1. INTRODUCTION

It is well established that there are at least three pathways carrying achromatic and chromatic information from the retina to the cortex. These pathways are derived from different cone-signal combinations to form the so-called, red–green (or parvocellular), blue–yellow (or koniocellular), and luminance (or magnocellular) pathways. The three cone types connect to horizontal and bipolar cells to form the substrate for excitatory-inhibitory spatio-temporal receptive field organization, allowing the visual system to respond to both increments and decrements of light. The magnocellular pathway, which is mainly responsible for the transmission of luminance information, receives input from both L- and M-cones that synapse with horizontal cells (namely H1 horizontal cells) and diffuse bipolar cells. The diffuse bipolar cells have a center-surround receptive field with excitatory-inhibitory properties that are formed through the sign-inverting H1 cells. This configuration is the initial stage of the neural substrate for detecting increments and decrements of luminance information.

Luminance information is simultaneously processed by the magnocellular and parvocellular pathways. De Valois and De Valois [1] described how much of the luminance (achromatic) information is multiplexed with chromatic information in the parvocellular pathway [2], although the magnocellular pathway can account for the classical luminosity function. The output of initial stages can be used by different pathways for different processing purposes as described in the model [2], and luminance information can directly influence color perception through blackness induction causing the appearance of brown, gray, and black [3,4]. Luminance information modulated at higher frequencies and speeds, however, is signaled by transient responses [5] and is considered to be processed in the magnocellular pathway. The magnocellular pathway, which mainly receives parafoveal and peripheral input, is critical for transmitting motion information, low spatial frequencies, and high temporal frequencies. De Lange [6] and Kelly [7] studied the temporal sensitivity of the visual system showing a band-pass shape with a peak at $\sim 10\mathrm{Hz}$ (for 7 Troland conditions). While these psychophysical experiments offer valuable information about the response dynamics of the visual system to temporal luminance modulation, they do not provide any insight into the different increment and decrement luminance response dynamics that have long been recognized as likely mediated by separate systems: “Die Neurophysiologie der Hell- und Dunkelwahrnehmung ist am besten durch ein Dualitätsschema mit zwei antagonistischen Neuronsystem für Licht- und Dunkelaktivierung verständlich zu machen” (Jung [8], p. 71) (see also Ehrenstein and Spillmann [9] and review by Fiorentini et al. [10]).

The magnocellular pathway has been studied extensively both psychophysically and physiologically. Rekauzke et al. [11] studied the V1 signal generated by neighboring black and white stimuli on a cat’s retina and found an asymmetry in the temporal responses to incremental and decremental changes in luminance. They showed that the activity that is generated by a decremental stimulus arrives at an earlier time at V1 than the activity that is generated simultaneously by an incremental stimulus. The authors suggested that this neural temporal imbalance of simultaneous increment and decrement luminance stimulation might be the basis for motion processing. Psychophysically, it has also been shown that the latencies of the decrement luminance signal are shorter compared to the increment signals [12,13] and that the decrement detection thresholds are lower when compared with the increment thresholds [14,15]. Li et al. [16] even found that the OFF responses to increments and decrements of luminance (i.e., the afterimages) show similar asymmetries as the ON responses, both with latency and amplitude; darker afterimages elicited a higher firing rate at the cat’s dLGN with shorter latency compared to brighter afterimages.

The detection of increments or decrements of lights and colors has been studied using stimulus-onset asynchrony to derive the impulse response function (IRF) of luminance and chromatic systems. Burr and Morrone [17,18] used increment luminance and color signals to study the increment response of the magnocellular and parvocellular pathways and showed that the former has a biphasic IRF while the latter is monophasic. Shi and Shinomori [19] studied the IRF to increments and decrements of light among young observers and found that the temporal profile of the IRF is similar between increment and decrement conditions, being unable, however, to measure timing differences smaller than the refresh rate of their equipment. The temporal similarity they found between the increment and decrement detection thresholds is supported by retinal studies in cats and macaques that showed that the temporal responses of the midget and parasol ganglion cells are similar between the two classes of stimuli [20–22]. Shi and Shinomori [19] also found that the threshold for decrements of light is lower than increments of light, meaning that the visual system is more sensitive to a luminance decrement than to an increment of light. This asymmetry might be explained by differences in neural populations in cortical area V1 rather than the retina. Xing et al. [23] found a larger population of V1 neurons that responded to black compared to neurons that responded to white in macaque monkeys, reinforcing the idea of a cortical contrast amplification mechanism.

These asymmetries found between the increment and decrement responses of the magnocellular pathway have implications for processing motion, adaptation, and contrast. Moreover, all of these functions change with age. Schefrin et al. [24] found a reduction in dark-adapted contrast sensitivity at low spatial frequencies suggesting that there might be aging processes that affect the responses of the magnocellular pathway. McKendrick et al. [25] showed that there is again some loss in the magnocellular pathway’s sensitivity to contrast in humans, while Yang et al. [26] showed that there are aging processes that affect the magnocellular pathway in macaque monkeys.

Shinomori and Werner [27] studied the response to a brief, achromatic, incremental pulse in order to derive the underlying impulse response across aging. They found that the first excitatory phase of the impulse response of a putative magnocellular pathway remains stable throughout the life span while there is a reduction in the response of the (second) inhibitory phase of the impulse response of the pathway, attributing these differences to neural rather than optical (primarily lens absorption) factors. Little is known however about the pathway’s responses to a decremental stimulus and how these responses change with age, although for an S-cone pathway, Shinomori and Werner [28] have demonstrated greater losses in amplitude of the IRF for decremental than incremental stimuli.

2. METHOD

A. Observers

Twelve young observers (7 male and 5 female) from 19 to 24 years old (mean 21.5 years, $\mathrm{SD}=1.8$) and ten older observers (6 male and 4 female) from 65 to 85 years old (mean 73.5 years, $\mathrm{SD}=6.9$) participated in this study. A certified ophthalmic technician performed an eye exam to exclude subjects with visual acuity worse than 20/25, intraocular pressure $>22\mathrm{mm}$ Hg, abnormal pupil responses, and ocular motility. An anterior and dilated-posterior examination was performed by an ophthalmologist or optometrist to exclude participants with any pathological conditions. Digital fundus photographs (TRC.501X Mydriatic Retinal Camera; Topcon Medical Systems, Inc.) were reviewed by a retinal specialist to eliminate individuals with retinal or optic nerve abnormalities. Significant lens opacity (or cataract) would be detected on clinical examination, and these participants were excluded from the study. A clinically significant opacity of the lens would be inconsistent with the inclusion criterion of 20/25 visual acuity or better. Each subject’s color vision was tested with the D-15 hue arrangement test, the F2, and HRR plates, the Neitz anomaloscope, and the Cambridge Colour Test (CCT). While Protan, Deutan, and Tritan CCT thresholds were higher for older observers, their values were in the color normal range for their ages, as described in our previous study [29]. All observers were classified as normal trichromats by these tests.

All procedures and experiments conformed to the principles expressed in the Declaration of Helsinki and were approved by the UC Davis Medical Center’s Institutional Review Board. Written informed consent was obtained from each subject before participating in the study.

B. Stimulus

We used the double-pulse method in which thresholds were measured for two flashes presented at the same position with varying stimulus onset asynchrony (SOA) [30]. The stimulus consisted of two incremental flashes or two decremental flashes in luminance on a $15\mathrm{cd}/{\mathrm{m}}^2$ (1.80 log Td.) background. Both flashes and background had the same chromaticity (CIE $x,y=0.33,0.33$). The flashes were disks of 2.26 deg. diameter filtered with a two-dimensional Gaussian envelope at 1 SD to eliminate artifacts caused by edges. The two flashes were presented in one of four quadrants defined by a central fixation cross. The positions of the test stimuli were located 1.70 deg. to either the left or right of the fixation cross and 1.70 deg. above or below the fixation cross.

The first flash was presented in a single frame on the cathode-ray tube (CRT) monitor (EIZO T566, EIZO Inc.) with a refresh rate of 150 Hz, controlled by a video board with 15-bit resolution in each phosphor’s intensity (VSG 2/4, Cambridge Research Systems Inc.). SOAs were also defined with a certain number of frames between flashes. The fourteen SOAs used in this study were: 13.3, 26.7, 33.3, 46.7, 53.3, 66.7, 73.3, 86.7, 93.3, 106.7, 126.7, 146.7, 166.7, and 186.7 ms.

The two flashes of the double-pulse stimulus always had the same luminance. We measured positive contrast thresholds for incremental flashes and negative contrast thresholds for decremental flashes, defined by the peak luminance ($I_T$) of the flash’s Gaussian envelope against the background intensity ($I_B$). To equally compare positive and negative contrast, the positive contrast ($C_P$) and the negative contrast ($C_N$) were defined by Eqs. (1) and (2) [19]. The negative contrast was transformed to its mirror value along the background intensity axis as shown in Fig. 1,

Luminance and chromaticity coordinates were adjusted individually by heterochromatic flicker photometry (HFP), in which the intensity of the blue and green phosphors were equated to the red phosphor (which is minimally affected by the lens absorption) to minimize the effect of age-related changes in the optical density of the ocular media. The temporal frequencies used were 12.5, 15.0, or 18.8 Hz. The stimulus was an annulus of 0.64 and 2.77 deg. inner and outer diameter, respectively, surrounding the central fixation cross. This annulus coincided with the center of the four flashes.

 

Fig. 1. Temporal profile of positive contrast for incremental flashes (left) and negative contrast for decremental flashes (right) in the double-pulse method.

Download Full Size | PDF

We used an astronomical telescope through which the subject viewed the CRT at optical infinity with a $2.16\times$ magnification with a 5.0-mm aperture stop positioned in front of the telescope’s objective lens. The aperture stop formed an image of 2.31 mm at the plane of the subject’s pupil, acting as an artificial pupil and ensuring equal retinal illuminance for all subjects. An adjustable chair and a dental-impression bite-bar assembly mounted to a milling-machine table were used to maintain the subject’s eye position. An experimenter used an auxiliary optical system with a reticule aligned with the optical axis of the telescope to permit precise alignment of the subject’s pupil.

The monitor was calibrated with a spectroradiometer/photometer (Photo Research, Model PR703-A) and a chromameter (Minolta, CS-100), producing measurement errors in luminance and chromaticity $<5\%$ and 3%, respectively. Phosphor excitation (response time) was about 1.2 ms for all phosphor guns. The diameter of the 1SD Gaussian patch was 106 pixels on a $640\times 480$ pixel display, and therefore the decay of the test stimulus at the vertical scan frequency was less than 1.5 ms from the maximum as measured by a photodiode (United Detector technology, Optometer 81, 10 mm diameter) and oscilloscope. The peak-to-peak timing error of the inter-stimulus intervals was less than 3%. Other details of apparatus and calibration have been described elsewhere [27].

C. Procedure

HFP was performed in the first experimental session, where we obtained more than six isoluminance points for each of the combinations of the blue, green, and red phosphors. These results of the HFP were used to adjust the intensity of the three phosphors for each subject.

After 5 min. of dark adaptation and 10 min. of adaptation to the screen background, observers were presented with the test stimuli preceded by a high pitch tone and followed by a low pitch tone. The observer was asked to press one of four buttons after the low pitch tone to indicate the quadrant in which the stimulus was detected. This four-alternative forced-choice (4AFC) was combined with a two-down, one-up staircase changing the luminance of the two flashes to obtain a contrast threshold. Thresholds were based on the last four of six reversals, corresponding to a probability of 70.7% detection. The 14 SOA settings were tested with 14 different staircases in a pseudo-random order. Thresholds for incremental and decremental flashes were measured in separate sessions. Each observer participated in at least five test sessions for each condition.

D. Derivation of the Impulse Response

We calculated the impulse response function (IRF) from the threshold data using Eq. (3) for an exponentially damped frequency-modulated sine wave as proposed by Burr and Morrone [17],

The fitted functions were based on a model of probability summation [33]. The proportion correct, $p(\tau )$, is described as

In the model fit, parameters of the IRF ($a_0-a_3$) were estimated by a nonlinear regression analysis to obtain the best fit to threshold data points for each SOA using the least-squares method. We started the optimization of parameters from at least three different sets of initial parameter values. It was occasionally found, however, that two different IRFs could be fitted about equally well since phase information of the impulse response function is lost in the double-pulse method. In those cases, we selected the IRF that was closest to the mean because the shape of the other IRF always looked unrealistic (e.g., low amplitude of excitation with high amplitude of inhibition) compared to physiological measures. Estimation of temporal resolution at the peak time of an IRF was limited by a one-time step size, $\mathrm{\Delta }t$ (1.333 ms). The accuracy of the peak amplitude (intensity) was also affected by the temporal resolution.

3. RESULTS

A. Contrast Thresholds and IRFs

The shape of the luminance IRF was calculated from the thresholds of two pulses in different SOAs. Left panels of Fig. 2 show the data from one younger (24 years old) and one older (84 years old) observer as typical results of the experiment. In these panels, open circles denote the mean of positive luminance contrast thresholds for incremental flashes and filled circles denote the mean of negative luminance contrast thresholds for decremental flashes, and error bars are $\pm 1$ standard error of the mean (SEM). Right panels of Fig. 2 show the luminance ON-IRF calculated from the contrast thresholds for incremental flashes (denoted by solid curves) and luminance OFF-IRF calculated from thresholds for decremental flashes (dotted curves). The OFF-IRF should theoretically have a first negative phase followed by a second positive phase because a decrement stimulus induced the OFF-impulse response. In this study, however, we describe the OFF-IRFs as having a first positive phase and a second negative phase to facilitate comparison between ON- and OFF-IRFs as defined in Eqs. (2)–(5).

 

Fig. 2. Threshold data (left panels) and luminance ON- and OFF-impulse response functions (right panels) for one younger (24-year-old, top panels) and one older (84-year-old, bottom panels) observer. Open and filled circles denote thresholds for luminance increments and decrements, respectively. Error bars show $\pm 1$ SEM. Solid and dotted curves are model fits to increment and decrement thresholds (left panels) and ON-IRFs and OFF-IRFs (right panels), respectively.

Download Full Size | PDF

As shown in the left panels of Fig. 2, thresholds for both incremental and decremental flashes were relatively low for short SOAs and increased around 50 ms of SOA. These thresholds do not change monotonically after their peak at $\sim 50\mathrm{ms}$. This non-monotonic threshold reduction yields an IRF that has more phases than the first positive and the second negative phases, as shown in the right panels where a third positive phase follows the second negative phase, as reported earlier [27].

The temporal profiles defined by the time to the peak and trough of the first and second phases are essentially the same between ON- and OFF-IRFs and observers’ age groups. Contrast thresholds were higher in the older than the younger observers. This elevation of thresholds is reflected in the lower amplitude of the first phase and the second phase in the IRF shape. Although the IRF amplitude is almost the same between ON- and OFF-IRFs for each observer, the amplitude is much higher in the younger observer’s IRF than the older observer’s IRF.

Figure 3 shows the average peak times (left) and peak amplitudes (right) for each observer group and each ON- or OFF-IRF. The error bars are $\pm 1$ SEM. Table 1 shows the numeric data. No statistically significant difference is found in the peak times between observer groups and between ON- and OFF-IRFs. Note that despite a greater range of ages within groups, the SEMs for peak times are almost identical. The peak amplitudes shown in the right panel are almost the same between ON- and OFF-IRFs for the same age group, however, they are different between younger and older observers. The peak amplitude of the first positive phase becomes significantly reduced with age for both the ON- and OFF-IRFs ($p<0.01$ in ON-IRF and $p<0.001$ in OFF-IRF by a two-tailed Mann–Whitney U test). On the contrary, the peak amplitude of the second negative phase is not statistically different between older and younger observers.

