Optimal Control of the Compensatory Eye Movement System

In this thesis, we utilized behavioural, electrophysiological, computational and stimulation techniques to delve our knowledge further into the functional neural network of the compensatory eye movement system (CEM). We first investigated the superposition violations and non-linearities in the CEM system with less predictable SoS stimuli (Chapter 2). We found that the superposition principle is violated in all CEM movement types tested (VOR, OKR, VVOR, SVOR) and that power spectra of SoS stimuli did not yield any distortion products. Natural stimulation of the CEM system generally involves complex frequency spectra. Thus, use of SoS stimuli is a step towards unravelling the signals that really drive CEM and the predictive algorithms they depend on. We then went on to do a feasibility study in the OKR paradigm whereby we identified the electrophysiological properties of modulatory purkinje cells in the cerebellar flocculus in response to SoS vertical axis stimuli (Chapter 3). Qualitatively we did see violations to the principle of superposition, as relative gains were not at unity and the modulation of complex spikes in the floccular purkinje cells was altered. In future studies, this methodology could confirm that complex spikes encode at least partially a motor component, rather than pure retinal slip, also during stimulation that is more complex than traditional sine waves. Turning to computational methods, we then tested our state predicting feedback control (SPFC) model against our physiological data to determine its robustness as an optimal control model for CEMs (Chapter 4a and 4b). Our working version of the Frens and Donchin state predicting feedback control (SPFC) scheme of the CEM system (Frens & Donchin, 2009a) was challenged in a broad range of experimental conditions. We found that the model successfully predicts the behaviour of the CEM system in all tested conditions. These results strongly support our results from Chapter 2 and Chapter 3. Finally, we moved on to stimulation techniques in the VOR paradigm and isolated the ability of the CEM to undergo cerebellar motor learning via the VOR in response to direct current stimulation (Chapter 5). Here we developed a mouse model of cerebellar DCS. We tested whether effects of stimulation depend on long-term potentiation (LTP) in the parallel-fiber (PF) Purkinje cell (PC) synapse. Our findings were that in a cerebellar motor learning task, initial learning rate was enhanced by anodal stimulation in wild types, but not in genetically ablated PF-PC LTP (L7-PP2B) mice. Additionally, neuronal activity should be assessed followed by cerebellar stimulation to understand the spatiotemporal aspects of DCS effects. Overall, the experiments presented in this thesis shed light on the optimal control of compensatory eye movements via cerebellar mechanisms and also provide a stable basis for future computational modelling and stimulation of the CEM system.