Radiohead has a well-earned reputation for writing music that’s depressing as hell. Reviewers have described the band as “pre-eminent doomsayers” with an “evergreen grimness.” This isn’t to say it’s bad music—far from it—it’s just not the kind of thing most people would play at a pool party. It’s more fitting for, say, the funeral of a suicidal cat.
It is, perhaps, the most depressing popular band in musical history. But of the roughly 160 songs in their catalog, which is the bleakest?
Analytics specialist Charlie Thompson looked to the data to figure it out. “As a Radiohead fan, my friends always get on me for listening to music that they consider to be depressing,” he explains, “and I was fascinated by the idea that you could quantify how sad a song made you feel.” In a blog post titled “ fitteR happieR ” that subsequently went viral, he outlined the process in detail. It’s fascinating, and it hints at a whole mess of analytical possibilities using similar techniques. What’s the dirge-iest song by Nick Drake? The happiest song by Metallica?
Thompson, who works at the startup VideoBlocks, used Spotify’s audio stats and the lyrics on Genius. Both companies offer an open application programming interface (API), which essentially allows anyone to utilize their data for projects like this. Spotify takes certain qualities of songs and quantifies them numerically in its database. This includes standard stuff like tempo and key, as well as more subjective values like “instrumentalness,” “acousticness,” “danceability” and “valence.” The latter is described by Spotify as “a measure from 0.0 to 1.0 describing the musical positiveness conveyed by a track.” The higher the number, the happier the song. Apparently.
Scraping the track data for each recording in Radiohead’s oeuvre gave Thompson a list of each song’s valence. Some of the bleakest, according to the data, include “Everything in Its Right Place” (0.0585), “Motion Picture Soundtrack” (0.0425) and, way at the bottom, “We Suck Young Blood” (0.0378).
Songs are more than their instrumentation, however, so what about the lyrics? Thompson used the Genius API to collect the lyrics of every Radiohead song. Then he used a sentiment analysis tool to determine the percentage of words known to elicit “sadness” in each song. In this respect, “High and Dry” was the saddest, at 36 percent. “Specifically, the algorithm picked out the words broke, fall, hate, kill and leave, ” Thompson says in his blog post. “The last of which was repeated 15 times in the chorus (‘Don’t leave me high, Don’t leave me dry.’).”
Thompson then calculated the “lyrical density” of each song, which basically weighs how crucial lyrics are to a given song, then combined these separate piles of data with what he calls the “gloom index.” At the risk of detaching the corneas of anyone who stopped caring about math after high school, the formula is (1 – valence ) + pctSad*(1 + lyricalDensity) / 2.
So the most depressing Radiohead song is, according to Thompson, “True Love Waits,” with a gloom index of 1, from its most depressing album, A Moon Shaped Pool , which has a total gloom index of 31.93. Listen and weep. For comparison, the second most depressing track the group has recorded is “Give up the Ghost,” with 6.46. Not even close.
Or was it? To the math-averse, it’s worth remembering that numbers can be deceptive and that Thompson’s results aren’t definitive. Though informative, his data analysis still arrives at a subjective result. The initial values used to quantify each song (valence, sentiment analysis, etc.) are themselves just approximations. Art is difficult to quantify; that’s part of its beauty.
Besides, our perceptions of art are deeply personal, constantly evolving depending on the context and our mood when we encounter it. Put differently, in the voice of some stoned hippie wandering through the woods: The most depressing Radiohead song is whichever one you find most depressing, man.
Bleak music isn’t binary. There’s often pleasure in listening to depressing music, whether it is a source of comfort after a bad breakup or, like a good Leonard Cohen lyric, an excuse to laugh at the darkness. Sometimes it crosses over into hilarity; consider campy early-2000s emo songs, or overwrought nu metal songs, or goofy teenage tragedy songs from the ’50s and ’60s. And let’s also not forget songs that are so bad they become depressing. There’s nothing quite like hearing a Maroon 5 song play in a CVS to make one lose the will to live.
Thompson acknowledges this. “The algorithm doesn't necessarily reflect what all people consider to be sad, but at the very least it's another opinion,” he says. “And I do think that algorithms like this can help us understand ourselves, because when we create them we have to pinpoint what exactly it is that makes us feel a certain way.”
For that matter, a relentlessly upbeat song like “Walking on Sunshine” could instantly become more depressing than any Radiohead song if, totally hypothetically, it plays after you find that your depressed cat finally got the guts to go through with it.