The motivation for this study was the remarkable ability of the brain to categorize musical styles, based solely on the information conveyed by the music. It was assumed that a computer system could accomplish similar results based only on the musical notation, and that an effective categorization of styles should reveal the main conceptual dimensions of tonal music style. The most outstanding characteristics of this study are its quantitative nature and the effort to avoid the subjectivity of structural analysis, considering only observable, measurable features. The work has been carried out exclusively on the basis of numerical continuous variables resulting from actual measurements effected by software on those aspects of musical style that are detectable in notation. The numerical data set resulting from the measurements suffices to allow data mining algorithms such as CART, Random Forests and GEP, to map music pieces to a multidimensional space and classify them by composer, with an accuracy related to the size of the database. This system has great potential in the areas of empirical musicology and authorship studies, and as the basis for a scientific theory of style.