Flicker ERGs Representing Chromaticity and Luminance Signals

Abstract

Purpose.: To test the hypothesis that electroretinograms (ERGs) reflect luminance activity when measured at high temporal frequencies and chromatic activity when measured near 12 Hz.

Methods.: The authors measured the responses to stimuli in which the output of red and green light-emitting diodes was modulated in counterphase at different ratios, varying the luminance content in the stimulus while keeping the red-green chromatic contrast and its phase constant.

Results.: The high temporal frequency electroretinography was determined mainly by the luminance contrast. At 12 Hz, electroretinographic response amplitudes and phases primarily reflected the red-green chromatic content of the stimulus. Control experiments, performed with a deuteranopic subject and with stimuli that silenced the rods and S-cones, excluded an explanation based on intrusion from rod- and S-cone–driven responses.

Conclusions.: It now is possible to perform noninvasive measurements of basic electrophysiological properties of the luminance and chromatic pathways on a retinal level, and their disease-related changes, in human observers.

The flicker electroretinogram (ERG) probably represents primarily the activity of bipolar cells, ¹ but there are several types of bipolar cells belonging to different postreceptoral pathways. The two main types are midget and diffuse bipolar cells. Midget bipolar cells receive input from one or a few cones and transfer the information to midget retinal ganglion cells. These, in turn, project to the parvocellular layers of the lateral geniculate nucleus. This parvocellular pathway is thought to be the physiological basis for red-green color vision. ^2–11 Diffuse bipolar cells contact a larger number of cones and make contact with parasol retinal ganglion cells. They are thus part of the magnocellular pathway, which is probably responsible for luminance vision. ^{2,4,8,10–13} The receptive fields of primate diffuse and midget bipolar cells have clear center-surround structures when using isochromatic stimuli, showing that some spatial antagonism is present at the level of the bipolar cells. ¹⁴ It is not clear how midget and diffuse bipolar cells respond to chromatic stimuli. However, it can be expected that they have distinct physiological properties and that they resemble, at least partially, those of the retinal ganglion cells belonging to the same retinogeniculate pathway. 

Until recently, there were few indications that the ERG can be correlated with chromatic activity. ^15–17 Luminance-like responses were described in the high temporal frequency flicker electroretinography because the long (L-) and middle (M-) wavelength-sensitive cone-driven responses are in phase. L- and M-cone–driven ERGs display an amplitude ratio that is generally larger than 1.0 at 30 Hz and reflect the ratio of L- and M-cone numbers in the retina. ^18,19 Furthermore, the spectral sensitivity of flicker electroretinography at high temporal frequencies closely resembles that of the psychophysical luminance channel. ^20–22 This correlation between the 30-Hz flicker electroretinography and the luminance channel still may not indicate a causal relationship between the two because a luminance-like electroretinography signal may also arise when it is driven by all available L- and M-cones and when the L- and M-cone–driven signals have similar phases. However, we previously found that chromatic adaptation has a similar influence on the L/M ratios in the 30-Hz flicker electroretinography and on luminance-mediated flicker detection thresholds. Making the background chromaticity redder increased the M-cone–driven responses and decreased the L-cone–driven responses so that the L/M ratios were strongly altered. The same chromatic adaptation had a similar effect on the psychophysically measured L/M ratio when the luminance channel mediated detection but not when the task was based on the red-green chromatic channel. ²³ This suggests that the luminance channel and the ERG share postreceptoral adaptation mechanisms and thus are more intimately related. 

