficiency function, whereas the data from normal and dichromatic subjects resemble several of Stiles’ n, mechanisms. At the long wavelength end of the visible spectrum the relative peak sensitivities and the shapes of the spectral sensitivity functions closely agree with Stiles’ x4 and n5 mechanisms. The spectral sensitivity of the short wavelength mechanism had the form of either the n, or n3 mechanism. The dichromat results can be fitted by the same mechanisms. The long wavelength mechanism of a protanope resembled K, and that of a deuteranope nj.
It is a currently accepted view that human vision is mediated by two types of photor~ptors: rods and cones, which are primarily
responsible
for achromatic
(night) and chromatic (day) vision, respectively. It is also widely accepted that normal color vision is mediated by three classes of cones, which have different spectral sensitivities. Because of severe technical problems, attempts to measure the absorption spectra of human cones have not yielded results that are precise enough to explain the details of color vision (Rushton, 1958, 1963, 1966; Brown and Wald, 1964; Marks, Dobelle and MacNichol, 1964; see also Riggs, 1967). For this reason, many theories of color vision are based on the more reliable data that can be obtained in psychophysical experiments. However, it is not clear whether psychophysical data can be used to infer the sensitivities of individual cone mechanisms. For example, Stiles (1939) measured the spectral sensitivities of a group of color mechanisms (“n-mechanisms”), but was careful to point out that these mechanisms differ in several details from the absorption spectra of known photopigments (Stiles, 1959). On the other hand Wald (1964) claimed to have measured the spectral sensitivities of human cone pigments by means of a method that was similar to that used by Stiles. The only serious discrepancy between Wald’s and Stiles’ data occurs for the red m~hanism. The sensitivity of Wald’s “red pigment” to short wavelength stimulation is lower than that of the Stiles rc5 mechanisms (Enoch, 1972). Because of the simplicity of Wald’s data, we used his sensitivity functions in a previous paper (Estevez and Spekreijse, 1974). Xn that paper we introduced a spectral compensation method which allows one to modu-
late one color mechanism independently of the modulation to another that has a different spectral sensitivity. The basis of the method is simple. Given a good approximation of the spectral sensitivity of a color mechanism, the intensities of two spectrally different stimuli can be adjusted so that they have an identical effect on one color mechanism. Then only mechanisms with a different spectral sensitivity will be modulated when the two stimuli are temporally modulated in counterphase (Fig. 1).
E ; f f % s
c3
400
500
600 Xl
x2
700 “l%l
Fig. 1. Spectral sensitivities of two hypothetical photoreceptor systems to illustrate the spectral compensation method. See text for explanation.
12Oh
0. Esrt’wc.
The compensation method was applied in order to determine the flicker characteristics of the separate red and green cone mechanisms subserving human color vision. The results led us to conclude that we did indeed succeed in isolating these two spectral mechanisms. We wished, however, to have more direct physiological evidence on the quality of isolation that was achieved with the compensation method. For this purpose we designed an experiment to measure electrophysiologically, by means of a visual evoked response technique. the spectral sensitivities of the mechanisms that were stimulated in each of our previously computed isolation conditions. Kellerman and Adachi-Usami (1972/3973) have used previously a visual evoked response (VER) technique for the same purpose. In their electrophysiological experiment they employed the experimental paradigm of Wald’s psychophysical method: monochromatic flashes of varying intensities were superimposed on strong chromatic adapting fields and the VERs were recorded by averaging methods. Although Kellerman and Adachi-Usami could identify three different cone functions, their results show the same shortcomings which so far have characterized most physiological attempts in this direction (Riggs, 1967). It is, however, possible to derive reliable spectral sensitivity functions from electrophysiological data. For example, Norren and Padmos (1973) used the electroretinogram (ERG) to measure the spectral sensitivity of the “blue” cone system of macaque monkey and man. They obtained their high quality data by employing a vector voltmeter technique (Padmos and Norren, 1972). In the present experiment we have succeeded in measuring by electrophysiological methods a set of spectral sensitivities whose quality is comparable to that of psychophysically obtained data. This was achieved by recording the electrical response that is elicited by changes in spatial contmst rather than by changes in lumirzance (Spekreijse, Van der Tweel and Zuidema. 1973). We made use of the fact that if other conditions remain constant, the amplitude of the contrast evoked response that is measured from the scalp over the occipital cortex depends on the relative depth of contrast modulation. Therefore the superposition of any homogeneous field of light upon the modulated pattern will change the amount of contrast and hence the amplitude of the evoked response. The test consisted of measuring as function of wavelength the intensity of a monochromatic light that was needed to reduce the amplitude of the contrast evoked potential to a prespecified criterion level. This technique, in which the effect of an adapting field with varying wavelength upon the periodic response that is elicited by a modulated beam is taken as measure for the spectral sensitivity, parallels closely the field radiance vs field wavelength technique introduced by Stiles to derive his 7c,,functions. The electrophysiological application of this idea has been realized previously by Van den Berg and Spekreijse (1974) for the ERG of goldfish and by Regan (1974) for the contrast evoked response in man. In both studies the chromaticity of the modulated beam was kept constant and chosen on the expectancy that it would stimulate mostly one type of receptors. This in contrast to the present experiment where a visual
(‘t 40’). Therefore to measure the spectral sensitivity of this subject on the basis of the contrast evoked response, we presented an achromatic, 4 cd/m2, 3” checkerboard field in which adjacent 50’ checks interchanged positions six times/set. This temporal frequency is just below the cut-off frequency of the spatial contrast mechanism at this luminance level. With this stimulus the continuous curve in Fig. 3 was obtained by adjusting the intensity of the adapting field so as to maintain a contrast evoked potential of 2 pV. Figure 3 shows a complete scan from the short to the long wavelength end of the visible spectrum and vice versa. The conventional CAT averaged responses indicated that along the spectral sensitivity curve of Fig. 3 the amplitude, phase and shape of the response remained constant, regardless of the wavelength of the adapting field. This suggests that a single (scotopic) spectral mechanism accounts for the spectral sensitivity curve of Fig. 3 and that the recording is not contaminated by other spectral mechanisms.
