When perceptual time stands still: long percept-memory in binocular rivalry

We have carried out binocular rivalry experiments with a large number of subjects to obtain high quality statistics on probability distribution of dominance duration (PDDD) for two cases where (a) the rival stimulus is continuously presented and (b) the rival stimulus is periodically removed, with stimulus-on and stimulus-off intervals T(on) and T(off) respectively. In the present study we have chosen to study the regime of relatively long stimulus-on time, i.e., T(on)> 1s, where the stimulus presentation duration is significantly longer than the human reaction and recognition time. In the case of periodically removed stimulus, the total probability for percept reversal during each of the successive stimulus-on intervals T(on) can be predicted using the PDDD for continuous viewing. More importantly, this total probability for percept reversal during any stimulus-on interval is independent of the length T(off) of the preceding blank time, which can be quite long. We argue that this suggests that, in the regime of long T(on) and T(off) considered here, the variables representing the perceptual state do not change significantly during long blank intervals. We discuss that these findings impose challenges to theoretical models which aim at describing visual perception.