Want Perfect Pitch? You Might Be Able To Pop A Pill for That, announces NPR (and dozens of other news outlets, verbatim). When I saw that, like the musician friend on Facebook who had posted it, no doubt, my eyes widened. Achieve perfect pitch — an incredibly useful ability only available to a tiny fraction of humanity — with a pill? I wanted it to be true.
So is it? The claim comes from a paper published by a group of international researchers in Frontiers in Systems Neuroscience. The researchers took a group of 24 college-aged men with little to no musical training and administered an anticonvulsant drug called valproate to half of them; the other half received a placebo.
Valproate (also known as VPA) is designed to inhibit the action of an enzyme called histone-deacetylase (HDAC). According to the researchers, HDAC is responsible for putting the brakes on that amazing time in our toddler years when our brains are plastic and we’re gobbling up information like candy. Perfect pitch, the ability to name any pitch you hear without needing a reference tone, is the poster child for this period — there are no known instances of people successfully acquiring the skill in adulthood. In general, the 0.01% of the population with this skill learned it before the age of six, through not only formal musical training but specific training in pitch identification.
So. If there’s a pill you can take that will stop the chemical that stops your brain from being malleable like a toddler’s, there’s a chance you could take that pill and learn all the things a toddler can learn. In fact, research has shown that adult rats on HDAC inhibitors can recover from a lazy eye, a problem that’s usually irreversible due to that lack of plasticity in adulthood.
The subjects took their pills as directed over two weeks, starting training on the second week so they’d be at full dosage. Each day of the training period, they watched an online video that taught them to associate six pitches with six names (they used people names, not note names, to ensure an even playing field). Then, they were tested.
So what happened? Did they get perfect pitch? Can I go out and buy this drug, already?
Well, yes and no. To quote a Magic 8 ball I once knew, reply hazy, try again.
What I didn’t mention is that this was a two-armed crossover study, meaning that whatever group was on the real drug during the first treatment arm would take a placebo during the second treatment arm, and vice versa. This is a handy way for researchers to get effectively double the data with the same number of subjects. This disclaimer at the beginning of the study puzzled me:
The strong hypothesis is that there might be improvement in the VPA condition in both arms of the design. However, since new training is introduced in the second treatment arm, successful learning in the first arm might carry over to and interfere with learning in the second arm. Effects are thus more likely in the first arm only.
It’s definitely possible that, as a mostly-layperson, I don’t have the training to understand why the second arm wouldn’t show results. That being said, how could successful learning in the first arm possibly interfere negatively with the second arm? If anything, wouldn’t it improve their pitch identification ability? They even used the other half of the Western scale in the second arm, so there were no double pitches learned. I didn’t get it, but I read on.
Sure enough, their hypothesis struck it rich. With their calculations determining “chance” as 3 correct answers, the first arm showed that the treatment group scored an average of 5.09, significantly above chance (p=0.002), while the placebo group scored an average of 3.5, or roughly chance (p=0.21).
(I’ll take a detour from this post to explain the wonder of P values to my non-scientific audience, as I’ve recently learned how to interpret them and that has made all the difference. To put it as simply as possible, the larger the P value is, the more likely the result was to have occurred by chance. The smaller it is, the more likely it was to have occurred because of the treatment. This is known as “significance” — depending on what the researchers decide before they start, a P value equal to or less than 0.01 or 0.02, respectively, is considered significant. So p=0.002 suggests that the treatment group’s awesome score was not a fluke.)
The second arm, just as they suggested, is where things fell apart. I might add that they lost six of the 24 subjects for the second treatment arm (none, they assure us, due to side effects). This could skew the results a bit, but still, extraordinary claims require extraordinary evidence, my friends. And extraordinary it was not: the average correct response was at chance for both groups — 2.75 in the treatment group and 3.33 in the placebo group.
Game over. You haven’t proven that you can “pop a pill to get perfect pitch” with a study that only showed an effect half the time. This doesn’t mean that valproate absolutely won’t help adults learn perfect pitch, but it does mean that we need some bigger, more robust studies to determine that. In the meantime, I’ll be waiting.