01, 400
’
500
600
700
Wavelength,
400
500
Red
600
n
nm
Fig. 4. Spectral sensitivity of the green and red mechanisms of two normal observers. Lines: position of neutral-density wedge that was required to maintain a constant evoked response amplitude. Circles: psychophysical thresholds, adjusted along the vertical axis for best fit with the electrophysiological data. These data have not been corrected for the spectral output of the adapting field.
Spectral sensitivities of isohted human colour mechanisms
I
I
1209
I
Deuteranope
i
Wavelength,
nm
Fig. 5. Spectral sensitivities of the long-wavelength mechanisms of two dichromatic and symbols as in Fig. 4.
of the compensation technique, stimulated only ooze of the two long wavelength mechanisms. The sensitivity functions that were obtained with this method in two color-normal subjects are shown in Fig. 4. These curves were obtained for a response criterion of 2pv. The psychophysical threshold settings that were obtained in the same stimulus situation are given by the dots in Fig. 4. For each subject the combination of stimulus contrast and response criterion was chosen on the basis of the signal-to-noise ratio of his contrast evoked potentialThis same combination was used when measuring both the red and the green systems. The stimulus intensities were so set that the adaptation levels for both systems were the same. The green spectral characteristics of the color-normal subjects resemble one another, as do the red spectral characteristics. The green spectral characteristic is somewhat narrower than the red, and its peak sensitivity is slightly higher. In both the electrophysiological and the psychophysical data obtained from our color-normal subjects the red and green sensitivity functions cross near 570 nm. (b) Dichromat results. When the same compensation values that were used to isolate the red and green mechanisms in color-normal subjects were used with dichromats, the results of Fig. 5 were obtained. Regardless of the relative content of the red and green lights in the stimulus, the dichromats gave only one type of sensitivity curve. The spectral sensitivity curves of the protanope closely resemble the green spectral sensitivity curves obtained in the two colornormal subjects, whereas the spectral characteristics of the deuteranope resemble the red curve of colornormal subjects. For the dichromats, as well as the normals, there is a close agreement between the form of the curves obtained by the electrophysiologi~l and psychophysical measurements. When the protanope was presented with the stimulus that elicits in normal observers a response of the red mechanism, he gave a response that had the spec-
subjects. Lines
tral position and form of the green mechanism but with much lower sensitivity. To raise this response to the level obtained with the stimulus for the green mechanism, the depth of modulation had to be increased by a factor of 10. Similarly, when the deuteranope was presented with the stimulus that elicits in normal observers a response of the green mechanism, the resulting response had the spectral sensitivity of the red mechanism, but depressed in ~nsiti~ty about a factor 6.4. We do not interpret this to mean that the dichromats had both of the normal long-wavelength mechanisms (but that the abundance of one was reduced). Rather, it shows that our compensation ratios were not perfect, so that the mechanism that should have received no modulation was in fact stimulated, when very great modulation depths were used. (c) The b/tie system. In the isolation of the red and green fovea1 sensitivities, the blue color mechanism was neglected because it is rather insensitive to wavelengths above 500 nm. The scarcity of blue cones in the fovea and the macular pigmentation further reduces a possible contribution by the short wavelength mechanism because the contrast evoked potential is predominantly a fovea1 response. In addition, the blue mechanism is extremely insensitive to a stimulus that is composed of 15’ checks. These circumstances, which were useful in the m~surements of the red and green spectral sensitivities, turned into a great disadvantage when we attempted to measure the spectral sensitivity of the blue mechanism. From previous experience we estimated that checks of about 40’ angular subtense are a preferred contrast stimulus for the blue mechanism. At the same time these sizes are a rather poor stimulus for the longer wavelength systems; they elicit a contrast evoked potential that has an amplitude that is less than one fourth that obtained with 15’ checks. Although this will favor the relative contribution of the blue mechanism in the contrast evoked potential, the two long wavelength mechanisms cannot be neglected since their sensitivity
-I
I
I
1
I
4-
Normal
A
Normal
6
3-
compensated with the four lights that were available. The electrophysiologically and psychophysically determined blue sensitivity of two color normals and two dichromats are presented in Fig. 6. The eorrespondence between the two types of measurements is quite fair. although the variability was greater than for the long wavelength systems.
