A model is presented for the early (retinal) stages of temporal processing of light inputs in the visual system. The model consists of a sequence of three adaptation processes, with two instantaneous nonlinearities in between. The three adaptation processes are, in order of processing of the light input: a divisive light adaptation, a subtractive light adaptation, and a contrast gain control. Divisive light adaptation is modeled by two gain controls. The first of these is a fast feedback loop with square-root behavior, the second a slow feedback loop with logarithm-like behavior. This can explain several aspects of the temporal behavior of photoreceptor outputs. Subtractive light adaptation is modeled by a high-pass filter equivalent to a fractional differentiation, and it can explain the attenuation of low frequencies observed in ganglion cell responses. Contrast gain control in the model is fast (Victor, 1987), and can explain the decreased detectability of test signals that are superimposed on dynamic backgrounds. We determine psychophysical detection thresholds for brief test pulses that are presented on flickering backgrounds, for a wide range of temporal modulation frequencies of these backgrounds. The model can explain the psychophysical data for the full range of modulation frequencies tested, as well as detection thresholds obtained for test pulses on backgrounds with increment and decrement steps in intensity.

As is well known, dark adaptation in the human visual system is much slower than is recovery from darkness. We show that at high photopic luminances the situation is exactly opposite. First, we study detection thresholds for a small light flash, at various delays from decrement and increment steps in background luminance. Light adaptation is nearly complete within 100 ms after luminance decrements but takes much longer after luminance increments. Second, we compare sensitivity after equally visible pulses or steps in the adaptation luminance and find that detectability is initially the same but recovers much faster for pulses than for increment steps. This suggests that, whereas any residual threshold elevation after a step shows the incomplete luminance adaptation, the initial threshold elevation is caused by the temporal contrast of the background steps and pulses. This hypothesis is further substantiated in a third experiment, whereby we show that manipulating the contrast of a transition between luminances affects only the initial part of the threshold curve, and not later stages.

We measured human psychophysical detection thresholds for test pulses which are superimposed on spatially homogeneous backgrounds that have abrupt onsets and offsets of high-contrast 25 Hz flicker. After the onset of the background flicker, test thresholds reach their steady-state levels within 20-60 ms. After the offset of the background flicker, test thresholds remain elevated above their steady-state level for much longer durations. Adaptation after onsets and offsets of background flicker is modeled with a divisive gain control that is activated by temporal contrast. We show that a feedback structure for the gain control can explain the asymmetric dynamics observed after onsets and offsets of the background contrast. Finally, we measure detection thresholds for tests presented on steadily flickering backgrounds as a function of the contrast of the background flicker. We show that the divisive feedback model for contrast gain control can describe these results as well.

As is well known, dark-adaptation in the human visual system is much slower than recovery from darkness. We show that at high photopic luminances the situation is exactly opposite. In psychophysical experiments on human subjects, we have studied detection thresholds for brief light flashes, at various delays with respect to decrement and increment steps in background luminance. Light adaptation was nearly complete within 100 ms after luminance decrements, but took much longer after luminance increments. In an effort to determine the nature of the threshold dynamics, we have compared sensitivity after equally visible pulses or steps in the adaptation luminance. Flash detectability was initially the same in the pulse and step conditions, but recovered much faster after pulses than after increment steps. This suggests that the initial threshold elevation is caused by the temporal contrast of the background steps and pulses, whereas the residual threshold elevation after an increment step shows an incomplete luminance adaptation. We have substantiated this by manipulating the contrast of a transition between luminances: we found that these contrast manipulations affected only the initial part of the threshold curve, not later stages. Finally, we measured detection thresholds for brief luminance decrements. For these tests with negative polarity, threshold recovery remained significantly faster after decrement than after increment steps in background luminance. Therefore, the asymmetry in adaptation dynamics that we report is indeed related to the step direction of the background luminance, and is not caused by interaction with the test polarity.