| How does the brain make perceptual decisions? Speed and accuracy of saccadic decisions regarding motion direction depend on the inherent ambiguity in the motion stimulus, and correlate with the temporal dynamics of firing rates in parietal and frontal cortical neurons of macaque monkeys. Some scientists claim that perception and decision-making can be described using Bayesian inference, which estimates the optimal interpretation of the stimulus given priors and likelihoods. However, such general statistical concepts do not propose brain mechanisms that enable perception and make decisions. Other neural models simulate some aspects of such data, but do not clarify important computations that need to occur between the motion stimulus and the saccadic response. The thesis develops the MOtion DEcision (MODE) model, which models interactions within and between Retina/LGN and cortical areas V1, MT, MST and LIP, gated by the basal ganglia, to provide a functional and mechanistic understanding of motion-based decision-making behavior in response to the experimental motion stimuli. Quantitative model simulations demonstrate how motion capture circuits in areas MT and MST gradually solve the informational aperture problem, while interacting with a noisy recurrent competitive field in LIP whose self-normalizing choice properties make probabilistic directional decisions in real time, without an appeal to Bayesian inference.;In order to further elucidate how directional grouping occurs in the brain and test some predictions of the MODE model, the thesis carries out two psychophysical studies that examine the ability of human subjects to estimate the direction of random dot motion stimuli under various parametric conditions. Main findings are that estimation accuracies are tuned to the spatial displacement between consecutive signal dot flashes, and not to speed, that this tuning function broadens with a peak shift towards larger spatial displacements as aperture size is increased, and that lowering contrast induces a rightward peak shift such that less contrast counterintuitively improves direction estimation of stimuli involving smaller spatial displacements.;These results are interpreted as behavioral correlates of pertinent neurophysiological data from area MT.;The thesis finally discusses potential MODE model extensions in the light of these new psychophysical data. |