We recently found that the 12-Hz L- and M-cone–driven flicker ERGs are in counterphase with each other, indicating cone opponent response properties. In addition, at this temporal frequency, the ratio of L- and M-cone–driven electroretinographic amplitudes is approximately 1 for all subjects, independent of the ratio of cone numbers. ²⁴ Other observations have shown that a ratio close to unity can also occur in the multifocal ERG and the multifocal visual evoked potential. ^25–27 It was found that the L/M ratio in psychophysical flicker detection thresholds is approximately unity when the red-green chromatic channel underlies flicker detection. ^19,28–30 Furthermore, the 12-Hz ERGs showed less interindividual variability and were less strongly influenced by chromatic adaptation. ²⁴ These are again properties that resemble those of the red-green opponent chromatic pathway. ^19,23,30 On the basis of these data, we inferred that, at 12 Hz, the ERG may represent activity in the red-green chromatic pathway. 

To be able to test the proposition that ERGs to mixed luminance and chromatic stimuli primarily represent activity in the red-green chromatic channel at a frequency near 12 Hz and in the luminance channel at high temporal frequencies, experiments should be performed in which luminance and red-green chromaticity contrasts are varied in a distinct manner. Furthermore, other explanations, such as photoreceptor-driven responses (electroretinographic signals that are correlated with the response properties of photoreceptors but not with postreceptoral visual processing), should be excluded. In this article, we present the results of experiments measured at a range of temporal frequencies using stimuli with a varying luminance contrast but a constant red-green chromatic content. Our hypothesis would predict that, at high temporal frequencies, the ERG would change according to the stimulus luminance contrast. At 12 Hz, the response amplitudes and phases are expected to be constant, similar to the red-green chromatic contrast. To be able to obtain data that are independent of those obtained previously, ²⁴ we used another set of stimuli with a four LED stimulator. This has a larger dynamic range than the CRT screen used previously and gives rise to better separation of the responses from the rods and the three cone classes. In addition, the stimulus waveform can be better reproduced, and higher luminances can be achieved with the LED stimulator. ^31–36 In addition, we performed two control experiments with which a photoreceptor-based explanation of the data could be excluded. 

The availability of ERGs representing chromatic and luminance activity may present the opportunity to study, with the use of noninvasive electrophysiology in human subjects, the processing of retinal signals that are correlated with visual perception. Furthermore, the experiments may find clinical application because it now may be possible to assess the functional integrity of visually relevant pathways in the retina. 

Methods

Subjects

Six healthy subjects participated in the present experiment. Five subjects had normal color vision. One subject was deuteranopic. The pupil of one eye was dilated with a drop of tropicamide (Mydriaticum; Pharma Stulln GmbH, Stulln, Germany). If requested, a topical anesthetic (Oxybuprocain; Alcon Pharma GmbH, Freiburg, Germany) was administered. The experiments adhered to the tenets of the Declaration of Helsinki and were approved by the institutional ethics committee. Informed consent was obtained from the subjects after explanation of the nature and possible consequences of the study. 

Stimuli

Measurements were performed using an electrophysiological recording system (RetiPort; Roland Consult, Wiesbaden, Germany). Sine wave stimuli were presented using a Ganzfeld bowl (Q450 SC; Roland Consult) with six arrays of differently colored light-emitting diodes (LEDS), one broad-band white (CIE coordinates: x, 0.3715; y, 0.4202) and five-narrow band LED arrays with peak wavelengths 452 (royal blue; bandwidth-at-half-height [bwhh], 18 nm; CIE coordinates: x, 0.1519; y, 0.0381), 469 (blue; bwhh, 22 nm; CIE coordinates: x, 0.1255; y, 0.0926), 523 (green; bwhh, 36 nm; CIE coordinates: x, 0.2016; y, 0.7371), 594 (orange; bwhh, 15 nm; CIE coordinates: x, 0.5753; y, 0.4240), and 638 nm (red; bwhh, 19 nm; CIE coordinates: x, 0.6957; y, 0.2966). 

