Exploring Vowel Categories Using the Human Markov Chain Monte Carlo Approach

Markov Chain Monte Carlo (MCMC) is a simple but powerful algorithm originally designed to sample from complex probability distributions. Recently, it has been adapted as a new psychophysical method to investigate how humans categorize complex stimuli. The innovative part of this approach involves placing the participant "in the loop" of the MCMC algorithm, enabling it to sample distributions corresponding to the participant's mental categories. This methodology has been successfully applied in the visual domain, shedding light on processes such as color perception, facial expression recognition, and animal categorization. The goal of this internship is to extend this method to address a long-standing question in the field of speech perception: how do humans recognize and categorize vowels? We will use the "human MCMC" framework, to map the contours of mental vowel categories.