DISCL’SSION
Deuteronope
‘Km
I
,
I
500
600
700 Wavelength,
/ >o
500
600
700
nm
Fig. 6. Spectral sensitivity of the blue mechanism of two normal and two dichromatic observers. Lines and symbols as in Fig. 4.
to short wavelengths remains rather high. Therefore, an ideal stimulus for the blue mechanisms would include complete compensation of the red and green mechanisms. In principle this can be done, but it would have required, for color-normal subjects, the addition of two more light sources to the existing four. For this reason the spectral sensitivity of the blue mechanism of color-normal observers was measured with a mixed technique: the compensation was adjusted so that the stimulus pattern was not seen by the green cone system and an intense red adapting field was superposed to bring the contrast in the red cone system below threshold. In the dichromats, however, the sensitivity of the blue mec~nism could be measured without the adapting field, since their single long wavelength mechanism could be completely
In order to estimate the variability of the raw data from the electrophysiological experiments, it is important to take into account the time lag of the lockin amplifier and wedge drive unit. This time lag is inherent in the use of a running average for improvement of signal-to-noise ratio. Because of this averaging, the wedge position that is plotted at any given time is in part determined by the response at some earlier time. The shift along the horizontal axis depends on the integration time that is selected and the speed at which the wavelength of the adapting field is changed. For the conditions that were used to record the data that have been presented, this shift was equivalent to about IOnm. If the raw data are displaced by this amount, the sensitivity functions measured wtth increasing and with decreasing adapting field wavelength generally agree to within better than 0.3 log units, which is comparable to the variability in the psychophysical data. After correction for the time lag and the spectral output of the adapting beam, the data for the photopit elec~ophysiological experiments resemble several of Stiles’ n, mechanisms (Fig. 7). Regan (1974) also used a constant-response method to obtain two spectral sensitivity functions, which have peak sensitivities in about the same regions of the spectrum as n4 and z5. However, the shapes of his sensitivity functions suggest that he achieved only partial separation of the two long-wavelength mechanisms. It is not possible to decide on the basis of our data whether our blue mechanism has the sensitivity of the 7~~or the 7~ me~~nism, because the relative sensitivities of these mechanisms are very similar over the range of wavelengths and adapting field luminances that was available to us. Unfortunately,
Normal A
700
400
Wavelength,
Fig. 7. Spectral of the adapting
500
600
700
nm
sensitivities of normal and dichromatic subjects after correction for the spectral output field, expressed as relative log quantum sensitivity. Lines: electrophysiological measurements. Symbols: Stiles nI, n4 and n5 mechanisms.
Spectral sensitivities of isolated human colour mechanisms
we cannot compare the maximum sensitivity of our blue mechanism with those of the red and green mechanisms, because we were forced to use larger check sizes, greater modulation and more intense stimuli to elicit a measurable response from the blue system. However, when recording the ~nsiti~ty of the red and green systems, we could use equivalent conditions, so that the peak sensitivities of these systems can be directly compared. Their sensitivities, as well as the shapes of the spectral sensitivity functions, suggest strongly that the red and green mechanisms that were measured were 7c5 and rcn4,respectively. None of our data require that opponent color processing is taking place at the neural level at which the contrast evoked response is generated. This is in accordance with the notion that the “red” and “green” signals are segregated until at least the stage at which the contrast evoked potential is generated (Regan and Spekreijse, 1974). It should be realized that this segregation of color channels does not imply that complete spectral characteristics of the red and green systems can be obtained simply by stimulating with red and green light. When the pro&mope was presented with the green stimulus, he gave a normal zr4 sensitivity. Similarly, the sensitivity of the deuteranope to the red stimulus was in no way different from a normal n5. However, when the protanope was presented with the stimulus that was designed not to stimulate the green system, his spectral sensitivity still resembled Q, but the modulation had to be increased 10 times to reach the same sensitivity. Similarly, when the deuteranope was given the stimulus that was designed not to stimulate the red system, his sensitivity profile remained that of x5, but the proper position along the sensitivity axis was only reached when six times the normal modulation was used. On the basis of the similarity of the protanope and deuteranope spectral sensitivities with the normal n,, and n5 mechanisms respectively, we can estimate that in normal observers the residual modulation of the green, when stimulating the red system, is loo/, that of the red and that when stimulating
the green system, the
red system is modulated by about 17% of the modulation of the green. The finding that the compensation for the red system is less complete than that for the green is due to the fact that the red compensation ratios were calculated from Wald’s “cone” function. As it turned out, Wald’s red function is not a good approximation to our measured red sensitivity. This result, deviating from the initial assumptions about the shape of the spectral ~nsitivities. could be obtained since an occipital response can be recorded only when the contrast modulation exceeds a certain value. For white light this value coincides with the psychophysical threshold (Campbell and Maffei, 1970). Therefore even if a stimulus condition is selected on the basis of an approximate spectral sensitivity function and as a result also stimulates another mechanism but to a lesser extent, one can, by choosing a low criterion response, eliminate the unwanted contribution of the second m~hanism. The resulting function will regect then the sensitivity function of only the mechanism that is most strongly stimulated. In this way the initial set of compensation factors that is employed for the isolation of a color coded
1211
channel can be improved on the basis of the spectral data obtained. The close agreement between our data and the rcmechanisms of Stiles suggests to us that the color coded channels as seen through the contrast EP are mediated through such m~hanisms. The displacement rules that Stiles used to obtain his z-functions have been questioned by Hurvich, Jameson and Krantz (1965). Our results, however, were obtained by using an isolation technique that does not require the use of displacement rules. Furthermore, Hurvich (1963) raised the objection that it is impossible to determine whether the criterion used by Stiles’ observers was based only on brightness contrast, or whether ~turation and hue also determined the thresholds. However, the present experiment used a response that is largely (and perhaps entirely) determined by the contrast of the stimulus (Spekreijse pt al., 1973). It is clear then that when appropriate stimulus conditions are used, the contrast EP can be mediated by the scotopic system and/or by at least three photopie mechanisms. Since greater modulation depth and larger check sizes were needed to obtain the responses from the rod system and from the blue cone mechanism, it may be that in normal vision these mechanisms contribute little to the perception of spatial contrast and that this is performed chiefly by the red and green (rc5 and zq) mechanisms. This difference between the patterns and modulation depths that were needed to elicit measurable responses from the different mechanisms cannot be due to chromatic aberration, which in everyday vision presumably degrades the stimulus seen by the blue mechanisms. In the present experiment, the stimuli were seen only by one mechanism at a time, and the subject naturally adjusted his accommodation so that the stimulus was in focus for that mechanism. The way in which the different mechanisms mediate color perception has not been investigated in this
study. Until now this question could not be directly studied because the chromatic adaptation method requires the s~Ippression of all but the mechanism that is investigated. With the spectral compen~tion method it is, however, possible to modulate each mechanism independently and simultaneously in any desired combination. Therefore this method shows considerable promise as a technique for studying the spatial and temporal characteristics of individual mechanisms and their interactions. Acknowle~gemeMts-This research was supported by the Organization of Health Research (TNO), The Hague and by the Netherlands Organization for the Advancement of Pure Research (ZWO), The Hague. REFERENCES
Berg T. J. T. P. Van den and Spekreijse H. (1974) ERG photometry in goldfish by the criterion response method. Documenta ophfh. (in press). Brown P. K. and Waid G. (1964) Visual pigments in single rods and cones of the human retina. Science, N.E: 144, 45-51. Campbell I;. J. and MatTei L. (1970) Electrophysiological evidence for the existence of orientation and size detectors in the human visual system. J. Physic& Land. 207, 635-652.
