An Exploration of the Applications of Neurally-Informed Models to Perceptual Decision-Making Research

Abstract

Understanding how the brain translates sensory input into action is of key importance to our understanding of all cognitive processes. Computational models utilise behavioural data to offer insight into the latent variables which underpin perceptual decision-making. However, competing models can often provide equally strong fits to the data, achieved through fundamentally distinct parameters. Examination of neural signals that index distinct stages of the decision-making chain can help us to overcome this. Through neurally-informed modelling, we can arbitrate between competing models based on their capacity to reproduce the neural markers of decision-making. Furthermore, we can constrain our models based on this neural data, allowing us to investigate increasingly more nuanced models. This thesis is designed to explore the possible applications of neurally-informed modelling to create data-driven, biologically-grounded representations of perceptual-decision making. Through this, this body of work furnishes novel insights into well-established features of perceptual-decision making including how ageing affects speed-accuracy tradeoffs and the decision process adjustments governing perceptual learning while providing novel model constraints and a method for improving the neural indices we rely on.