Brain oscillations are dynamic entities, rapidly varying in time and frequency that are extensively recorded in the mammalian brain. In this thesis I investigated the role of cortical rhythms in human auditory cortex during speech perception, both with computational and experimental methods. We developed a neural microcircuit model of nested oscillations for early auditory processing, involving a fast Gamma rhythm (30-100 Hz) coupled to a slow Theta rhythm (3-8 Hz). The model is capable of parsing speech into its constituents (i.e. syllables) and extracting the syllabic information for latter categorization. We then analysed EEG signals from intracranial recordings in humans using cross-frequency coupling and Granger causality. We found that Bottom-up information is dominated by Gamma oscillations, while Top-Down flow is conveyed through Theta oscillations. Both flows fluctuate over time at a rate of ~1-3 Hz, suggesting that sensory information is conveyed using distinct frequencies and via discrete time windows.