부산대학교 도서관

Andrew Hughes' Late Medieval Liturgical Offices afforded chant scholarship more melodies than it knew what to do with. Until now, chant scholarship involving 'Big Data' usually meant comparing individual feasts to the whole corpus or looking at general trends with respect to 'word painting' or stereotyped cadences. New research presented here, using n-gram analysis, networks, and Recurrent Neural Networks (RNN) looks to the nature of the gestural components of the melodies themselves. By isolating the notes preceding, and proceeding from, the naturally occurring semitones in the medieval church modes, we find significant recurrence of particular phrases, or riffs, which we propose could have been used to help 'build modes' from the inside out. Special care needed to be brought to the question of assumed B-flats that were not given explicitly in the manuscripts represented in Hughes' work. Understanding modes not as 'scales' but as a collection of associated smaller musical gestures, has resulted in a set of recurring riffs that appear as the identifiers of their larger contexts and confirming the influence of an earlier, oral / aural culture on these late medieval chants where musical literacy was expected.