 

Fig. 3. Mean peak time (left panel) and mean peak amplitude (right panel) for luminance ON- and OFF-IRFs for younger and older observers. Error bars are $\pm 1$ SEM. For each condition, the left bar denotes the peak of the positive phase, and the right bar denotes the peak of the negative phase. Asterisks denote statistically significant differences; the peak amplitude of the first positive phase is reduced significantly with age for both the ON- ($p<0.01$) and OFF-IRFs ($p<0.001$).

Download Full Size | PDF

Table 1. Mean of Peak Time and Peak Amplitude in First and Second Phase^a,b

View Table

B. Individual Difference in IRFs

We compared the peak time and the peak amplitude of the ON- and OFF-IRFs for each observer. Figure 4 shows the comparison of the first and second peak time for all observers. Open triangles and gray circles denote younger and older observers’ data points, respectively. For both younger and older observers, the peak time of the first phase falls within a small range, and the peak time of the second phase falls within a wider time range. These results indicate that incremental and decremental stimuli are processed at the same speed for all observers.

 

Fig. 4. Comparison of the first (positive) and second (negative) peak time between luminance $(\mathrm{L}+\mathrm{M})$ ON- and OFF-IRFs for all observers. Open triangles and gray circles denote younger and older observers’ data points, respectively.

Download Full Size | PDF

In Fig. 5 we compare the first positive peak amplitude (top panel) and the second negative peak amplitude (bottom panel) of the ON- and OFF-IRFs for all observers. The correlation between the first positive peak amplitude of the ON- and OFF-IRFs is statistically significant ($p<0.001$, $r=0.92$). The first phase amplitude of younger observers in both ON- and OFF-IRFs is significantly larger than the amplitude of older observers ($p<0.01$), as also shown in Fig. 3. The correlation is significant for the second negative peak amplitude between the ON- and OFF-IRFs ($p<0.001$, $r=0.74$). However, the second negative phase amplitude is not statistically different between younger and older observers. These results suggest that the amplitude of luminance ON- and OFF-IRFs decreases with age in the first positive phase but not in the second negative phase. The significance of age-related changes does not follow our previous aging study on ON-IRF [27]; age-related reduction of the first and second phase amplitudes were significant ($p<0.001$, and $p<0.01$, respectively). It is not a surprising difference, however, because as shown in Fig. 6 of the previous study [27], the observer’s variance in the same age range was much larger in the second phase amplitude compared to the first phase amplitude.

 

Fig. 5. Comparison of the first positive peak amplitude (top panel) and the second negative peak amplitude (bottom panel) between luminance $(\mathrm{L}+\mathrm{M})$ ON- and OFF-IRFs. Open triangles and gray circles denote younger and older observers, respectively. Solid lines are the regression lines for all data points.

Download Full Size | PDF

 

Fig. 6. Comparison between negative contrast thresholds for decremental flashes and positive contrast thresholds for incremental flashes for each SOA for twelve younger observers (top panel) and ten older observers (bottom panel). Crosses denote the six and four data sets of the younger and older groups, respectively, that showed no significant difference between positive and negative contrast thresholds (see text for explanation). Open and filled symbols show individual observers’ data sets for which positive contrast thresholds are significantly larger or smaller than negative contrast thresholds, respectively.

Download Full Size | PDF

C. Comparison Between Positive and Negative Contrast Thresholds

We showed that the first and second peak amplitudes were almost identical between ON- and OFF-IRFs for each observer. In Fig. 6, we compare positive luminance contrast thresholds for incremental flashes and negative luminance contrast thresholds for decremental flashes for each observer in our study at all SOAs for investigation of possible sensitivity differences between ON- and OFF-pathways. In both age groups, the correlations between the negative and positive contrast thresholds were statistically significant [$p<0.001$, $r=0.93$ for young observers (Fig. 6 top panel) and $p<0.001$, $r=0.79$ for old observers (Fig. 6 bottom panel)].

We used a Wilcoxon signed-ranks test (two-tailed) to compare the sets of positive and negative contrast thresholds for each SOA (14 data points for one observer). In twelve younger observers, five had significantly higher positive contrast thresholds for incremental flashes ($p<0.05$) as denoted by open symbols in the top panel of Fig. 6, and one observer had significantly lower positive contrast thresholds ($p<0.01$) as shown by filled circles. The other six younger observers showed no significant difference between positive and negative contrast thresholds (denoted by crosses in the top panel of Fig. 6). In ten older observers, four observers had significantly higher positive contrast thresholds ($p<0.05$) as denoted by open symbols in the bottom panel of Fig. 6, and two observers had significantly lower positive contrast thresholds ($p<0.05$) as shown by filled symbols. The other four older observers showed no difference from the group mean (denoted by crosses in the bottom panel of Fig. 6). This analysis indicates that the balance of sensitivities between luminance ON- and OFF-pathways is different between individual observers, although there is no difference in the average sensitivities of our cohort between ON- and OFF-pathways for the two age groups.

4. DISCUSSION

A. Age-Related Changes in Luminance ON- and OFF-IRFs

The results indicate that the temporal profile of luminance ON- and OFF-IRFs is stable with age. The amplitude of the first positive phase of the ON-IRF and the first negative phase of the OFF-IRF, however, is reduced in the older observers’ group, reflecting higher thresholds in the detection of the flashes. Additionally, the data suggest that the mechanisms of the ON- and OFF-luminance pathways show the same temporal and amplitude characteristics regardless of age.

We calculated average ON- and OFF-IRFs from all younger and older observers. For this calculation, we used the mean time and the mean amplitude at the peaks of the first positive phase and the second negative phase shown in Table 1. The average IRFs are shown in Fig. 7 (top) where open squares show the ON-IRFs and filled triangles the OFF-IRFs. Black solid and dotted lines are the IRFs of the younger group, while gray solid and dotted lines show the IRFs for the older group of observers. The four IRFs are the best fit for a set of two data points (amplitude and time for the first positive and second negative phases). As expected from the mean values, the IRF timing is almost identical between younger and older observer groups. The only significant difference between the two groups is that the first phase amplitude of younger observers’ IRF is higher than that of the older observers. The bottom panel of Fig. 7 shows the temporal contrast sensitivity function (tCSF) calculated from the IRFs. All tCSFs have a band-pass shape, consistent with previous literature [7]. The peak temporal frequencies of the tCSF are 9.7 Hz (young ON-IRF), 8.7 Hz (old ON-IRF), 8.9 Hz (young OFF-IRF) and 9.2 Hz (old OFF-IRF). These data were obtained at $15\mathrm{cd}/{\mathrm{m}}^2$ (63.1Td) and are comparable to peak temporal frequencies of 12.7 and 6.8 Hz measured at 77 and 7.1 Td, respectively, by Kelly [7].

 

Fig. 7. Top panel: average luminance ON- and OFF-IRFs for younger and older observers, as calculated from the mean of the first and second peaks. Open squares and filled triangles denote mean peak points of ON- and OFF-IRFs, respectively. Error bars indicate ±1 SEM. Horizontal error bars are not visible as they were small. Bottom panel: temporal contrast sensitivity function (tCSF) calculated from the IRFs. Solid and dotted curves denote IRFs and tCSFs of ON- and OFF-IRFs, and black and gray curves show the functions of younger and older observers, respectively. The tCSFs were normalized to the peak sensitivity of the OFF-IRF of the younger observers.

Download Full Size | PDF

The temporal profiles of chromatic pathways do not show the second negative phase in their IRF, and they have a slower oscillation speed [28,34–38]. The IRFs of this study show a second negative phase, and they are faster compared to the chromatic IRFs, in accordance with previous studies [27,35]. These findings suggest that the detection of the flashes is mediated by the magnocellular pathway. Luminance information is also processed by the parvocellular pathway. However, the fast IRF oscillation under our conditions suggests that the signal is processed by a mechanism with transient responses rather than sustained [5]. The fact that the stimulus size does not influence the peak frequencies in the tCSFs is described in previous research using Ganzfeld flicker (68 deg.) [7], and also supports the notion that the stimuli were detected by the magnocellular pathway. The similar temporal profiles between ON- and OFF-pathways suggest that the detection of stimuli occurs at similar speeds, regardless of the stimulus being an increment or a decrement compared to its background. Under conditions of natural viewing this becomes especially important for detection of fast-moving objects that may pose a threat (i.e., moving cars). Additionally, the ability to process decremental or incremental stimuli with similar speeds does not change with age. This suggests that older observers (at least for the age range of our cohort) have the same processing speed as younger observers when it comes to contrast-adjusted stimuli containing luminance information.

