The aim of this thesis is to develop a novel technique for the estimation of firing rate dynamics from single-unit recordings of neural pulse trains. This method applies an offline digital filtering technique to extract information transmitted by a neuron in teens of a rate code. While there is increasing evidence that the traditional rate coding cannot account for all the information transmitted by a cell, and that information may also be contained in the precise timing of spikes, the firing rate signal remains the benchmark by which the vast majority of electrophysiological studies relating neural activity to functional behaviour have been interpreted. Nevertheless, there does not seem to be an agreement on a single definition of a rate code let alone a consensus on an optimal estimation method. This study raises significant concerns about the validity of some of the most common methods in systems neuroscience, and proposes a simple yet more robust alternative. This latter is based on the convolution of the spike train with an optimally designed Kaiser window. Using computer-simulated as well as experimental data obtained from single-unit recordings of vestibular canal afferents, the proposed technique is shown to consistently outperform the current methods and even to permit robust estimations under time-varying conditions. These results suggest that estimates acquired with the conventional methods are biased and hence models of neural dynamics based on these latter may not be reliable.

It has been hypothesized that head-unrestrained gaze shifts are controlled by an error signal produced by a feedback loop. It has also been hypothesized that the superior colliculus (SC) is within this feedback loop. If the feedback-to-SC hypothesis is valid, an unexpected mid-flight perturbation in gaze trajectory should be quickly followed by a concurrent change in the discharges of collicular saccade-related neurons. To verify this prediction experimentally, primate head movements were unexpectedly and briefly halted during head-unrestrained gaze shifts in the dark. Perturbed gaze shifts were composed of first a gaze saccade made when the head was immobilized by the head-brake, followed by a period where gaze was immobile, called a gaze plateau. The latter was composed of an initial period when the eyes and head were immobile, followed by a period wherein the head was released and the eyes counter-rotated to stabilize gaze. The plateau ended with a corrective gaze saccade to the goal location. In perturbed gaze shifts, there was widely distributed activity on the SC map during gaze plateaus, and there was no evidence that the initial motor program was aborted; the corrective gaze saccades were not "fresh" small stand-alone movements. Cells on the SC map responded at short latencies to head accelerations and associated gaze shift perturbations and carried a gaze position error (GPE = final - instantaneous gaze position) signal. As a large gaze shift progressed there was a caudo-rostral moving hill of activity on the SC map that encoded, not instantaneous veridical GPE, but a filtered version of it (time constant 100ms). Recordings from both the motor map and the so-called "fixation zone" in the rostral SC during perturbed head-unrestrained gaze shifts reveal gaze feedback control and a gaze feedback signal to the SC. However, these results do not prove that the SC is within the online gaze feedback loop, only that such a loop exists and that the collicular map is informed about its calculations.

No description available.

No description available.