Selbstregulation in biologischen Systemen illustriert an Modellen zur visuellen Wahrnehmung, zur ökologischen Populationsdynamik und aus der Neurophysiologie

Motion that is occurring suddenly is known to be extremely eyecatching. This observation leads to the question of how such salient beginning movements are processed by neurons in the early visual system. Self-regulative mechanisms, which reduce the cells´ firing rates if they have been too high during the recent past, are shown to be capable of both speeding up the rise time of retinal activity following a motion start as well as causing postretinal neurons to signal motion onset saliency by their firing rate. The speed of retinal processing of dynamic stimuli is further investigated by Bayesian estimation of stimulus properties from the experimentally determined response to abrupt changes in velocity. The second part of this work deals with examples of self-regulated systems in which the regulation sets in after a certain time lag. Here, the assumption of a singular delay is substituted by proposing that time lags in biological systems are diversely distributed within a certain range. Therefore, when computing the feedback strength in a model of such a system, its states during a whole past time interval have to be considered. Broadening the interval during which past states contribute to the regulation yields more regular behaviour in the predicted time course compared to the case of a singular time lag. Three different models of various complexity are examined. In a model of ecological population dynamics, oscillations can be reduced in amplitude if regulation depends on population densities integrated over a past time interval. The same is true for the Mackey-Glass-system that describes the concentration of circulating blood cells. Furthermore, in a model of neuronal recurrent inhibition in the hippocampal complex of mossy fibres, CA3 pyramidal cells and basket cells, stronger, more regular membrane potential fluctuations in pyramidal cells resembling burst-like activation occur in case of a broad distribution in the feedback time lags.