The sensation of rhythm in music is evoked by accents that follow a more or less regular structure in time. This structure takes the shape of a rhythm hierarchy consisting of note onsets, the pulse of a piece, rhythmic patterns, and, finally, the units of the musical form. Many computational analysis methods concentrate on one level of the rhythm hierarchy in isolation, and ignore the importance of melody for rhythm. This project proposes a Bayesian framework for the analysis of rhythm, which takes into account the relations between the hierarchy levels, and the impact of melody on rhythm.Most current approaches are further limited in their universality, because they are tailored towards certain paradigms of Western music, such as the well-formedness of meter. By analyzing music which contradicts common paradigms, we aim at a set of analysis tools that gives us access to a wider variety of music. The developed methods will be evaluated on Western and on non-Western music, which will ensure the universality of the proposed methods.This project will contribute directly to the state-of-the-art in Music Information Retrieval (MIR), an area with growing economic impact. The proposed methods will permit implementing novel computer-based learning and music recommendation platforms for types of music that currently cannot be analyzed by the state-of-the-art. Geographical location and chosen musical context maximize the direct economic impact of such learning and retrieval platforms. The project will profit from a wide variety of expertise. The fellow is a specialist in rhythm analysis, and the host institution has high expertise in Bayesian rhythm modeling and machine learning, consultants of the project are experts in musicology and melodic analysis of non-Western music.