The reduction of sensitivity to increments and decrements of flashes shown in our data can be ascribed to a reduction of L- and M-cone sensitivity rather than ocular media changes, as we compensated for the age-related increases in lens optical density using the HFP measurement. Werner and Steele [39] showed that all three cone types show a parallel reduction in their sensitivity with age. Results from color discrimination experiments [40,41] also suggest a reduction in cone signals with age. The Weber fraction of color detection thresholds increased with age, meaning a reduction in color discrimination. This reduction can be caused either because the L- and M-cone signals are reduced with age and/or because the inherent noise of the visual system increases with age, although additional experiments suggested that the noise is stable with age [42].

Shinomori and Werner [27] measured the impulse response by luminance increment flashes for 70 observers at $10\mathrm{cd}/{\mathrm{m}}^2$. The IRFs of that study [27] suggested a first phase peak time of 22.4 ms for younger observers (average age of 21.5 years) and 22.3 ms for older observers (73.5 years). The peak time of the second phase was 62.9 ms and 75.2 ms for younger and older observers, respectively. The difference of the peak times between the 2003 study and this study is about 8%, which is small considering the difference in the background intensity ($15\mathrm{cd}/{\mathrm{m}}^2$ in this study). The IRF amplitude depends on the luminance contrast at threshold; therefore, the relatively small difference of absolute background luminance may be ignored, and the amplitude of IRFs can be directly compared between these studies. Shinomori and Werner [27] calculated the peak amplitude of the first and second phases of the IRF for luminance increments of younger and older observers. The amplitude differences between the two studies were 13% and 3% for the first phase and 4% and 21% for the second phase for younger and older observers, respectively. These differences show that the difference of absolute background and stimulus luminance can affect the luminance contrast thresholds.

Overall, we conclude that in the magnocellular pathway, the processing time of luminance stimulation is stable with age, and the strength of the cone signals is reduced with age, even if we compensate for the age-related increases in ocular media density. These findings suggest that the temporal summation of the luminance system does not change with age, resulting in a reduction in sensitivity with age. This implies that the visual system does not compensate for the reduction in cone signals by increasing the time of temporal summation, resulting in a sensitivity loss. It could be speculated that this helps to maintain processing speed in the magnocellular pathway and thereby contributes to stable motion perception with age. Further investigation should be performed to determine the relation to motion.

B. Age-Related Changes in Luminance ON- and OFF-IRFs

The findings of this study show that the luminance ON- and OFF-IRFs are almost identical. Both the temporal profile and amplitude of the first and second phase are the same between ON- and OFF-IRFs for each age group. On the contrary, Shi and Shinomori [19] suggested that positive luminance contrast thresholds for incremental flashes are higher than negative luminance contrast thresholds for decremental flashes, causing the amplitude of the OFF-IRF to be higher than that of the ON-IRF. However, there was an age-dependent increase of luminance contrast thresholds for both incremental and decremental flashes when the data were considered for each subject. In five relatively young observers (mean 24.3 years), three observers showed higher positive contrast thresholds than negative contrast thresholds, but two observers showed almost the same thresholds between positive and negative contrast thresholds [19]. As described in the results, in twenty-two observers, nine observers (five younger and four older) had significantly higher positive contrast thresholds for incremental flashes ($p<0.05$), and three observers (one younger and two older) had significantly lower positive contrast thresholds ($p<0.05$). The other ten observers (six younger and four older) showed no significant difference between positive and negative contrast thresholds. Thus, while there is individual variation in the sensitivity of luminance ON- and OFF-pathways, overall, there is no clear evidence that negative contrast sensitivity is higher than positive contrast sensitivity as might be expected from physiological studies [16,23].