Tuesday, August 08, 2006

Are dirty and suggestive lyrics forcing our children into sex? (Probably not.)

There's been a ridiculous amount of attention (see here for a bunch of Google News results) over the last few days to a study purportedly showing that exposure to "sexually degrading" lyrics in popular music leads to "risky" sexual behaviour among teens. Very lurid, and very much plays into conservative fantasies about controlling young people's sex lives.

I don't seriously think any reported actually read the study, though. Full text is here.

Unless you're a total science geek, you can usually get away with reading the abstracts on these studies. They'll tell you everything you really need to know, while avoiding technical details of how studies were constructed and experiments were performed. (Not that that stuff is irrelevant, but it takes some expertise to figure out what it all means.)

BACKGROUND. Early sexual activity is a significant problem in the United States. A recent survey suggested that most sexually experienced teens wish they had waited longer to have intercourse; other data indicate that unplanned pregnancy and sexually transmitted diseases are more common among those who begin sexual activity earlier. Popular music may contribute to early sex. Music is an integral part of teens' lives. The average youth listens to music 1.5 to 2.5 hours per day. Sexual themes are common in much of this music and range from romantic and playful to degrading and hostile. Although a previous longitudinal study has linked music video consumption and sexual risk behavior, no previous study has tested longitudinal associations between the content of music lyrics and subsequent changes in sexual experience, such as intercourse initiation, nor has any study explored whether exposure to different kinds of portrayals of sex has different effects.
While the language is a bit loaded, the hypothesis is clear: teen sex is considered a problem, teens like music, perhaps there is a correlation or causal connection between teen sex and music.
DESIGN AND PARTICIPANTS. We conducted a national longitudinal telephone survey of 1461 adolescents. Participants were interviewed at baseline (T1), when they were 12 to 17 years old, and again 1 and 3 years later (T2 and T3). At all of the interviews, participants reported their sexual experience and responded to measures of more than a dozen factors known to be associated with adolescent sexual initiation. A total of 1242 participants reported on their sexual behavior at all 3 time points; a subsample of 938 were identified as virgins before music exposure for certain analyses. Participants also indicated how frequently they listened to each of more than a dozen musical artists representing a variety of musical genres. Data on listening habits were combined with results of an analysis of the sexual content of each artist's songs to create measures of exposure to 2 kinds of sexual content: degrading and nondegrading.
What should be flagged here is the division of music into these two categories: how the division is made might sway the results. Other than that, the design seems fair.
OUTCOME MEASURES. We measured initiation of intercourse and advancement in noncoital sexual activity level over a 2-year period.
Again, fair enough.
RESULTS. Multivariate regression analyses indicated that youth who listened to more degrading sexual content at T2 were more likely to subsequently initiate intercourse and to progress to more advanced levels of noncoital sexual activity, even after controlling for 18 respondent characteristics that might otherwise explain these relationships. In contrast, exposure to nondegrading sexual content was unrelated to changes in participants' sexual behavior.
So, they have a statistical correlation between exposure to degrading (the meaning of which we will have to read on to find out) sexual content in music and initiating more sexual activity at 13 to 18 years old (the initial 12 to 17 year old group plus one year).
CONCLUSION. Listening to music with degrading sexual lyrics is related to advances in a range of sexual activities among adolescents, whereas this does not seem to be true of other sexual lyrics. This result is consistent with sexual-script theory and suggests that cultural messages about expected sexual behavior among males and females may underlie the effect. Reducing the amount of degrading sexual content in popular music or reducing young people's exposure to music with this type of content could help delay the onset of sexual behavior.
It's important to note that scientific papers of any significance will rarely, if ever, give a normative conclusion. And this is no exception: all that's claimed is a possible causal connection (indeed, the implied test of the causal connection seems to be something like Mill's method of concomitant variation -- scroll down for the definition), which could be deployed to control behaviour.

Immediately, then, the headlines are all wrong. It's not all sexual lyrics -- only "degrading" ones. And there's no suggestion that it is better or worse for teens to engage in sex earlier or later -- no normative conclusions are being drawn.

So, to deal with the lingering issue (about whether this study really shows anything interesting at all): what does "degrading" mean? Well, this:

Two raters independently coded the lyrics, obtained from Internet Web sites, of all songs (N = 193) from each of the 16 albums. The unit of analysis was the song. Raters first judged whether a song contained 1 or more references to sexual behavior (implicit or explicit references to intercourse, oral sex, or other sexual acts). For each song deemed to contain ≥1 sexual reference, raters then judged whether the song contained only nondegrading references to sex or contained ≥1 degrading sexual reference. Thus, these classifications of content were mutually exclusive, and the degrading/nondegrading designation accounted for all of the instances of sexual content.
So, the division into degrading/nondegrading was done by two people (who are unnamed and whose potential for bias has not been explored in the study). These two people ignored all songs which had no references to sex -- which is a pity, as that might've been an interesting control, but it is within the bounds of the study. However, these references could be "implicit", which is a dicey proposition at best. What marks off an implicit sexual reference that's actually in a song from an implicit sexual reference that a person listening to the song insists is there? Again, no information is provided. Among the list of songs with references, "raters then judged whether the song contained only nondegrading references to sex or contained ≥1 degrading sexual reference." So, in total, two people reviewed the songs and, based on their own judgement (as it could be "implicit"), decided if there was at least one sexual reference, and then decided, based on their own judgement, if at least one of the references was degrading.

In short, it all comes down to whether we can trust the judgement of these two people. What do we know about these people? Well, we know that they tended to agree with each other:

To establish interrater reliability for classifying the type of content in a song, raters double-coded one third (n = 63) of all of the songs. These songs were selected via stratified random sampling, with artists as strata. Interrater reliability was satisfactory (Cohen's {kappa} ranged from 0.74 to 0.92).
And that's it. So, the only check on whether the judgement of the raters was valid was whether the judgement of the raters was reliable. In other words, the study is junk science par excellence: there's no basis given on which we can conclude it measures what it claims to be measuring. Thus there's no reason to take it seriously.

No comments: