Tuesday, November 8, 2016

Earworms Revisited

Every few months we have another article on the topic of "earworms", those catchy bits of melody that get stuck in your head. The latest is The Science Behind “Earworms”:
Don’t worry, there’s a reason why “Don’t Stop Believin'” gets stuck in your head for days every time you hear it. In fact, there are a couple of reasons. And they’re backed by science.
Songs that get trapped in your head for long periods of time, commonly called “earworms,” are the subject of a study by Durham University (in England) researcher Dr. Kelly Jakubowski, who recently published a paper on the subject in Psychology of Aesthetics, Creativity, and the Arts. Jakubowski and her team found that earworms have three distinct qualities that separate them from other songs: pace, melodic shape, and unique intervals.
All these articles have the same characteristics:

  1. They claim to have the answer to some age-old and puzzling question
  2. They take a scientific approach, and
  3. They display a near-total ignorance of hundreds of years of music history and theory
Given those shortcomings, it is not surprising that the "answers" they come up with are shallow and, to any musician, unconvincing. Let's have some more from the researchers at Durham University:
Pacing, the team found, is crucial. Many commonly cited earworms have upbeat, danceable tempos, but are still slow enough to easily track. Most earworms follow the melodic preferences of Western pop music, which in turn follows many of the melodic contour patterns in nursery rhymes. “Twinkle, Twinkle, Little Star,” for instance, has a rising pitch in the first phrase that falls in the second, a common trait of earworms. (Maroon 5’s “Moves Like Jagger” was specifically called out by the study for this.)
But childlike simplicity and a peppy tempo aren’t enough to make an earworm. A true earworm changes its game up with at least one unusual interval structure, defined by the study as “unexpected leaps,” repeated notes, or any other idiosyncratic tick in the song’s composition that makes it memorable, in addition to catchy.
The basic problem with this kind of "research" is that it is nearly always a case of using the wrong set of tools. Science, at least this kind of science, is nothing more than statistics and surveys with a thin veneer of technical vocabulary. The vocabulary has been eliminated from the news story, but you can see it in the original paper here where an "earworm" is given the scientific moniker "involuntary musical imagery". The idea of sounds being misnamed "images" already makes me uneasy. Looking over the original paper to get a sense of how they approached the question musically, I see that again, it is purely a question of statistics:
First- order features are features that are calculated based on the intrinsic content of a melody itself, such as the average note duration, average interval size, or pitch range of the melody. Second-order features, also called corpus-based features, are features that compare a melody to a larger collection or corpus of melodies (generally comprised of music from the same genre or style as the melodies that are being analyzed, such as pop songs or folk songs). For instance, one example of a second-order feature might measure to what degree the average interval size within a particular melody is common or uncommon with respect to the distribution average interval sizes within a large corpus of comparable melodies.
This is inevitable if the tools used are scientific. In science, you can only examine those things that you can measure and you can only measure those things that you can assign numbers to. Now, of course, you can look at music entirely in terms of numbers: tempo is how many beats per minute, pitch is how many vibrations per second and so on. But those things are nothing but the externals and bear as much relation to a musical performance as a recipe does to the dish on the table. The musical expression as implied and encoded in the score and as evoked by the performer and as experienced by the listener is simply another order of reality from the notes on the page and the numbers that science can deduce from them.

Let me explain why. Really catchy earworms each have a memorable quality. This is why they, and not a thousand generic examples, are memorable. It is their individuality that makes them memorable. This individuality is precisely the quality that cannot be captured by any methodology based on statistics. To make another metaphor, imagine trying to identify Albert Einstein when he was a patent clerk by doing a statistical survey of patent clerks. That is the equivalent of trying to identify those qualities that make for an earworm by doing a statistical survey of catchy tunes. If you look at the quotes above you will see that each description begs the question (as in "assumes the conclusion"). A rising pitch in the first phrase that falls in the second, a "common trait in earworms" is found in thousands of melodies that are NOT earworms. This same critique applies to every single characteristic claimed as being indicative or characteristic of earworms: they are all characteristic of all popular melodies, nursery rhymes, Haydn string quartets, Bach minuets and, for all I know, Brazilian sambas (though I haven't done the research on the latter).

The question of what makes a particular piece of music "catchy" is unanswerable using scientific methodology. What is extremely odd is that the researchers, their subjects, the journalists reporting and nearly all of the readers fail to realize this!

In related news, science also has no answers for questions regarding aesthetic quality, ethics, theology and why I am having trouble organizing my latest composition.

Now for our envoi, and we deserve a good one today. Here is one of the most catchy compositions ever written, by Wolfgang Amadeus Mozart. It's a little divertimento that he referred to as "a little night music" in a letter to his father and so, ever since, it has been known as "Eine kleine Nachtmusik". It was composed in 1787. This performance is by Concerto Koln:


Anonymous said...

Your argument seems to be that there are too few earworms for the concept to lend itself to statistics. But that's doubtful. A significant fraction of pop tunes qualify as earworms. The question the researchers ask is whether suitable software could make a good prediction of whether a new tune winds up being an earworm or not. Perhaps this is impossible to do but that's an empirical question that cannot be decided "a priori." As long as a group of people can more or less agree on whether a tune is an earworm or not, the question above is certainly a scientific one. Of course it does not mean we can answer it.

Bryan Townsend said...

Excellent riposte! But my argument is not that there are too few earworms to be statistically analyzable, it is that what makes a powerful earworm has to do with, among other things, its memorability. To be memorable, the tune has to have some unique quality. Remember that for every catchy earworm there are hundreds or thousands of pieces in very similar styles that are not memorable earworms. My contention is that to examine large numbers of pieces of music to try and isolate earwormness is a category error in the sense of Gilbert Ryle's The Concept of Mind.

Now, mind you, my contention that an earworm has to have a uniquely memorable quality might simply be incorrect. Perhaps there are certain musical techniques such as the "millennial whoop" that I talked about here


...that just stick in people's heads. The more they are used, the less memorable they are, of course. So why would a particular exemplar stick in your head? So I go back to my uniquely memorable notion.

Anonymous said...

My point is that your "unique memorability" hypothesis can be tested. Suppose a computer could predict 90% of the time if a new tune ends up being an earworm. Its existence would invalidate your hypothesis. But the existence of such a thing can be tested. If such a program existed, we still might not know what makes a tune an earworm but we would know that statistics suffices to predict earwormhood. I think that knowledge would be interesting to have. Just like it's interesting to know that a computer plays Go better than humans.

Bryan Townsend said...

Well, I guess we can wait until someone manages to do that!