Music has been associated with drug use for decades -- from the flower children smoking weed at Woodstock to jazz great Charlie Parker getting hooked on heroin, it seems that every type of music has a drug that we associate with it. Last month we discussed a study where college students were asked what type of substances they thought music fans of 14 different genres of music were likely to use. Here are the results:
But are these stereotypes accurate? Rentfrow and Gosling ran into an ethical barrier when they tried to confirm whether the stereotypes they found were true: They didn't want to ask under-age college students about their alcohol and illegal drug preferences.
Assisted by the anonymity of the internet, we decided to see if we could fill the void in Rentfrow and Gosling's work with our own survey, and despite the privacy concerns, we received over 500 responses. So how did our readers' preferences compare to the stereotypes identified by Rentfrow and Gosling? Here are our results:
This graph charts the correlation between respondents' preferences for a each genre rated on a scale of 1 to 7 and their estimation of their likelihood of using each substance, again on a scale of 1 to 7. Because of our large sample size, any correlation below -.1 or above .1 is statistically significant (p < .05). As you can see, our results match fairly well with the stereotypes: Classical music listeners prefer wine to beer, while rock and rap fans are more likely to drink beer than wine. However, although rock fans are stereotyped as preferring marijuana to wine, our results show no statistical difference in rock fans' likelihood of using marijuana over wine. And while the stereotype shows rock fans as more likely to use Ecstasy than rap fans, our results showed no significant correlation between liking rock music and using Ecstasy, while rap fans are extremely likely to use Ecstasy.
One potential problem with Rentfrow and Gosling's method is that they only asked music fans to rate their preference for music on a scale of 1 to 7. Thus, one person could rate several different genres as "7". How do we know whether she is a rock fan or a classical fan if she gives each genre the same rating?
To address that issue, we had fans rate their preferences in a different way (in addition to rating on a scale of 1 to 7): they assigned a percentage to each genre; all genres had to add up to 100 percent. If you rate both classical and rock a "7" on Rentfrow and Gosling's scale, but you actually listen to classical music 90 percent of the time, doesn't that make you more of a classical fan? Let's see how the results change when we use this method of rating:
Now only Rap, Electronic, and to a lesser extent Rock fans expressed any significant positive correlation with substance use. Rock fans were significantly more likely than chance to use only beer and hallucinogens. Classical fans had significant negative correlations with every substance except wine, which was still a marginally significant negative correlation with r = -.08. Far from being wine snobs, it appears that most Classical fans prefer to avoid pretty much all mind-altering substances. The preferred genre of wine drinkers? Alternative, with a .12 correlation between wine drinking and listening. No other genre was significantly correlated with wine drinking in this measure.
Some other interesting correlations: Heroin and country, .24, heroin and folk, .14; heroin and soundtracks, .17, barbiturates and soul, .17, cocaine and soul, .20.
By popular demand, we also asked readers about subgenres in their favorite musical genre. We received a vast array of subgenres, but there were so many that it's going to be a significant project to analyze the data. Now that I've stripped out the IP addresses of respondents (and deleted the original data file), I'd be happy to share this data with anyone who's interested.
Well done. I participated and was anxious to see the results. Very interesting.
So what does a negative correlation imply in this study? I'm assuming a positive correlation implies you're more likely to use the substance and a nonsignificant correlation means you're not likely to use it (or is that what the negative correlation means - if so, what does nonsignificant mean?) Interesting stuff!
Very interesting, I was looking forward to seeing the results.
Though I don't seem to correlate with the correlations.
I'm not surprised by the "heroin and country" correlation; rather, I'm surprised that there aren't any other strong correlations for the country music genre. I suspect/observe a lot of substance abuse in the rodeo crowd (both participants and spectators), to be honest. Out of curiosity...what percentage of your respondants listed country as their top genre?
A positive correlation between X and Y means that more of X corresponds to more of Y. A negative correlation means that more of X corresponds to less of Y. So there might be a positive correlation between number of cars on the road and pollution. There might be a negative correlation between, say, number of bikes on the road and pollution.
Nonsignificant correlations means there is no relationship between the two values, such as number of pigs in Russia and number of violins in Peru.
Could you please rerun the second two graphs using different symbols in addition to different colors? For us semi-color-blind types, the three dark lines all look the same. Thanks.
I hate to be a graph-nazi, but line graphs are inappropriate for categorical data. Lines indicate progression over time. So unless you're trying to show a progressive drug-use escalation trend, scatterplots or bar graphs should be used.
Interesting findings nonetheless.
I like the line graphs because they help me see the connection between points better than the color alone would, with a bar or scatter plot. The slopes of the lines don't have any plausible interpretation, so for me at least, it's not confusing.
But I have to ask you about the statement "most Classical fans prefer to avoid pretty much all mind-altering substances". Classical fans have about a -.2 correlation on most things. I think that means that the more serious classical fans are on average, *slightly* less likely to use drugs than less serious fans, but that individual variability far outweighs the variability accounted for by the music. It doesn't address mean or median, which is what your statement is really about.
I'm no statistician though -- please set me right if I've got things confused.
Thank you for doing this study and sharing it. The results are more interesting than the original study.
I participated in the study and found the questions a little tricky, especially the parts about actual drug use. I understand that privacy concerns may make collection of this data difficult. However, disregarding that issue this is how I might have posed the drug use questions:
1) Which of the following recreational substances do you use? (Then the list)
2) Of these substances, how often do you use them?
Beer: a) daily b) more than once a week c) once a week d) between once a week and once a month e) once a month f) less than once a month
There are many substances on the list that I have never tried and most likely will not - however, the question was posed (if I recall correctly) as 'how likely are you to use...' It is somewhat likely that given the opportunity (say, traveling abroad to a place where it was legal (haha)) I might try some of the substances. However, in reality, I've only consumed a very few of the items in question.
I think another important addition to the drug list would be caffeine and nicotine. Certainly they are drugs and likely the most abused of all the ones listed.
How is it unethical to ask underage people about drug and alcohol use? In fact, isn't this important information to have? How do you think we identify risk factors for risky behavior? I think that was an excuse rather than a well-thought out reason to not complete a study.
Would be a nice Last.fm project
Thanks for this post! This study seems like it confirms lots of the stereotypes about classical music listeners - funny how it follows the curve for religious music listeners.
Maybe you could also get smoother lines if you rated beer a stronger drug than wine - we wouldn't get that odd effect around wine. But then again, maybe that's the point - wine's social status changes how we perceive its strength.
Fred Child mentioned this on his Performance Today blog - check it out: