Auditory perceptual representations (i.e., “sounds”) reflect the brain’s ability to group or segregate acoustic features based on detected regularities in the acoustic environment. These perceptual representations provide information on the putative sound sources in the environment, which are combined with knowledge of auditory categories to both imbue sounds with meaning and inform appropriate behavioral actions. Critical aspects regarding the cortical mechanisms responsible for regularity representation and perception as well as how prior knowledge of auditory categories influences perceptual judgments remain unanswered. This dissertation had two main goals: (1) to test how neural activity encodes regularity representation and perception; and (2) to test how a listener uses prior category knowledge to inform categorical judgments when a stimulus’ category membership is ambiguous. To achieve these goals, I employed a combination of neurophysiological, behavioral, and computational analyses in humans. I found that the phase of population-level neural activity is a more reliable indicator of regularity than power and that a variety of brain regions exhibited reliable modulations that distinguished stimulus and behavioral differences related to regularity violation. Additionally, I found that human listeners learn approximations of auditory categories and are varied in their ability to use prior category information to inform categorical judgments. Finally, I found that categorization behavior was consistent with an ideal decision strategy that includes trial-by-trial variability in a listener’s estimates of the prior probability of each category. These findings build upon previous work on the mechanisms underlying regularity processing in auditory perception and that future research should focus on a variety of brain regions beyond the classical auditory pathways in cortex. Additionally, the categorization findings are the first to extend previous work in visual categorization into the auditory domain and reformulates the issue of categorization in a manner that can help to interpret the results of previous research within a generative framework.