Luminances of the LEDs were set by input voltage. The LEDs were calibrated by measuring the luminance at approximately 4000 voltage values. Data were stored in a look-up table that was used to define the output in real time during a recording. The relationship between LED luminance and voltage was linear. The waveform, mean luminance, Michelson contrast, and modulation phase of each LED array could be set independently. In the main experiment, only the 638-nm and the 523-nm LEDs were activated. Mean luminance of each of the two LED arrays was 100 cd/m². The pupils of all subjects were dilated to approximately 8 mm, resulting in a retinal illuminance of approximately 10⁴ td. Mean hue was yellow (CIE coordinates: x, 0.5813; y, 0.4030). The luminance of the 638-nm and 523-nm LEDs was sinusoidally modulated in counterphase. In each measurement, the fraction of red (638 nm) modulation R/(R+G) was varied between 0 (only the 523-nm LED was modulated; the 638-nm LED was constant at 100 cd/m²) and 1 (only the 638-nm LED was modulated; the 523-nm LED was constant at 100 cd/m²) in steps of 0.1, with finer steps of 0.05 between 0.7 and 0.3. The total modulation (R+G) was kept constant in all conditions. A sketch of the stimuli is given in the left plots of Figure 1. In total, 15 conditions were used. The condition R/(R+G) = 0.5 was measured twice to check for changes in recording conditions. Measurements were repeated at seven temporal frequencies: 4, 8, 12, 16, 20, 24, and 36 Hz. We calculated the stimulus strength for each of the three cone types and for the rods in terms of cone and rod contrast. In brief, cone contrast is quantified by the Michelson contrast of photoreceptor excitation modulation. To calculate the excitation, the emission spectra of the LED arrays were multiplied with their luminance and with the cone and rod fundamentals and were integrated over wavelength. ^37,38  

Figure 1.

 

Left: sketch of the stimulus configuration in the main experiment. Six different stimuli with varying R/(R+G) values are displayed. Middle and right plots: sections of averaged responses measured in a trichromatic subject (ARR). Total duration of the averaged responses was always 1 second. Responses were measured at 36 Hz (middle column) and 12 Hz (right column). Upper, middle, and lower rows: original responses at an R/(R+G) values of 0, 0.5, and 1, respectively. Clearly, the response amplitudes go through a minimum at 36 Hz. This cannot be observed in the 12-Hz responses. Vertical dashed lines: responses at R/(R+G) values 0 and 1 are in counterphase at 36 Hz, whereas the responses have similar phases at 12 Hz.

Figure 2A shows the photoreceptor contrasts elicited by each of the 15 stimulus conditions. Zero cone or rod contrast means that the stimulus does not modulate the excitation of the particular photoreceptor and thus represents a silent substitution condition. Response phase changes by 180° at this point. Response amplitudes (in arbitrary units) of luminance and red-green chromatic channels are displayed in Figure 2B. These plots assume that S-cone input to the luminance and the red-green chromatic channels is nonexistent (or negligible) and that the mean luminance of 200 cd/m² is too high for substantial rod input. ^39–41 For the red-green chromatic channel, it was assumed that the L- and M-cone–driven responses contribute in a counterphase manner and with a ratio of 1 (Chrom _amp = L − M ^{19,23,30,38,42} ; Chrom _amp is the response amplitude of the red-green chromatic channel). The two cone types contribute additively to the responses of the luminance channel (Lum _amp = aL + M in which Lum _amp is the response amplitude of the luminance channel). The factor a varies between different subjects but is generally larger than 1, which is related to the fact that most trichromatic subjects have more L-cones than M-cones. ^18,19,43 The factor a may also depend on the state of adaptation. ²³ To calculate the response of the luminance channel, we have not included saturation effects because in electroretinographic measurements, saturation effects were not measurable over a large range of stimulus intensities. ^44,45 The temporal frequencies of chromaticity and luminance modulation are identical. 

Figure 2.

 

(A) Cone contrasts (upper) and cone phase (lower) and (B) estimated response amplitude of a luminance and red-green chromatic pathway as a function of the stimulus condition, expressed as R/(R+G), in the main experiment. In the responses of the postreceptoral pathways it was assumed that saturation does not play a role. (C, D) Contrasts and postreceptoral responses in the second control experiment as a function of the stimulus condition (expressed in terms of equivalent R/(R+G) in the main experiment).

