It would be informative to have an electrophysiological method to study, in an objective way, the effects of mercury exposure and other neurotoxics on human color vision performance. The purpose of the present work was to study human color discrimination by measuring chromatic difference thresholds with visual evoked potential (VEP). Six young normal trichromats (24 ± 1 years old) and one deutan (26 years old) were tested. The stimuli consisted of sinusoidal isoluminant chromatic gratings made from chromaticity pairs located along four different color directions centered on two reference points. Heterochromatic flicker photometry (HFP) protocol was used to obtain the isoluminance condition for every subject and for all chromaticity pairs. Spatial frequency was 2 cycles/deg. Presentation mode comprised onset (300 ms)/offset (700 ms) periods. As previously described, we found a negative deflection in the VEP which was related to the chromatic difference: as chromatic difference increased, amplitude increased and latency decreased. VEP response amplitude was plotted against distance in the CIE 1976 color space between the grating chromaticities and fitted with a regression line. We found color thresholds by extrapolating the fitting to null amplitude values. The thresholds were plotted in the CIE 1976 color space as MacAdam ellipses. In normal trichromats the ellipses had small size, low ellipticity, and were vertically oriented. In the deutan subject, the ellipses had large size, high ellipticity, and were oriented towards the deutan copunctal locus. The VEP thresholds were similar to those obtained using grating stimuli and psychophysical procedures, however smaller than those obtained using pseudoisochromatic stimuli (Mollon-Reffin method). We concluded that transient VEP amplitude as a function of contrast can be reliably used in objective studies of chromatic discrimination performance in normal and altered human subjects.

A novel noise-masking technique was used to test D'Zmura and Knoblauch's (1998) idea that subjects employ off-channel looking in detecting chromatic test stimuli embedded in spatiotemporal chromatic noise. Detection thresholds were obtained for stationary, isoluminant, Gaussian-windowed (σx = σy = 2.25 deg; σt = 0.25 s), 135 deg (yellow/blue) or 160 deg (orange/blue–green), sinusoidal test gratings (11 deg × 11 deg; 0.75 cycle/deg) superimposed on each of a series of dynamic, random-check chromatic noise masks varying in azimuth in DKL space. Thresholds for detecting the test in the presence of these variable masks were again measured in the presence of an additional (auxiliary) noise mask created from colors falling on azimuths of 0 deg or 90 deg (135-deg test) or 0 deg or 135 deg (160-deg test). The effectiveness, kvar, of the variable noise masks in elevating grating detection thresholds was determined by fitting the detection data to the Pelli-Legge equation relating test detection energy to variable noise-mask energy: Et = K + kvarNvar. Differences in the calculated values of kvar for detection data obtained with and without the auxiliary masks were consistent with off-channel looking and were well accounted for by a simple model based on the idea that subjects possess a multichannel array of linear chromatic detectors spanning the isoluminant plane of DKL space, and they can choose the channel that has the highest signal-to-noise ratio.

It is widely accepted that human color vision is based on two types of cone-opponent mechanism, one differencing L and M cone types (loosely termed “red–green”), and the other differencing S with the L and M cones (loosely termed “blue–yellow”). The traditional view of the early processing of human color vision suggests that each of these cone-opponent mechanisms respond in a bipolar fashion to signal two opponent colors (red vs. green, blue vs. yellow). An alternative possibility is that each cone-opponent response, as well as the luminance response, is rectified, so producing separable signals for each pole (red, green, blue, yellow, light, and dark). In this study, we use psychophysical noise masking to determine whether the rectified model applies to detection by the postreceptoral mechanisms. We measured the contrast-detection thresholds of six test stimuli (red, green, blue, yellow, light, and dark), corresponding to the two poles of each of the three postreceptoral mechanisms. For each test, we determined whether noise presented to the cross pole had the same masking effect as noise presented to the same pole (e.g. comparing masking of luminance increments by luminance decrement noise (cross pole) and luminance increment noise (same pole)). To avoid stimulus cancellation, the test and mask were presented asynchronously in a “sandwich” arrangement (mask-test-mask). For the six test stimuli, we observed that noise masks presented to the cross pole did not raise the detection thresholds of the test, whereas noise presented to the same pole produced a substantial masking. This result suggests that each color signal (red, green, blue, and yellow) and luminance signal (light and dark) is subserved by a separable mechanism. We suggest that the cone-opponent and luminance mechanisms have similar physiological bases, since a functional separation of the processing of cone increments and cone decrements could underlie both the separation of the luminance system into ON and OFF pathways as well as the splitting of the cone-opponent mechanisms into separable color poles.