Computer Assisted Analysis of the Music of Elton John. Rupert Till & Phillip Allcock (University of Huddersfield)
[abstract] This paper explores what part computational methods can play in the analysis of popular music, and how they can be combined with other approaches to form a better understanding of the analytical subject. This project investigates the use of Humdrum, a powerful computer toolkit that in the past has mostly been used to analyse classical music. Although it uses musical scores as its source, It offers a high level of flexibility, and can provide valuable objective data about musical content.
The music of Elton John is used as a case study, and the use of traditional musicological analytical tools is compared with computational methods, as is what other approaches might be required to deal with cultural issues such as gender, identity, and sexuality. The value of Humdrum in exploring the changes in musical style between the four distinct eras of Elton John s music is discussed.
The paper goes on to briefly contrast other approaches to popular music analysis involving computer based techniques, including Sonic Visualizer; computer tools developed for the analysis of experimental electro-acoustic and acousmatic music; and commercial software designed for popular music production, such as Logic and Melodyne. It compares the potential of various tools, pointing out the many areas in which computer software cannot provide a complete solution, and addresses the restrictions of analysis based on musical scores. It concludes by suggesting in which situations and combinations various computer tools might prove most useful, and exploring where further… [abstract ends here].
Rupert starts with core questions
- How can music analysis inform socio-cultural issues such as gender?
- How can we use music analysis to understand the role of gender in the performances of Elton John?
- Is there a role for scores?
- Is computer analysis tool Humdrum of use?
- What are the pros and cons of such computer tools?
He cites methodological challenges, and specifies the way that the EJ canon has had to be thinned to create a manageable dataset. Phillip now presents a section, discussing the Humdrum software itself and acknowledging its capabilities and limitations. He notes that it requires a score as input.
EJ’s career is divided into sections for historically-contextual song analysis. 1969-1976 are the ‘Eligible Bachelor’ years. The rather glamorous album cover is suggested, perhaps playfully, to be ‘less than macho’, but Phillip accepts that in historical context pink baseball jackets (for example) were commonplace in the glam era. Era two (1977-83) is referred to as ‘Problems’ and we hear Blue Eyes, the problems in question relating to sexuality, personal relationships etc. The lack of sequins and glitter is noted in this period. Musically, the harmony has become more sophisticated and less rock ‘n’ roll, and Phillip considers that some cadences seem ‘forced’ [JB comment –an example of this adjective is not given, and I have to wonder how Humdrum was used to conclude this?].
The third era is called ‘Artifice’ and refers to 1984-93, referring to biographical artificiality and superficiality. It was in this period that he divorced his wife and came out. The music of this period has a small melodic range and almost exclusive diatony in harmony and melody. The final era, 1994-present, is referred to as ‘Rejuvenation’. Lion King and Billy Elliott are cited as extra-pop projects. Harmonic complexity increases but not to the level of the ‘Problems’ years.
Phillip notes the limitations of Humdrum, and identifies where it incorrectly identifies the key. He notes that Humdrum was designed to work with (notation of) Western classical music [JB note – I wonder here whether its poor effectiveness at identifying the much-discussed mixolydian nature of rock/pop harmony?].
Most interesting, for me, were the melodic pitch ranges, and it is here I think that Humdrum seems to be making its most effective contribution to the project [JB note – there is clearly a strong case for getting a human to identify the key signature, time allowing!]. Chord frequency is shown in pie chart form, which is a clear way of demonstrating graphically the drift into (or away from) diatony in the different composing periods.
Rupert now points out that the software, despite its limitations, does at least provide scale, helping the project to analyse 43 songs. [is scale a strong methodological argument for not using human analysis with a corpus of only 43 songs?]. He notes its objectivity and pitch range analysis [for me, the best feature for this project] and can provide a dataset for analysis. He briefly mentions Sonic Visualiser, KIAALs, Logic etc. We see a SV graph of Your Song and Rupert bravely plays us a musically wrong auto-transcription of EJ’s vocal. Melodyne is also cited, and although its recently-acquired [amazing!] mix-extraction polyphony is not mentioned, it is implicit in the audio that Rupert plays.
The data correlations imply that there is an observable correlation between the biographical data and the musicological content, particularly relating to complexity. The link to gender is less clear. The paper concludes that scores still are useful, but that they are only a partial solution.
A questioner notes that Humdrum has now been superseded by a new application called Music21.