1212
0. EsT~:
Enoch J M. (1972) The two-color threshold technique of Stiles and derived component color mechanisms. In ~a~book ?f &wary ~~z~s~~~~~~~, Vol. VII/4 (Edited by Jameson D. and Hurvich L. M.). pp. 537-567. Springer. Berlin. Estbvez 0. and Spekreijse H. (1934) A spectral compensation method for determining the flicker characteristics of the human color mechanisms. Vision Res. 14,823-830. Green D. G. (1968) The contrast sensitivity of the color mechanisms of the human eye. J. Physioi., Land. 196, 415-429. Hurvich L. M. (1963) Contributions to color-discrimination theory: review. summary and discussion. J. opt. See. Am. 53, 196-201. Hurvich L, M., Jameson D. and Krantz D. H. fi965) Theoretical treatments of selected visual problems. In Handbook of ~uth~~tica~ Psvcholouv (Edited bv Lute R. D., Bush”R. R. and Gala&r E.) %I: 3, pp. l&-159. Wiley. New York. Kelierman F. J. and Adachi-Usami E. (1972/1973) Spectral sensitivities of color mechanisms isolated by the human visual evoked response. Ophthal. Res. 4, 199-210. Marks W. B., Dobelle W. and MacNichol E. F. Jr. (1964) Visual pigments of single primate cones. Stirnce. N.Y. 143, 1181-l 183. Norren D. V. and Padmos P. (1973) Human and macaaue blue cones studied with electroretinographv. I . - Msion jips. 13. t241-1254. Padmos P. and Norren D. V. 119721The vector voitmeter as a tool to measure electrorctinogram spectral sensitivity and dark adaptation. fnoestcc Ophtlz. Il. 7X3-788. Regan D. and Spekreijse H. ft974) Evoked potential indications of co&r blindness. i/isiotz Rrs. 14. X9-95.
Regan D. (1974) ~iectr~physiological evidence for colour channels in human pattern vision. Natrne. Ix&. Z-SO, 437439. Riggs L. (1967) Electrical evidence on the trichomatic theory. Ir?oescae OpAt!+zI.6, 6-17. Rushton W. ‘4. H. (1958) The cone pigments of the human fovea in colour blind and normal. In Visual P~o~~~~ls ($Ccoiour. Vol. 1. pp. 71-105. H.M.S.O., London. Rushton W. A. H. (1963) Cone pigment kinetics in the protanope. J. Physbl., Lonif. 188,-374-388. Rushton W. A. H. (1966) Densitometry of pigments in rods and cones of normal and color defective subjects. IITrcstw O&h, 5. 233%241_ Snekreiise H. (1966) Analysis of EEG resnonses in man, ~evokid by sine &ve modulated light. T&z&. Junk, The Hague. Spekreijse H., Tweel L. H. van der and Zuidema Th. (1973) Contrast evoked responses in man. &ion Rrs. 13, 1577,-1601.
Stiles W. S. (iY39) The directional sensitivity of the retina and the spectral sensitivity of the rods and cones. Proc. R. Sot. B. 127, 64-105. Stiles W. S. (1959) Color vision: the approach through increment threshold sensitivity. PPOC. fxatn. rlcad. Scr’. I!.S.A. 45, 10s 113. Twccl L. H. van der and Saekreiise H. (19731 Psvchonhvsits and electrophysiolog; of a iod-a&rom&. $roc. ‘X& I.S.C.I%R.G. Synqx tos At~&es. pp. 163%l73. Junk, The Hague. Wald G. (1964) The receptors of human color vision. Sc‘irrxr. N.x 145. 1~7-10~6.
Press 1975. Punted in Great Bntain.
