PURPOSE. To investigate two apparent anomalies of the human electroretinogram: the "on" and "off" components of the cone based PIII are unequally sized, and transitions from red to green, which are electroretinographically silent, yield reverse transitions (green to red) in which a-waves develop. METHODS. Ganzfeld electroretinograms were obtained with intense 100 msec flickering flashes from red and green light-emitting diode. Such stimuli light-adapt the retina, and the responses are caused by the excitation of long and medium wavelength cones. RESULTS. In the 10-20 msec after the beginning of a flash (black to green or black to red) the beginning of rapid receptor-generated a-wave is seen. Ten to twenty milliseconds after the end of the flash, the beginning of a rapid positive-going off response, also derived from receptors can be seen. If the retina is stimulated by the abrupt change from one wavelength of light to another (eg, from "green" to "red"), at times > 20 msec after the change there are always slow changes in potential (presumably caused by postsynaptic activity) regardless of the relative intensities of red and green. However, if the two light intensities are adjusted appropriately, 10-20 msec after the transition from green to red no electroretinographic a-wave (or off response) develops--the transition is "silent." When the transition reverses (changes back from red to green), an a-wave occurs. In the same way if a red-to-green transition is made silent by altering the relative light intensities, the green-to-red reversal evokes an a-wave. This occurs for numerous pairs of red and green intensities. Rod intrusion or minor electroretinogram components do not explain this result. The relative red:green intensity in two color-anomalous subjects is different to that in three normal subjects. The rule for a silent transition is that the decrease in excitation in one cone type should be twice the increase in excitation in the second cone type. CONCLUSIONS. The most likely cause is a reduction in the amplitude of cone receptor potentials 20-50 msec after the onset of the stimulus, caused by a sign-reversing feedback mechanism such as that described in amphibians. This implies that the chromatic signals for color vision required by theorists are partly generated in the cones.