This note is prompted in part by the unsettling report of Mollon and Polden (1978) on the time constants of tachistoscopes. They tested a number of tachistoscopes, including an Iconix fitted with the lamps used by Banks and Barber (1977), and found that they typically required about 18 msec to reach 900/0 of full output when they were turned on or to decline to 10% of output when turned off. Furthermore, the spectral composition of the light changed Over the rise and fall periods. The change in spectral composition is particularly important because the conclusion we drew (Banks & Barber, 1977), that iconic memory was not strictly a rod phenomenon, depended in part on use of stimuli that present zero contrast to the rods, despite being clearly distinguishable to the cones. In that research we created letter arrays by using one color of paper for figure and another for ground such that the cones, but not the rods, could see the contrast between the colors and thus only the cones could be used to detect the letters. The fact that these arrays gave typical iconic decay curves and a partial report superiority seemed to us to be good evidence that iconic memory did not require the rods. We realized, of course, that the matching of the colored papers depends on the spectral composition of the illuminant, and we made our matches using as an illuminant the steady-state light of the stimulus field of the Iconix, filtered through the two mirrors through which the stimulus was seen in Our experiments. If, however, the spectral composition of the light during the rise and fall periods of the tachistoscope lamps was not the same as it was during the steady-state period, then the stimulus arrays would not have zero contrast for the rods during these periods, and the rods might have been able to mediate iconic memory during experimental trials. Another motivation for this note comes from Sakitt and Long (1979), which shows that a small