THE SPECTRAL SENS~~~IES OF ISOLATED HUMAN COLOR ~E~ANIS~S DETER~I~D FROM CONTRAST EVOKED POTENTIAL MEASURELESS 0. EST&QZZ,H. SPEKREIJSE,T. J. T. P. VAN
DEN
BERG and C. R. CAVONIUS
Laboratory of Medical Physics, University of Amsterdam, and The Netherlands Ophthalmic Research Institute, Wilhemina Gasthuis, Amsterdam, The Netherlands (Received 14 October 1974)
Abstract-Previous attempts to measure human spectral sensitivity curves by physiological methods have not yielded results with the precision and over a sufficiently wide intensity range to enable useful predictions to be made about the fine details of color vision. In the present experiment we have used a contrast evoked response technique to measure a set of spectral sensitivities whose quality is comparable to that of psychophysically obtained data. For this purpose a visual stimulus was prepared that would presumably be seen as a pattern by only one color mechanism. The EP data obtained from one rod-monochromat correspond well with the CIE 1951 scotopic relative luminous efficiency function, whereas the data from normal and dichromatic subjects resemble several of Stiles’ n, mechanisms. At the long wavelength end of the visible spectrum the relative peak sensitivities and the shapes of the spectral sensitivity functions closely agree with Stiles’ x4 and n5 mechanisms. The spectral sensitivity of the short wavelength mechanism had the form of either the n, or n3 mechanism. The dichromat results can be fitted by the same mechanisms. The long wavelength mechanism of a protanope resembled K, and that of a deuteranope nj.
It is a currently accepted view that human vision is mediated by two types of photor~ptors: rods and cones, which are primarily
responsible
for achromatic
(night) and chromatic (day) vision, respectively. It is also widely accepted that normal color vision is mediated by three classes of cones, which have different spectral sensitivities. Because of severe technical problems, attempts to measure the absorption spectra of human cones have not yielded results that are precise enough to explain the details of color vision (Rushton, 1958, 1963, 1966; Brown and Wald, 1964; Marks, Dobelle and MacNichol, 1964; see also Riggs, 1967). For this reason, many theories of color vision are based on the more reliable data that can be obtained in psychophysical experiments. However, it is not clear whether psychophysical data can be used to infer the sensitivities of individual cone mechanisms. For example, Stiles (1939) measured the spectral sensitivities of a group of color mechanisms (“n-mechanisms”), but was careful to point out that these mechanisms differ in several details from the absorption spectra of known photopigments (Stiles, 1959). On the other hand Wald (1964) claimed to have measured the spectral sensitivities of human cone pigments by means of a method that was similar to that used by Stiles. The only serious discrepancy between Wald’s and Stiles’ data occurs for the red m~hanism. The sensitivity of Wald’s “red pigment” to short wavelength stimulation is lower than that of the Stiles rc5 mechanisms (Enoch, 1972). Because of the simplicity of Wald’s data, we used his sensitivity functions in a previous paper (Estevez and Spekreijse, 1974). Xn that paper we introduced a spectral compensation method which allows one to modu-
late one color mechanism independently of the modulation to another that has a different spectral sensitivity. The basis of the method is simple. Given a good approximation of the spectral sensitivity of a color mechanism, the intensities of two spectrally different stimuli can be adjusted so that they have an identical effect on one color mechanism. Then only mechanisms with a different spectral sensitivity will be modulated when the two stimuli are temporally modulated in counterphase (Fig. 1).
E ; f f % s
c3
400
500
600 Xl
x2
700 “l%l
Fig. 1. Spectral sensitivities of two hypothetical photoreceptor systems to illustrate the spectral compensation method. See text for explanation.
12Oh
0. Esrt’wc.
The compensation method was applied in order to determine the flicker characteristics of the separate red and green cone mechanisms subserving human color vision. The results led us to conclude that we did indeed succeed in isolating these two spectral mechanisms. We wished, however, to have more direct physiological evidence on the quality of isolation that was achieved with the compensation method. For this purpose we designed an experiment to measure electrophysiologically, by means of a visual evoked response technique. the spectral sensitivities of the mechanisms that were stimulated in each of our previously computed isolation conditions. Kellerman and Adachi-Usami (1972/3973) have used previously a visual evoked response (VER) technique for the same purpose. In their electrophysiological experiment they employed the experimental paradigm of Wald’s psychophysical method: monochromatic flashes of varying intensities were superimposed on strong chromatic adapting fields and the VERs were recorded by averaging methods. Although Kellerman and Adachi-Usami could identify three different cone functions, their results show the same shortcomings which so far have characterized most physiological attempts in this direction (Riggs, 1967). It is, however, possible to derive reliable spectral sensitivity functions from electrophysiological data. For example, Norren and Padmos (1973) used the electroretinogram (ERG) to measure the spectral sensitivity of the “blue” cone system of macaque monkey and man. They obtained their high quality data by employing a vector voltmeter technique (Padmos and Norren, 1972). In the present experiment we have succeeded in measuring by electrophysiological methods a set of spectral sensitivities whose quality is comparable to that of psychophysically obtained data. This was achieved by recording the electrical response that is elicited by changes in spatial contmst rather than by changes in lumirzance (Spekreijse, Van der Tweel and Zuidema. 1973). We made use of the fact that if other conditions remain constant, the amplitude of the contrast evoked response that is measured from the scalp over the occipital cortex depends on the relative depth of contrast modulation. Therefore the superposition of any homogeneous field of light upon the modulated pattern will change the amount of contrast and hence the amplitude of the evoked response. The test consisted of measuring as function of wavelength the intensity of a monochromatic light that was needed to reduce the amplitude of the contrast evoked potential to a prespecified criterion level. This technique, in which the effect of an adapting field with varying wavelength upon the periodic response that is elicited by a modulated beam is taken as measure for the spectral sensitivity, parallels closely the field radiance vs field wavelength technique introduced by Stiles to derive his 7c,,functions. The electrophysiological application of this idea has been realized previously by Van den Berg and Spekreijse (1974) for the ERG of goldfish and by Regan (1974) for the contrast evoked response in man. In both studies the chromaticity of the modulated beam was kept constant and chosen on the expectancy that it would stimulate mostly one type of receptors. This in contrast to the present experiment where a visual
(‘t 40’). Therefore to measure the spectral sensitivity of this subject on the basis of the contrast evoked response, we presented an achromatic, 4 cd/m2, 3” checkerboard field in which adjacent 50’ checks interchanged positions six times/set. This temporal frequency is just below the cut-off frequency of the spatial contrast mechanism at this luminance level. With this stimulus the continuous curve in Fig. 3 was obtained by adjusting the intensity of the adapting field so as to maintain a contrast evoked potential of 2 pV. Figure 3 shows a complete scan from the short to the long wavelength end of the visible spectrum and vice versa. The conventional CAT averaged responses indicated that along the spectral sensitivity curve of Fig. 3 the amplitude, phase and shape of the response remained constant, regardless of the wavelength of the adapting field. This suggests that a single (scotopic) spectral mechanism accounts for the spectral sensitivity curve of Fig. 3 and that the recording is not contaminated by other spectral mechanisms.
