Thesis defense James Cooke (Donders series 404)
18 October 2019
Promotor : prof. dr. P. Medendorp
Co-promotors : dr. L. Selen , dr. R. van Beers
Statistical models of object tracking
Our ability to accurately track our position in space as well as the changing locations of objects or other agents around us is key in our daily life, such as in traffic, or on a sports field. The goal of this thesis is to understand the brain’s computations that underlie this ability. A complicating factor for these computations is that the sensory inputs to the brain are noisy and often ambiguous, which means that probabilistic approaches are needed to both develop and compare different theoretical models. Regarding the latter, the thesis presents a novel algorithm that facilitates the design of stimulus presentation schedules that distinguish efficiently (i.e. using a low number of trials) between different models of cognitive processing. The thesis also provides a probabilistic model, underpinned with psychophysical measurement, in which noisy sensory information is optimally integrated with predicted sensory information, to keep track of multiple objects randomly moving in the world. Finally, the thesis deals with locating objects based on ambiguous visual motion cues These ambiguous cues can be caused by the moving observer, by the object itself, or by both. A probabilistic formulation of this causal inference process could account for the behavior that humans show in a spatial constancy task under self-motion.