Burn down the disco, Hang the blessed DJ, because the music that they constantly play, It says nothing to me about my life

There’s been a certain amount of breathless reportage about a new study linking personality and musical taste done by Adrian North at the Heriot-Watt University in Edinburgh, UK. It’s hard to be sure where things went awry, but by the time the media got a hold of the results they were badly exaggerated.

The BBC, for instance, says the research “suggested classical music fans were shy, while heavy metal aficionados were gentle and at ease with themselves.” North himself, interviewed by the BBC, makes similarly bold claims: “If you know a person’s music preference you can tell what kind of person they are, who to sell to. There are obvious implications for the music industry who are worried about declining CD sales. One of the most surprising things is the similarities between fans of classical music and heavy metal. They’re both creative and at ease but not outgoing.” These statements imply a very strong effect—the sort of effect that should mean that if you know somebody’s musical taste you can make fairly accurate predictions about his or her personality.

If you’re starting to be skeptical, you’re not alone. Jonathan Bellman, at Dial M, has a pretty good rant you might want to read.

Knowing that the media is notoriously bad at accurately reporting scientific findings, I wanted to know what we’re really talking about here, so I e-mailed Professor North with a few questions.

First, the correlations he found are actually quite weak. “Typically only 5-10% of the variance,” he told me. So the average differences among the populations is actually pretty small, although he assured me that the results met the standard criteria for statistical significance (p<.05), so the findings are probably real. But with such a low correlation, it’s not reasonable to say that “classical music fans [are] shy;” at best you can say that on average classical music fans are slightly more shy than the general population. Which is potentially interesting, but doesn’t sell newspapers.

The next question is whether the experimental design was any good, and whether these results mean what they sound like they mean. The news reports and the press release offered by the university talk about personality characteristics on six axes: “Self-esteem,” “Creative,” “Hardworking,” “Outgoing,” “Gentle,” and “At ease.” These seemed like fairly strange characteristics to choose, so I asked North how he collected this data. He told me that “We used a short form of the big five. This means that the measure of personality is no less or more valid than that used by most other research in psychology.”

The “Big Five” is a set of five dimensions of personality which are broadly accepted by psychologists, and for which there are standard assessment methods. The five factors are Openness to Experience, Conscientiousness, Extraversion, Agreeableness, and Neuroticism. You will have noticed that these five factors are different from the six characteristics in the press release and the news reports, which is indeed strange. My best guess is that somebody decided that the public wouldn’t be able to handle the psychology terms and came up with their own, so probably “Creative” is “Openness,” “Hardworking” is “Conscientiousness,” “Outgoing” is “Extraversion,” “Gentle” is “Agreeableness,” and “At Ease” is “Neuroticism.” That leaves “Self-esteem” unaccounted for. Perhaps North assessed self-esteem separately and neglected to mention it in his e-mail? Or perhaps he decided to split “Neuroticism” into “At Ease” and “Self-esteem?” It’s not clear. What is clear, though, is that whoever decided to abandon the standard terminology increased the risk of misinterpretation of the results.

And the results have already been badly abused. Chris Green, writing in The Independent, says:

“People who listen to indie bands are miserable shaggy-haired layabouts, while fans of rap music are bold, brash and brimming with self-confidence.

Rather than mere narrow- minded stereotyping, these are the results of an extensive psychological survey of more than 36,000 music lovers, which confirms, once and for all, that our musical tastes really do reflect our personality. But the study’s most remarkable discovery is that refined lovers of classical music share a high number of personality traits with those who prefer rocking out to heavy metal.”

The results of the study, of course, say no such thing. (It’s not pertinent to the main discussion, but note as well Green’s snobbishly condescending surprise at the idea that classical music fans and heavy metal fans have many personality traits in common.) Green can’t be held solely responsible for his exaggerations, though, because the Heriot-Watt press release is almost as bad, giving no indication of the smallness of the effect North actually found.

Even if we accept that the results of the study are accurate, it’s very difficult to say what they actually mean. We certainly can’t make any claims about causation—did taste in music shift personalities, did personality traits impact musical taste, or did some third factor drive both? (Increases in ice cream sales correlate with deaths by drowning, but it’s not because ice cream causes drowning.) I don’t know whether North controlled for demographics like gender, race, and nationality. Weak correlations between musical taste and personality traits could result from stronger correlations between musical taste and gender and personality traits and gender. We also know that musical taste tends to be bundled with subculture affinity. The results of this study could easily be a product of stronger correlations between personality traits and sub-culture grouping. Perhaps people with similar personality profiles tend to cluster together, and people who group themselves tend to develop similar aesthetic tastes. Furthermore, one of the challenges of the Five Factor model is that it relies on self-reporting, and self-reporting can change based on a person’s attitude toward the factor in question. Belonging to a cultural group that, for instance, self-identifies as non-hardworking might well skew your answers to the questionnaire. I’m sure there are any number of additional possibilities as well.

Given the difficulty in knowing what the results of this study actually mean, it can be dangerous to leap, as the media so often does, to faulty conclusions. When Eric Harris and Dylan Klebold killed twelve students and a teacher at Columbine High School, they were initially mistakenly identified as members of the Goth subculture. There was a national backlash against Goths. Bands like Marilyn Manson and KMFDM were blamed. Kids who wore the wrong clothing or listened to the wrong music were considered suspect and potentially dangerous. That’s an extreme and somewhat limited case, but if Chris Green’s readers come away from his article thinking that Adrian North has scientifically proven that fans of indie music are “miserable shaggy-haired layabouts” and other such nonsense it will be a real shame.