01, 400
’
500
600
700
Wavelength,
400
500
Red
600
n
nm
Fig. 4. Spectral sensitivity of the green and red mechanisms of two normal observers. Lines: position of neutral-density wedge that was required to maintain a constant evoked response amplitude. Circles: psychophysical thresholds, adjusted along the vertical axis for best fit with the electrophysiological data. These data have not been corrected for the spectral output of the adapting field.
Spectral sensitivities of isohted human colour mechanisms
I
I
1209
I
Deuteranope
i
Wavelength,
nm
Fig. 5. Spectral sensitivities of the long-wavelength mechanisms of two dichromatic and symbols as in Fig. 4.
of the compensation technique, stimulated only ooze of the two long wavelength mechanisms. The sensitivity functions that were obtained with this method in two color-normal subjects are shown in Fig. 4. These curves were obtained for a response criterion of 2pv. The psychophysical threshold settings that were obtained in the same stimulus situation are given by the dots in Fig. 4. For each subject the combination of stimulus contrast and response criterion was chosen on the basis of the signal-to-noise ratio of his contrast evoked potentialThis same combination was used when measuring both the red and the green systems. The stimulus intensities were so set that the adaptation levels for both systems were the same. The green spectral characteristics of the color-normal subjects resemble one another, as do the red spectral characteristics. The green spectral characteristic is somewhat narrower than the red, and its peak sensitivity is slightly higher. In both the electrophysiological and the psychophysical data obtained from our color-normal subjects the red and green sensitivity functions cross near 570 nm. (b) Dichromat results. When the same compensation values that were used to isolate the red and green mechanisms in color-normal subjects were used with dichromats, the results of Fig. 5 were obtained. Regardless of the relative content of the red and green lights in the stimulus, the dichromats gave only one type of sensitivity curve. The spectral sensitivity curves of the protanope closely resemble the green spectral sensitivity curves obtained in the two colornormal subjects, whereas the spectral characteristics of the deuteranope resemble the red curve of colornormal subjects. For the dichromats, as well as the normals, there is a close agreement between the form of the curves obtained by the electrophysiologi~l and psychophysical measurements. When the protanope was presented with the stimulus that elicits in normal observers a response of the red mechanism, he gave a response that had the spec-
subjects. Lines
tral position and form of the green mechanism but with much lower sensitivity. To raise this response to the level obtained with the stimulus for the green mechanism, the depth of modulation had to be increased by a factor of 10. Similarly, when the deuteranope was presented with the stimulus that elicits in normal observers a response of the green mechanism, the resulting response had the spectral sensitivity of the red mechanism, but depressed in ~nsiti~ty about a factor 6.4. We do not interpret this to mean that the dichromats had both of the normal long-wavelength mechanisms (but that the abundance of one was reduced). Rather, it shows that our compensation ratios were not perfect, so that the mechanism that should have received no modulation was in fact stimulated, when very great modulation depths were used. (c) The b/tie system. In the isolation of the red and green fovea1 sensitivities, the blue color mechanism was neglected because it is rather insensitive to wavelengths above 500 nm. The scarcity of blue cones in the fovea and the macular pigmentation further reduces a possible contribution by the short wavelength mechanism because the contrast evoked potential is predominantly a fovea1 response. In addition, the blue mechanism is extremely insensitive to a stimulus that is composed of 15’ checks. These circumstances, which were useful in the m~surements of the red and green spectral sensitivities, turned into a great disadvantage when we attempted to measure the spectral sensitivity of the blue mechanism. From previous experience we estimated that checks of about 40’ angular subtense are a preferred contrast stimulus for the blue mechanism. At the same time these sizes are a rather poor stimulus for the longer wavelength systems; they elicit a contrast evoked potential that has an amplitude that is less than one fourth that obtained with 15’ checks. Although this will favor the relative contribution of the blue mechanism in the contrast evoked potential, the two long wavelength mechanisms cannot be neglected since their sensitivity
-I
I
I
1
I
4-
Normal
A
Normal
6
3-
compensated with the four lights that were available. The electrophysiologically and psychophysically determined blue sensitivity of two color normals and two dichromats are presented in Fig. 6. The eorrespondence between the two types of measurements is quite fair. although the variability was greater than for the long wavelength systems.
