Corpus Approaches to Popular Music Analysis

A seminal paper by Temperley (2007) drew attention to a phenomenon in Western popular music dubbed the “melodic-harmonic divorce,” a term he attributed to Alan Moore (1995). While a well-orchestrated relationship between melody and harmony is the lynchpin of Western common-practice music, Western popular music, on the other hand, frequently eschews the traditional formal rules of harmony and counterpoint.  Temperley, for instance, argues that pop/rock melodies are less likely than classical melodies to have non-chord tones (NCTs) resolve down by step, are more likely to have multiple NCTs in a row, or contain NCTs left unresolved.  The phenomenon is also argued to be more prominent in verse melodies than chorus melodies: a phenomenon coined by Temperley (2007) as the “loose-verse, tight-chorus” (LVTC) model.

In this paper we examine the propensity of the melodic-harmonic divorce phenomenon via a large-scale corpus analysis of a set of 414 expert-encoded popular melodies—dubbed the “CoCoPops” project, with 212 melodies pulled from Burgoyne et al.’s McGill Billboard corpus of harmonic annotations (Burgoyne et al., 2011),  as well as 200 melodies encoded as a part of the RS200 project (Temperley & deClercq)—with the aim of shedding light on the basic questions of when (and where?) does melodic-harmonic divorce occur? And, how prevalent of a phenomenon is it in popular music in general? Our paper will discuss the contentious issues of defining a chord (Doll, 2013), chord labeling and harmonic analysis, especially of popular music, and how our practices impact the interpretability and reliability of results.  

Harmonic tension and closure in 21^st-century pop defy common-practice theoretical concepts; a melody often appears independently above the cyclic repetitions of the chord-loop, where multiple tonal centers are perpetually in non-alignment with each other. Instead, formal sections adapted from Electronic Popular Music (EDM)—riser, drop, and dance chorus—create musical tension and closure through its sonic function (Peres 2016, Barna 2020). The characteristic equivocal tonal centers in popular music have been analyzed as multiple modal tonics (Richards 2017), “fragile, emergent, and absent tonics” (Spicer 2017), “hybrid tonics” (Duinker 2019), or “divorced” of musical layers (Nobile 2015, De Clercq 2019). While there have been case studies of “ambiguous tonality” in notable chord-loops schemas (Richards 2017, Duinker 2019), there is no expert-encoded harmony dataset of 21^st-century popular music to corroborate these harmonic theories.

Through a meta-corpus study of an expert-encoded dataset over 300 songs, I uncover the shift in harmonic syntax of chord-loops after 2000. I compiled and encoded 21^st-century pop harmony using a “relative-key” approach (De Clercq 2019), which reveals a shared harmonic grammar consistent within a style: it allows multiple interpretations of chord-loop’s tonality, while ensuring unequivocal identification of the chord-loop schema. Moreover, the use of chord-loop pairs—two related chord-loop schemas that always appear as a pair—is normative to EDM-pop especially in respect to “Axis” (Richards 2017) and “Plateau” (Duinker 2019) loop family. I close my paper by observing the prominence of the EDM-influenced chord-loop syntax across various 21^st-century popular music genres.

While pitch material has been thoroughly studied in pop/rock (White and Quinn 2018, Doll 2017, Temperley 2018), other, less-studied parameters play a potentially larger role in these styles (Tagg 1982). For example, guitar timbres take on a signifying role for distinguishing between genres (Lavengood 2020, Howie 2020, Gjerdingen and Perrot 2008), and prosodic stress and rhyme can contribute to alternative rhythmic layers (Komaniecki 2021, Eron 2020, Condit-Schultz 2017). Following these insights, this paper studies microtiming in punk vocals to investigate genre boundaries. To do so, we constructed a corpus of pop-punk and post-punk songs—the All the Small Things Punk Corpus—consisting of both quantized transcribed musical segments, and inter-onset intervals. By analyzing the microtiming of vocal lines, we show that microtiming deviations vary significantly between subgenres of punk music. This suggests that microtiming adds specific stylistic markers of music, and that these deviations therefore may play some role in listeners’ ability to discriminate genres.