What Does a Corpus of Music Represent? Commentary on “Diversity in Music Corpus Studies”

In Volume 30 (1) of this journal, Shea et al. (2024) introduce a novel sampling method called the "Anti-Discriminatory Alignment System" (or ADAS), which is designed to adjust the demographic distribution of artists in a corpus of music.In this commentary, I consider what a corpus created using the ADAS represents.I begin with a general discussion of corpus work in the context of foundational statistical principles.I then posit two different types of corpora that could be created to model the real world: music as heard and music as produced.A corpus created with the ADAS, in contrast, appears to be neither of these two types; instead, it appears to be a corpus of music as imagined, which models only a hypothetical statistical population.As a result, statistical findings from a corpus created with the ADAS may have limited explanatory power.If a researcher hopes to diversify the demographic distribution in a corpus of music, I argue that these attempts should occur not at the sampling stage but rather at an earlier stage, when the researcher is determining the particular statistical population to study.