DISCL’SSION
Deuteronope
‘Km
I
,
I
500
600
700 Wavelength,
/ >o
500
600
700
nm
Fig. 6. Spectral sensitivity of the blue mechanism of two normal and two dichromatic observers. Lines and symbols as in Fig. 4.
to short wavelengths remains rather high. Therefore, an ideal stimulus for the blue mechanisms would include complete compensation of the red and green mechanisms. In principle this can be done, but it would have required, for color-normal subjects, the addition of two more light sources to the existing four. For this reason the spectral sensitivity of the blue mechanism of color-normal observers was measured with a mixed technique: the compensation was adjusted so that the stimulus pattern was not seen by the green cone system and an intense red adapting field was superposed to bring the contrast in the red cone system below threshold. In the dichromats, however, the sensitivity of the blue mec~nism could be measured without the adapting field, since their single long wavelength mechanism could be completely
In order to estimate the variability of the raw data from the electrophysiological experiments, it is important to take into account the time lag of the lockin amplifier and wedge drive unit. This time lag is inherent in the use of a running average for improvement of signal-to-noise ratio. Because of this averaging, the wedge position that is plotted at any given time is in part determined by the response at some earlier time. The shift along the horizontal axis depends on the integration time that is selected and the speed at which the wavelength of the adapting field is changed. For the conditions that were used to record the data that have been presented, this shift was equivalent to about IOnm. If the raw data are displaced by this amount, the sensitivity functions measured wtth increasing and with decreasing adapting field wavelength generally agree to within better than 0.3 log units, which is comparable to the variability in the psychophysical data. After correction for the time lag and the spectral output of the adapting beam, the data for the photopit elec~ophysiological experiments resemble several of Stiles’ n, mechanisms (Fig. 7). Regan (1974) also used a constant-response method to obtain two spectral sensitivity functions, which have peak sensitivities in about the same regions of the spectrum as n4 and z5. However, the shapes of his sensitivity functions suggest that he achieved only partial separation of the two long-wavelength mechanisms. It is not possible to decide on the basis of our data whether our blue mechanism has the sensitivity of the 7~~or the 7~ me~~nism, because the relative sensitivities of these mechanisms are very similar over the range of wavelengths and adapting field luminances that was available to us. Unfortunately,
Normal A
700
400
Wavelength,
Fig. 7. Spectral of the adapting
500
600
700
nm
sensitivities of normal and dichromatic subjects after correction for the spectral output field, expressed as relative log quantum sensitivity. Lines: electrophysiological measurements. Symbols: Stiles nI, n4 and n5 mechanisms.
Spectral sensitivities of isolated human colour mechanisms
we cannot compare the maximum sensitivity of our blue mechanism with those of the red and green mechanisms, because we were forced to use larger check sizes, greater modulation and more intense stimuli to elicit a measurable response from the blue system. However, when recording the ~nsiti~ty of the red and green systems, we could use equivalent conditions, so that the peak sensitivities of these systems can be directly compared. Their sensitivities, as well as the shapes of the spectral sensitivity functions, suggest strongly that the red and green mechanisms that were measured were 7c5 and rcn4,respectively. None of our data require that opponent color processing is taking place at the neural level at which the contrast evoked response is generated. This is in accordance with the notion that the “red” and “green” signals are segregated until at least the stage at which the contrast evoked potential is generated (Regan and Spekreijse, 1974). It should be realized that this segregation of color channels does not imply that complete spectral characteristics of the red and green systems can be obtained simply by stimulating with red and green light. When the pro&mope was presented with the green stimulus, he gave a normal zr4 sensitivity. Similarly, the sensitivity of the deuteranope to the red stimulus was in no way different from a normal n5. However, when the protanope was presented with the stimulus that was designed not to stimulate the green system, his spectral sensitivity still resembled Q, but the modulation had to be increased 10 times to reach the same sensitivity. Similarly, when the deuteranope was given the stimulus that was designed not to stimulate the red system, his sensitivity profile remained that of x5, but the proper position along the sensitivity axis was only reached when six times the normal modulation was used. On the basis of the similarity of the protanope and deuteranope spectral sensitivities with the normal n,, and n5 mechanisms respectively, we can estimate that in normal observers the residual modulation of the green, when stimulating the red system, is loo/, that of the red and that when stimulating
the green system, the
red system is modulated by about 17% of the modulation of the green. The finding that the compensation for the red system is less complete than that for the green is due to the fact that the red compensation ratios were calculated from Wald’s “cone” function. As it turned out, Wald’s red function is not a good approximation to our measured red sensitivity. This result, deviating from the initial assumptions about the shape of the spectral ~nsitivities. could be obtained since an occipital response can be recorded only when the contrast modulation exceeds a certain value. For white light this value coincides with the psychophysical threshold (Campbell and Maffei, 1970). Therefore even if a stimulus condition is selected on the basis of an approximate spectral sensitivity function and as a result also stimulates another mechanism but to a lesser extent, one can, by choosing a low criterion response, eliminate the unwanted contribution of the second m~hanism. The resulting function will regect then the sensitivity function of only the mechanism that is most strongly stimulated. In this way the initial set of compensation factors that is employed for the isolation of a color coded
