A computational model of retinal circuitry predicts stimulus duration and intensity effects on visual persistence and afterimages

摘要

Visual persistence refers to a temporal characteristic of the visual processing in which a visual stimulus remains visible for up to a few hundred milliseconds after the stimulus physically disappears. Interestingly, a stimulus with higher contrast and/or longer duration persists shorter. While persistence is generally observed with 25~500ms of stimulation, negative afterimages (perceived images at the absence of physical stimulation that appear as weak polarity-reversed versions of previously presented stimuli) involve relatively long-term process and are typically observable only after more than few seconds of stimulation. Importantly, the effects of stimulus properties on afterimages are opposite to those for persistence such that a higher contrast and/or longer duration stimulus generates a stronger afterimage. While these two phenomena are important for investigating temporal vision and the seemingly related effects of the stimulus properties on them are intriguing, the underlying mechanisms producing these phenomena are unclear. In this study, we propose a theoretic /computational model that sheds light on the mechanisms of visual persistence and afterimages by simulating biologically plausible retinal circuitry for achromatic processing. According to the model, both persistence and afterimages are outcomes of a retinal light-gating process, which is largely determined by response kinetics and functional connections of horizontal and amacrine feedback layer cells onto the photoreceptor, bipolar, and ganglion feed-forward layer cells. Model simulations suggest that transient inhibition from the horizontal and amacrine cells to feed-forward layer cells differently shape ON and OFF ganglion cell responses and modulates persistence. On the other hand, afterimages are produced by slow changes of horizontal cell response kinetics that affect photoreceptors and bipolar cell responses to the background illumination. Overall, our results imply that the retinal circuitry decodes visual inputs into complicated temporal and spatial patterns, which consequently alter perceptual experiences.