Abstract
To successfully catch the prey or escape from hunting, organisms need to make prediction of the future trajectory of the object in order to compensate the time delay resulting from signal transmission and processing. The anticipative ability of retina has been demonstrated in simple motion such as a moving bar with constant velocity or with directional reversal. However, these studies lack a general applicability under noisy stimulations as in most cases in nature. Therefore, a stochastic moving bar is applied to the bullfrog’s retina and we use time-shifted mutual information to quantify the predictive information shared between bar positions and firing rates of a single ganglion cell. Results show that retina reveals predictive computation of the future bar position under stochastic moving bar. Furthermore, we found that the orientation of moving bar, correlation times and the statistics of motion will influence the retinal prediction ability of future position. Among these factors, the orientation-selective prediction of HMM forms a stark contrast to the predictive firing pattern under motion with constant velocity, which is general in different motion orientations. More concentrated studies of the effect of motion orientation on prediction is needed in order to understand the retinal circuit of prediction.