Control Experiment

In the main experiment, all four photoreceptor types were modulated (Fig. 2A). To check to what extent the intrusion of rod and S-cone–driven signals might influence the results, the measurements were repeated with three trichromatic subjects using stimuli that did not stimulate the S-cones and the rods. To be able to obtain the appropriate stimuli, the stimulus settings had to be changed. First, in addition to the 638-nm and 523-nm LEDs, the 594-nm and 469-nm LEDs of the stimulator were activated. In that way, the four photoreceptor types could be stimulated independently. Each LED had a mean luminance of 50 cd/m². As a result, the mean luminance was the same as in the main experiment, but the time-averaged chromaticity was different (CIE coordinates: 0.3408, 02443). Second, the L- and M-cone contrasts obtained in the main experiment were beyond the dynamic range of the 4-LED stimulator. Therefore, the contrasts were decreased by a factor of 5, and stimuli equivalent to those with R/(R+G) values lower than 0.35 in the main experiment were omitted. In total, eight stimuli were presented (equivalent to red fractions 0.35, 0.4, 0.45, 0.5, 0.6, 0.7. 0.8, and 1.0 in the main experiment). Resultant stimuli in terms of cone contrast and stimulus strength in the luminance and chromatic channels are displayed in Figures 2C and 2D. 

Electroretinographic Recordings

ERGs were recorded with a DTL electrode positioned on the lower limbus of the eye. The forehead and the ipsilateral temple were cleaned with abrasive gel, and gold cup electrodes filled with electrode paste were positioned. These electrodes served as ground and reference electrodes, respectively. 

In the main experiment, the stimuli were presented for 24 to 48 seconds to obtain a large signal-to-noise ratio (SNR) in most conditions and temporal frequencies. Recordings were averaged 24 or 48 times, depending on the total recording time, so that the averaged signals spanned 1-second episodes. The middle and right plots of Figure 1 display some sections of these averaged recordings in a trichromatic subject measured at 36 and 12 Hz. Stimulus strengths in the control experiment were much smaller (compare the contrasts given in Figs. 2A and 2C). In this case, for each stimulus, between 80 and 160 1-second sweeps were presented to increase the SNR. 

Results

Figure 3 shows the response amplitudes and phases of the first harmonic components as a function of red fraction (R/(R+G)) measured in a trichromatic subject. Higher harmonics may play a role, especially at lower temporal frequencies (see, for example, the double peaks in the 12-Hz data in Fig. 1), but these are not considered here. Results are typical of all data obtained with trichromatic subjects (Fig. 4). The dependence of the response amplitude and phase on the red fraction differed at the six temporal frequencies. At 36 Hz, the first harmonic amplitude of the response goes through a clear minimum at which the phase changes by 180°. This is reminiscent of the estimated luminance response displayed in Figure 2B. Although this response behavior was found for all subjects, the red fraction at which the minimum occurred varied slightly for the different subjects (Fig. 4). This reflects the interindividual variability in the numbers of L- and M-cones. ^18,19,46 As temporal frequency decreased, the minimum in the first harmonic component was not as sharp as at 36 Hz and shifted toward smaller red fractions. Phase changed more gradually. At 12 Hz, the first harmonic response amplitude and phase were relatively constant for all red fractions and resembled the red-green chromatic activity displayed in Figure 2B. At 4 and 8 Hz, the data were noisier because of artifacts from small eye movements. Amplitudes decreased as the red fraction increased, whereas phase did not substantially change. 

Figure 3.

 

Response amplitude (left) and phase (right) of the first harmonic components in the Fourier transform on the electroretinographic recordings given as a function of stimulus condition. Responses were measured in subject ARR. Data are plotted separately for different temporal frequencies. Solid lines: fits with a model that assumes a vector addition of luminance (given by the responses at 36 Hz) and red-green chromatic pathway driven responses (shown in Fig. 2B). Dashed lines: results of a fit with a model that assumes a vector addition of L- and M-cone–driven responses at high temporal frequencies and an additional rod-driven response at low temporal frequency.

