The Beatles: "In My Life" authorship analysis paper finally released

Close to a year ago, I wrote an article about statisticians who tried to analyse the songs attributed to Lennon and McCartney to see who really wrote them. You see, between the Beatles getting signed to Parlophone and John Lennon leaving the band in 1969, John Lennon and Paul McCartney had an informal agreement: if any of them wrote a song, it would be credited to both of them. Sure, they usually got feedback from each other, but a lot of the time, they wrote solo. Fortunately, when the band broke up, they’d clarify individual songwriting credits. Roughly speaking, these rules of thumb as whoever sung the lead on a given song was likely the main songwriter (“Every Little Thing” and “Day Tripper” appear to be the main exceptions). But there were some songs that they couldn’t agree on. One of them was “In My Life.”

Authorship accounts summarised:

John Lennon, The Playboy Interviews: “Now Paul helped with the middle-eight melody. The whole lyrics were already written before Paul had even heard it. In ‘In My Life,’ his contribution melodically was the harmony and the middle eight itself.” (Note: the song doesn’t actually have a proper middle-eight, just the second half of the verses with a different melody)
Paul McCartney, 1976 interview with Paul Gambaccini, “I liked ‘In My Life.’ Those were words that John wrote and I wrote the tune to it. That was a great one.”
Paul McCartney, 1996 interview for Barry Miles’ Many Years From Now: “As I recall, he didn’t have a tune to it, and my recollection, I think, is at variance with John’s. I said, ‘Well, you haven’t got a tune, let me just go and work on it.’ And I went down to the half-landing, where John had a Mellotron, and I sat there and put together a tune based in my mind on Smokey Robinson and the Miracles… I recall writing the whole melody. And it actually does sound very like me, if you analyse it.”

So, analyse it they did. Harvard Professors Mark E. Glickman and Ryan B. Song collaborated with Dalhousie’s Professor of Mathematics Jason I. Brown to settle the matter once and for all. The band deconstructed every song the duo composed between “Love Me Do” and “She Said She Said” and took small parts of each song and divided them up into blocks to identify 121 different musical features that divided into five categories:

Individual Chords
Individual Notes
Chord Progressions
Note Progressions
Musical Contours.

Eventually, they synthesised it all into a model to predict whether a song was composed by Paul McCartney or not. Now, normally, I would have loved to read the whole paper before I wrote my first article. There was one problem: it wasn’t available. It was all over my NME newsfeed, but nobody had actually read the paper. Around the time I wrote my article, Dad and I emailed Glickman for the paper. Only on the first day of June did he get around to responding.

The article, entitled “(A) Data in the Life: Authorship Attribution of Lennon-McCartney Songs”, has finally been approved for publication in the Harvard Data Science Review (HDSR). To quote the news on Glickman’s website:

The online version will be available mid-to-late June, and the print version (which will be a bundling of the first two online HDSR issues) will be ready in October. The journal graciously funded the development of several interactive demos that will be embedded within the online version on the HDSR site. The online version will also contain sound samples to illustrate the relevant musical concepts on which our analyses rely. We encourage you to take a look at the HDSR version when it is published. The preprint reflects an improved model over that which was reported earlier in the media, and all details are included in the article. Thank you for your patience in waiting to read our manuscript.

The pre-printed version can be found here. Things may change in the transition to the HDSR version. It may be a bit dry and not terribly accessible to people without the sort of background in music and statistics, but thankfully, I have some knowledge in those areas.

So, what changed in the intervening 10 months? Quite a bit, actually. The initial press release claimed that there were 149 musical features that went into the algorithm, and it looks like they streamlined it and dropped 28 from it. In addition, the initially reported .018 probability has increased significantly. The paper concludes: “Our model produces a probability of 18.9% that McCartney wrote the verse, and a 43.5% probability that McCartney wrote the bridge,with a large amount of uncertainty about the latter.” Surprisingly little of the 43-page paper covers “In My Life.” I suspected a lot of ink to be spilled on the minutiae of the chords, the melody, and their contours, but the most salient other point about it comes when they say “McCartney … stated he composed the song in the style of Smokey Robinson and the Miracles (Turner, 1999), but actually wrote in the style of Lennon, whether consciously or unconsciously.” At any rate, it sounds more like Lennon than Smokey:

It should be noted that Macca still has yet to respond to this study.

Of course, it turns out that there are limitations to this model. In the appendices, the authors apply it to 70 songs from the era known to be written by one or the other. Sometimes, the model gives a low McCartney probability score to a song known to be written by Paul, or a high McCartney probability score to a song known to be written by John. In fact, some songs from Revolver get it very wrong and even attribute it to whichever one of the duo DIDN’T EVEN PARTICIPATE IN THE RECORDING OF THE SONG.

Case in point: “For No One”, McCartney probability 18.4% [John and George not present for recording]

“She Said She Said.” [Inspired by something Peter Fonda told John while tripping on acid, Macca eventually refused to take part in the recording of this song for reasons related either to his refusal to take LSD or John not liking his proposed changes to the arrangement.]