Using Bayesian frameworks to explore simple cognition

In this thesis I explore simple cognition from two perspectives, both using Bayesian approaches. Firstly, the statistical benefits of Bayesian estimation in cognitive models of decision-making are demonstrated. In this research stream, I cover fundamentals of Bayesian estimation using Markov chain Monte Carlo sampling techniques. I demonstrate the benefits of such techniques with two applications of a well-validated evidence accumulation model, the Linear Ballistic Accumulator (LBA: Brown & Heathcote, 2008). The first application highlights the importance of the Bayesian estimation of the LBA in a clinical setting; I uncover latent mechanisms involved in the inability of individuals with major depressive disorder to disengage from negative emotional stimuli. The second application explores the foundational aspects of the LBA mechanisms from a neural perspective. To conclude this section, I describe a novel development in joint modelling. A model framework was developed which quantitatively links neural and behavioural data streams. The framework allows both data streams to jointly influence parameters, via Bayesian estimation. The second stream of research focuses on Bayesian frameworks for the description of cognition, rather than as tools to measure it. Motivated by current debates in the literature I directly and quantitatively compare human and Bayesian inference, demonstrating that human inference deviates from statistically optimal Bayesian inference, at least for simple inference. I finally describe work which explores the potential mental representations of alphabetic letters using a novel graphing algorithm based on Bayesian principles. The work in this thesis highlights the benefits of Bayesian statistical framework, both demonstrating and adding to their lasting contribution in cognitive psychology. Conversely, this thesis provides a direct critique of Bayesian cognitive frameworks, demonstrating some short-comings and adding to existing debates about the suitability of the analogy.