Figure 4.

 

Same data as displayed in Figure 3 for three additional subjects. Only the 36- and 12-Hz data are shown.

To illustrate that the data were similar for all trichromatic subjects, we display in Figure 4 the 12- and 36-Hz data for three additional subjects. For all subjects, the response amplitude versus R/(R+G) plot is V-shaped, with a sharp minimum at 36 Hz. As amplitude reaches a minimum, phase changes by 180°. In contrast, amplitude and phase change less strongly as a function of red fraction at 12 Hz. 

To model the data, it was assumed that a combination of independent luminance and chromatic signals, interacting according to a vector addition, determined the ERG (see Refs. 45, 38, 47 for detailed descriptions of the vector addition procedure). Briefly, amplitude and phase of the responses of the luminance and the chromatic mechanisms were converted into vectors, the lengths of which corresponded to amplitude, whereas the angle they made with the positive x-axis represented phase. The vectors were then added to obtain the vector representations of the measured responses. There were four free parameters—the amplitude factors and phase shifts of the two underlying mechanisms—in the fitting procedure. These parameters were used to adjust absolute amplitude and phase to describe the measured responses optimally. The best fits were defined as the least square distances in this vector space. It was assumed that, at 36 Hz, the responses were fully determined by luminance activity, without intrusion by any other mechanism. To account for the above mentioned interindividual differences in the luminance channel, the curve measured at 36 Hz was used as a template to describe the luminance activity in the same subject at the other temporal frequencies. The calculated output of the red-green chromatic activity as displayed in Figure 1B was used as a template for its activity in the vector addition model. Templates of the luminance and chromatic activities had different absolute amplitudes. Therefore, the amplitudes of the putative luminance and chromatic channels in the fits cannot be compared in absolute terms. The lines in Figures 3 and 4 are fits of the data with this vector addition. 

In Figure 5, the means of the estimated amplitude and phase of the six observers are displayed as a function of temporal frequency. The frequency dependency of the parameters was similar in all subjects. The response of the putative luminance channel goes through a minimum at 12 Hz. Although a small luminance component can still be identified in some subjects at this frequency, the signals are determined primarily by red-green chromatic activity. At 4 and 8 Hz, the amplitude of the luminance-driven signal increases again. The phase of the luminance response is relatively stable. The amplitude of the red-green chromatic channel is low-pass. Furthermore, there is a monotonic increase in phase as temporal frequency decreases. 

Figure 5.

 

Mean (±SD) of amplitudes and phases of the luminance and red-green chromatic pathway-driven signals obtained from the fitted model to the data of all subjects. Dashed lines: results for the luminance-driven response in the deuteranopic observer (derived from the data given in Fig. 6).

As shown in Figure 2A, rods and S-cones are also substantially modulated in most of the stimulus conditions. Despite the photopic conditions, it cannot be excluded that the data result from a weighted activity originating in the four photoreceptor types instead of the two postreceptoral channels. With photoreceptor-driven responses we do not suggest that the electroretinographic response directly originates in the photoreceptors themselves but that the ERG-mediating mechanism reflects photoreceptor excitation. Previously, we found that rod- and S-cone–driven responses cannot always be neglected, even when the responses are performed at high retinal illuminance. ³⁴ This is supported by psychophysical studies. Rod activity can suppress cone sensitivity at high temporal frequencies. ³⁶ At low temporal frequencies, rod and cone signals show linear summation when mediated through the same pathway and probability summation when mediated through different pathways. ⁴⁸ To check whether an interaction between the photoreceptor-driven signals can also explain the data, we modeled the data once more by vector adding the 36-Hz data with a rod-driven signal. There were two assumptions. First, the data at 36 Hz are exclusively driven by L- and M-cones. As explained in the introduction, the similarity with a luminance system would then not be the result of a causal relationship. It would occur because both the electroretinographic and the luminance pathways draw their input from all available L- and M-cones in the retina, with small phase differences between the L- and M-cone–driven signals. As a result, both reflect the ratio of L- to M-cone numbers. This ratio does not change when temporal frequency is altered; the interaction between L- and M-cones can be described by the 36-Hz data allowing an overall change in amplitude and phase. Second, we only considered the rods because the responses of rods and S-cones were similar with the stimuli used (Fig. 2A). Results of this model are shown as dotted lines in Figure 3. Overall, the model fits with rod responses were also satisfactory, although the goodness-of-fit (defined as the sum of squared distances in vector space) were generally poorer (because the sum of squared distances was larger), especially at frequencies between 8 and 16 Hz, indicating the fits to be slightly worse. 