1211
channel can be improved on the basis of the spectral data obtained. The close agreement between our data and the rcmechanisms of Stiles suggests to us that the color coded channels as seen through the contrast EP are mediated through such m~hanisms. The displacement rules that Stiles used to obtain his z-functions have been questioned by Hurvich, Jameson and Krantz (1965). Our results, however, were obtained by using an isolation technique that does not require the use of displacement rules. Furthermore, Hurvich (1963) raised the objection that it is impossible to determine whether the criterion used by Stiles’ observers was based only on brightness contrast, or whether ~turation and hue also determined the thresholds. However, the present experiment used a response that is largely (and perhaps entirely) determined by the contrast of the stimulus (Spekreijse pt al., 1973). It is clear then that when appropriate stimulus conditions are used, the contrast EP can be mediated by the scotopic system and/or by at least three photopie mechanisms. Since greater modulation depth and larger check sizes were needed to obtain the responses from the rod system and from the blue cone mechanism, it may be that in normal vision these mechanisms contribute little to the perception of spatial contrast and that this is performed chiefly by the red and green (rc5 and zq) mechanisms. This difference between the patterns and modulation depths that were needed to elicit measurable responses from the different mechanisms cannot be due to chromatic aberration, which in everyday vision presumably degrades the stimulus seen by the blue mechanisms. In the present experiment, the stimuli were seen only by one mechanism at a time, and the subject naturally adjusted his accommodation so that the stimulus was in focus for that mechanism. The way in which the different mechanisms mediate color perception has not been investigated in this
study. Until now this question could not be directly studied because the chromatic adaptation method requires the s~Ippression of all but the mechanism that is investigated. With the spectral compen~tion method it is, however, possible to modulate each mechanism independently and simultaneously in any desired combination. Therefore this method shows considerable promise as a technique for studying the spatial and temporal characteristics of individual mechanisms and their interactions. Acknowle~gemeMts-This research was supported by the Organization of Health Research (TNO), The Hague and by the Netherlands Organization for the Advancement of Pure Research (ZWO), The Hague. REFERENCES
Berg T. J. T. P. Van den and Spekreijse H. (1974) ERG photometry in goldfish by the criterion response method. Documenta ophfh. (in press). Brown P. K. and Waid G. (1964) Visual pigments in single rods and cones of the human retina. Science, N.E: 144, 45-51. Campbell I;. J. and MatTei L. (1970) Electrophysiological evidence for the existence of orientation and size detectors in the human visual system. J. Physic& Land. 207, 635-652.
1212
0. EsT~:
Enoch J M. (1972) The two-color threshold technique of Stiles and derived component color mechanisms. In ~a~book ?f &wary ~~z~s~~~~~~~, Vol. VII/4 (Edited by Jameson D. and Hurvich L. M.). pp. 537-567. Springer. Berlin. Estbvez 0. and Spekreijse H. (1934) A spectral compensation method for determining the flicker characteristics of the human color mechanisms. Vision Res. 14,823-830. Green D. G. (1968) The contrast sensitivity of the color mechanisms of the human eye. J. Physioi., Land. 196, 415-429. Hurvich L. M. (1963) Contributions to color-discrimination theory: review. summary and discussion. J. opt. See. Am. 53, 196-201. Hurvich L, M., Jameson D. and Krantz D. H. fi965) Theoretical treatments of selected visual problems. In Handbook of ~uth~~tica~ Psvcholouv (Edited bv Lute R. D., Bush”R. R. and Gala&r E.) %I: 3, pp. l&-159. Wiley. New York. Kelierman F. J. and Adachi-Usami E. (1972/1973) Spectral sensitivities of color mechanisms isolated by the human visual evoked response. Ophthal. Res. 4, 199-210. Marks W. B., Dobelle W. and MacNichol E. F. Jr. (1964) Visual pigments of single primate cones. Stirnce. N.Y. 143, 1181-l 183. Norren D. V. and Padmos P. (1973) Human and macaaue blue cones studied with electroretinographv. I . - Msion jips. 13. t241-1254. Padmos P. and Norren D. V. 119721The vector voitmeter as a tool to measure electrorctinogram spectral sensitivity and dark adaptation. fnoestcc Ophtlz. Il. 7X3-788. Regan D. and Spekreijse H. ft974) Evoked potential indications of co&r blindness. i/isiotz Rrs. 14. X9-95.
Regan D. (1974) ~iectr~physiological evidence for colour channels in human pattern vision. Natrne. Ix&. Z-SO, 437439. Riggs L. (1967) Electrical evidence on the trichomatic theory. Ir?oescae OpAt!+zI.6, 6-17. Rushton W. ‘4. H. (1958) The cone pigments of the human fovea in colour blind and normal. In Visual P~o~~~~ls ($Ccoiour. Vol. 1. pp. 71-105. H.M.S.O., London. Rushton W. A. H. (1963) Cone pigment kinetics in the protanope. J. Physbl., Lonif. 188,-374-388. Rushton W. A. H. (1966) Densitometry of pigments in rods and cones of normal and color defective subjects. IITrcstw O&h, 5. 233%241_ Snekreiise H. (1966) Analysis of EEG resnonses in man, ~evokid by sine &ve modulated light. T&z&. Junk, The Hague. Spekreijse H., Tweel L. H. van der and Zuidema Th. (1973) Contrast evoked responses in man. &ion Rrs. 13, 1577,-1601.
Stiles W. S. (iY39) The directional sensitivity of the retina and the spectral sensitivity of the rods and cones. Proc. R. Sot. B. 127, 64-105. Stiles W. S. (1959) Color vision: the approach through increment threshold sensitivity. PPOC. fxatn. rlcad. Scr’. I!.S.A. 45, 10s 113. Twccl L. H. van der and Saekreiise H. (19731 Psvchonhvsits and electrophysiolog; of a iod-a&rom&. $roc. ‘X& I.S.C.I%R.G. Synqx tos At~&es. pp. 163%l73. Junk, The Hague. Wald G. (1964) The receptors of human color vision. Sc‘irrxr. N.x 145. 1~7-10~6.
