Simulation of Spontaneous Perception Switching with a Nonlinear Dynamics Model Using Perception-Attention-Memory Coupling

Abstract

In the present work new simulation results of an extended version of a nonlinear dynamics model of bistable perception due to ambiguous visual stimuli [Fürstenau 2006] are presented. Specifically long range correlations of the perceptual duration times are predicted which have recently been found experimentally by Gao et.al.(2006) via determination of the self similarity (Hurst) parameter H (> 0.5) of the reversal time series. The basic model couples the dynamics of a macroscopic (behavioral) perception state order parameter with an adaptive attention (feedback gain) control parameter with additive band limited white noise (Fürstenau2004), 2006). Memory effects are introduced by allowing for the adaptation of the originally constant perception bias parameter which balances the subjective preference of one of the two percepts. Determination of the Hurst parameter of perception reversal time series with the log(variance) vs. log(sample size) method used by Gao et.al. (2006) as well as with the more sophisticated method of Kettani and Gubner (2002) yield significant long range correlations in the simulations when coupling the bias parameter to the perception state, i.e. allowing for learning and memory.