Brief sensory inputs to intact conscious subjects commonly trigger complex long-lasting motor responses, in which higher cerebral mechanisms, or even voluntary action, may be integrative parts. However, long-lasting motor discharge following brief afferent stimulation is also observed in reduced preparations, such as decerebrate or spinal animals. Reverberating activity in closed neuronal loops has been postulated to be an important integral mechanism underlying various phenomena, ranging from temporal integration in the vestibulo-ocular reflex to maintenance of “initial” memory. Such far-reaching notions stand in sharp contrast to the virtual absence of experimental evidence for sustained reverberating activity in well-defined neuronal circuits. Present quantitative knowledge of the sensory input and efferent output in the vestibulo-ocular reflex has prompted several investigators to postulate a neuronal integrator in the intervening CNS. There have been several attempts to model this neuronal integrator. The assumption common for all these models is a network of neurones with positive feedback loops. The bistable properties of motoneurones allow them to act as very simple integrators, like flipflops, which are set at one of two levels by short excitatory or inhibitory inputs. However, when the whole motoneuronal pool is considered, many different levels can be maintained by recruitment of new units.