New study revives a Mozart sonata as a potential epilepsy therapy

Could this be the return of the “Mozart effect”?

In 1993, researchers reported that after college students listened to a particular Mozart piano sonata for 10 minutes, they showed better spatial reasoning skills than they did after listening to relaxation instructions designed to lower blood pressure — or to nothing at all. And their IQ scores jumped by 8 or 9 points in what became known as the Mozart effect. Even though the benefits were hard to reproduce (and wore off within minutes), the fad of Mozart for babies’ brain development was born.

In that early research, scientists also said listening to Mozart’s “Sonata for Two Pianos in D Major” (K448), helped patients with epilepsy by reducing spikes in neuron activity that can lead to seizures. Again, the results were not consistent enough for other scientists to replicate them. But now, 28 years later, newer methods are reviving the possibility that music can be the calm that prevents the brain’s electrical storm.

Research published Thursday in Scientific Reports says listening to the sonata for at least 30 seconds may be associated with less frequent spikes of certain electrical activity in the brains of people whose epilepsy does not respond to medication. In the study’s 16 patients, spikes fell by two-thirds throughout the brain, but they dropped the most in the brain’s left and right frontal cortices, where emotional responses are regulated.

Study participants listened to this clip from Mozart’s “Sonata for Two Pianos in D Major” (K448).

Robert Quon, a graduate student at Geisel School of Medicine at Dartmouth and lead author of the study, talked with STAT about his lab’s initial skepticism, the tools that allow for better measurement of brain activity, and just what it might be about this piece of music (and not Judas Priest or Buddy Holly) that could prove helpful to people with epilepsy. This interview has been edited and condensed for clarity.

What led you to reexamine the Mozart effect?

When I first joined Barbara Jobst’s lab, I was fortunate to work with a postdoc in the lab, Grace Leslie, who was conducting a study looking at how these different auditory tones — not music, but more pure tones and frequencies — affected the brain waves. The first study we performed was looking at how 40 hertz or this gamma stimulation affected these interictal brain waves — brief events that occur sporadically between seizures — in patients with epilepsy. It sounds essentially just like a buzzing in the speaker.

What is it about these spikes?

Increased numbers of these spikes are correlated with memory loss or reduced cognitive outcomes, and even increased seizure frequency. Therapies that can maybe reduce these spikes could have some proven benefits to patients with epilepsy.

Why did you switch to music?

I was testing patients in that first experiment for only 15 seconds at a time, but even with just those repeated 15 seconds of that gamma tone, a lot of the patients complained that it wasn’t too nice to listen to.

Did you think it would work?

Our group definitely approached this with a very healthy skepticism. And we still are skeptical. A lot of questions have emerged from this paper, but we’re hoping to answer more of them over the years.

How did you determine the music’s effect on brain activity, and how does it compare to methods of the 1990s?

A lot of those studies were using a scalp EEG [electroencephalogram], and that led to a lot of contradictory findings. Signals generated by the cells that are firing in the brain have to travel across a lot of layers, including bone, and a lot of the signal gets damaged or even lost. The group of patients that we tested actually have electrodes placed directly onto their brain surfaces rather than on top of their scalp. They might be considered for surgical resection, so before doing a surgical resection, intracranial electrodes are implanted in the patients to localize where exactly the seizures are coming from so they can cut out the brain region that might be generating these seizure events to hopefully lead to seizure freedom.

That provided us with an optimal window into the brain, enabling us to test the hypothesis that Mozart may or may not reduce these intractable spikes.

Did you try other music?

We played essentially nine different musical stimuli. One of them was Mozart. And then we also modulated Mozart with that gamma tone, like an envelope of sorts. We tested music from different genres and then patients selected the song that they liked the most.

Your paper lists Wagner’s wedding march, but also Buddy Holly, Judas Priest, and Nickelback. What happened?

We were really hoping that one of these preferred pieces, when we modulated the frequencies, would result in significant therapeutic effects. But unfortunately only the original Mozart composition showed significant response. So that led us to investigate further what properties might be responsible for this.

What is it about this piece of music?

When people develop these expectations about a musical piece and when the expectations are suddenly violated — you have the same beat or tempo going on in the song and then a sudden violation to that expectation in your brain — that’s linked with positive emotional responses. We used a machine learning technique and a music information retrieval technique to break this song into segments. And we predicted that transitions out of these longer segments would be correlated with enhanced power in the emotion networks of the brain. And that’s exactly what we found with our analysis.

What’s next?

We want to look at how different elements within the music itself and see how important they are for eliciting these effects. Our next step would be to test whether we could detect other musical stimuli with these preexisting properties, like tempo or timbre, or add these properties through signal alteration. We’ll eventually, hopefully, be able to define an anti-epileptic genre so maybe listening to music from this playlist will reduce chances of having a seizure.

There’s a lot of time needed to test these theories and to tease apart exactly what is important. And we could definitely use machine learning to help us do that. Then after we do that, we can even use machine learning to look for these features within just large Spotify playlists of sorts and test that later.

We have a lot more questions than we have answers from our study. But, you know, that’s kind of common to most of research, right? I think the biggest take-home is that we just want to replicate this in a larger group of subjects for a prolonged period of time to really prove to ourselves that this is happening.

You must hear that sonata a lot.

I listen to it quite often. I’m in the room with the patient, so I hear it every time I’m testing it. And other than that, I’m analyzing the structure of the musical signals from a computational point of view. I’m pretty familiar with it.