To provide further evidence that the ERGs reflected chromatic and luminance activity rather than those of pathways in which all photoreceptor responses, including those of S-cones and rods, converge, we performed two control experiments. In the first experiment, we performed the same measurements in a deuteranopic subject who lacks M-cones and a red-green chromatic channel. If the ERGs reflect luminance and chromatic activity, we would expect a large electroretinographic response difference in this subject in comparison with the trichromatic subjects, especially when the red-green chromatic channel is supposed to mediate the ERG in the trichromatic subjects. In contrast, if the low-frequency ERGs were driven largely by rods, S-cones, or both, we would expect that the electroretinographic data in the deuteranopic subject would not differ much from those of the trichromatic subjects. 

Results of the measurements with the deuteranopic subjects are shown in Figure 6. It can be seen that response amplitudes and phases depend in a similar manner on R/(R+G) at nearly all temporal frequencies. All curves were, therefore, fitted with a scaled and phase-shifted version of the 36-Hz (luminance) template that is similar to the response characteristics of the M-cones, without intrusion of rod- or S-cone–driven signals. The fits generally gave satisfactory results. ERGs measured at 4 and 8 Hz were noisier than those measured in the trichromatic subjects. Deviations from a scaled 36-Hz response, especially near zero values of R/(R+G), suggest that intrusion of rod or S-cone signals, or both, at these temporal frequencies may play a significant role. Estimates of the free parameters are shown as dashed lines in Figure 5. Obviously, they have characteristics similar to those of the trichromatic luminance-driven electroretinographic response. In summary, electroretinographic responses in deuteranopic subjects are probably driven by only one mechanism, consistent with the absence of a red-green chromatic mechanism. S-cone– or rod-driven signals are relatively small. 

Figure 6.

 

Amplitude and phase data measured in a deuteranopic subject. Data at the 30-Hz and lower frequencies are fitted with the scaled and phase-shifted responses recorded at 36 Hz. Amplitude scaling factors and phase shifts are given in the upper plots of Figure 5.

In the second control experiment, measurements with the same ratios of L- and M-cone contrasts were used in combination with silent substitutions for the rods and the S-cones. Apart from an overall decrease in contrast, the red-green chromatic and luminance channels were expected to display the same response behavior as in the main experiment (Fig. 2D). Thus, in case of photoreceptor-driven electroretinographic signals, amplitude and phase can be expected to differ strongly in comparison with the main experiment. ERGs reflecting luminance and chromatic activity should show response characteristics similar to those measured in the main experiments. As described in the methods section, in these measurements the 594-nm and 469-nm LEDs were also activated. The measurements were performed in three subjects. 

Results for one representative subject are shown in Figure 7. As mentioned in the methods section, we used lower contrasts and stimuli with equivalent R/(R+G) values above 0.35 to stay within the dynamic range of the stimulator. The data show important similarities with those from the main experiment. There is a substantial change in response amplitude at 36 Hz, whereas the electroretinographic amplitudes are constant at 12 Hz. 

Figure 7.

 

Response amplitude and phase of the first harmonic components as a function of equivalent R/(R+G) in the second control experiment measured in subject JK.

As in the main experiment, the data were fitted by a vector addition of the putative luminance response and the response of the red-green chromatic channel. Estimated parameters are shown in Figure 8. Although the estimates are more variable, because of the lower signal-to-noise ratio, they are similar to those obtained from the main experiment (see Fig. 5), suggesting that the same processes underlie the responses in the two experiments. 

Figure 8.

 

Mean (±SD) estimated amplitudes and phases of the luminance and chromatic mechanisms obtained from the model fits to the responses in the second control experiment.

Discussion

The electroretinogram is a powerful tool for studying the functional integrity of the retina in a noninvasive manner. Therefore, it is often recorded in clinical practice. In addition, correlations with photoreceptor activities can be obtained relatively easily because photoreceptor activity directly determines a large part of the a-wave and drives other electroretinographic components that originate in postreceptoral neurons. It is generally difficult to correlate electroretinographic responses with activity in the luminance and chromatic pathways, though electroretinographic signals with postreceptoral origins have been described in humans and nonhuman primates. ^1,47,49–52 Recently, we found indications of a separation between electroretinographic signals representing luminance and chromatic activity at high temporal frequencies and near 12 Hz, respectively. ²⁴ In those experiments, a CRT screen was used as stimulator. For the experiments shown in the present study, an LED stimulator with important advantages (see the Introduction) was used. 

Results of the present experiments confirm and further strengthen the original proposal. The ERGs measured in the deuteranopic observer reflect primarily luminance (or L-cone) activity at all temporal frequencies. This is in agreement with our hypothesis because the deuteranopic subject lacks a red-green chromatic channel. Rod- and S-cone–driven signals might have been present in the responses to 4- and 8-Hz stimuli with R/(R+G) close to zero. The control experiment in which stimuli were applied but did not stimulate the rods and the S-cones and in which the luminance and red-green chromatic signals were varied in the same manner as in the main experiment also indicated that the influence of temporal frequency on the electroretinographic data cannot be attributed to intrusion of rod and/or S-cone driven electroretinographic signals but rather reflect luminance and chromatic activity. 

We performed independent electroretinographic experiments in human subjects in which the modulated red and green LEDs were sequentially presented and had the form of raised cosine waves (Parry N, unpublished data, 2008). By using raised cosines (instead of simple sine-wave stimuli, as used in the experiments in the present study), the luminance modulation in these experiments was at twice the temporal frequency than the red-green chromatic modulation so that for each stimulus the response to the luminance content and the chromatic content could be separated on the basis of the Fourier spectrum. The advantage, in comparison with the data presented here, may be that the electroretinographic responses reflecting chromatic and luminance activity can be obtained on the basis of only one stimulus and not on the basis of a series of responses in which R/(R+G) values are varied. The data show that, in agreement with our data presented here, a response to the chromatic component in the stimulus can be observed only at low temporal frequencies. This is an additional indication of the existence of luminance and chromatic reflecting electroretinographic responses in human observers. 

The results of other experiments can also be explained on the basis of this idea. A change in cone weightings in multifocal ERGs and multifocal visual evoked potentials has been described ^25–27 indicating that the L/M ratio can depend on the stimulus and recording technique. As mentioned in the introduction, we found previously that a change in chromatic adaptation changed the L/M ratio in the ERG measured at 30 Hz and in luminance-mediated psychophysical flicker detection thresholds of L- and M-cone isolating stimuli, ²³ from which it can be concluded that the spectral sensitivities of the psychophysical luminance channel and the high-frequency flicker electroretinography are similarly altered by chromatic adaptation and therefore share postreceptoral adaptation mechanisms. 

Although we have discussed a relationship between flicker ERGs and psychophysical data, we have not performed psychophysical experiments with the stimuli described in this article. Measuring flicker detection thresholds at different temporal frequencies for different R/(R+G) values would provide additional information about the correlation between the ERGs and the different postreceptoral psychophysical channels. Similar psychophysical measurements for the luminance channel have been performed, ⁵³ and the results are qualitatively similar to the electroretinographic data at high temporal frequency. It would be interesting to repeat the psychophysical experiments at temporal frequencies known to favor the chromatic channel for comparison with the 12-Hz ERGs. 

We previously described a frequency-dependent change in the responses to L- and M-cone–specific CRT stimuli and proposed the intrusion of a rod-driven mechanism with low-pass characteristics to explain the electroretinographic response changes. ⁵⁴ The data presented here confirm those data and indeed suggest the presence of a low-pass mechanism. However, on the basis of our present data and those presented in our recent paper, ²⁴ we propose that this low-pass mechanism reflects chromatic activity rather than rod-driven responses. 

An intriguing finding is that the estimated frequency dependency of the amplitude and phase of the putative luminance mechanism found in the present study (Fig. 5, upper plots) closely resemble previously published data on the electroretinographics response amplitudes and phases to luminance flicker that were performed in monkeys and human observers ^47,52,55 because they all show a minimum in the amplitudes at frequencies between 10 and 20 Hz. At the same temporal frequencies, a phase decrease can be observed. This suggests that the electroretinographic responses might have reflected activities of the same mechanisms in the two experiments. The absence of any chromaticity in the stimuli used in the previous experiments is in agreement with the notion that the ERGs reflect luminance activity. 

The small amplitude of the luminance response in the first harmonic component may explain why the response at 12 Hz can reflect chromatic activity even in the presence of a substantial luminance component in the stimulus. This does not mean that a luminance response is completely absent. First, a residual response is still present (reflected by the small but non-zero amplitude of the luminance component in Fig. 5), the amplitude of which may vary between individuals. Second, we have not considered higher harmonic components in our analysis. Obviously, these higher harmonic components may play an important role in the responses (e.g., resulting in double peaks in the 12-Hz response, as shown in Fig. 1) and may reflect the activity of additional retinal pathways. Further analysis of the harmonic components is needed for identification of these pathways. 

The control experiments showed that the basic effect is also present when rods and S-cones are not stimulated, indicating that the measured effects cannot be attributed to S-cone or rod-driven responses. Nevertheless, in these conditions, some response properties were different from those measured in the main experiment. For instance, the results indicate that the stimulus conditions required for a minimal response at high temporal frequency was different (Figs. 3, 7). We propose that this might have been caused by the different mean chromaticities of the stimuli used in the two experiments. As mentioned, mean chromaticity can have a large influence on the L- to M-cone weighting ratio at high temporal frequency responses. ²³ In the experiments described, a change in the L/M cone weighting ratio results in a change of stimulus red fraction for a minimal response. 

In conclusion, on the basis of the data presented here and previously, ²⁴ we are confident that it is indeed possible to study luminance and chromatic activity in the human ERG. These results will be helpful to study the retinal properties of these pathways in an objective and noninvasive manner in human observers. For instance, using a stimulus presentation, as described here, combined with a spatial variation (e.g., checkerboards for pattern ERGs instead of a spatially homogeneous stimulus) could possibly reveal the spatial properties of the two pathways and the influence of retinal eccentricity on the two. In addition, if the two mechanisms are differently affected by a retinal disorder, these effects may now possibly be studied. 

Footnotes

 Supported by German Research Council (DFG) Grant KR 1317/9-1, CAPES-DAAD PROBRAL Grant 182/2007, and CNPq Grant 550671/2007-2. JK is a Fellow in the Excellence Program of the Hertie Foundation. ARR had a CAPES fellowship for graduate students. LCLS and MSF are CNPq research fellows.

Footnotes

 Disclosure: J. Kremers, None; A.R. Rodrigues, None; L.C. de Lima Silveira, None; M. da Silva Filho, None

The authors thank Neil Parry and Andrew Zele for their comments on the manuscript.