As some of you already know (since you were participants), I recently conducted a little pilot study to investigate some aspects of music cognition. My main focus was how different levels of repetition factored into a listener’s preference for a song. In other words, if a certain part of a song is repeated more or less often, does this change affect how much you like or dislike a song?

In popular music, I think we can take for granted that the repetition of portions of a song is one of the basic tools that songwriters use to structure their tunes: “second verse, same as the first.” The chorus of a song, for instance, often reappears exactly the same way at multiple points in the song: lyrics, harmony, melody, instrumentation are all identical from iteration to iteration. Does it matter, then, how often these identical parts get repeated?

I don’t think it’s too much of a stretch to say that at some point, we as listeners would get bored with a song that repeated the same section over and over without any change. Of course, many songs have repetitive harmonic structures, but often a balance is negotiated (in a good song) by having other domains change and thus add interest: melodies get embellished, instruments get added, rhythms change. It’s this balance of variety versus unity - a balance that lots of composers talk about - that keeps our interest in a song. For a stripped down guitar/vocal arrangement, there is going to be more pressure for harmonic and melodic variation than were more instruments involved.

Part of this balance is captured by theories from the realm of experimental aesthetics. There is this one guy, D. E. Berlyne, who basically posits that our optimal preference level occurs when we are experiencing a moderate level of arousal (insert “arousal” joke here). Certain factors increase our arousal: complexity, novelty, ambiguity, etc. Other opposite factors decrease our arousal: simplicity, familiarity, clarity, etc. If a work of art is too complex and/or too unfamiliar, for example, then our preference for that work of art is lower than were the artwork less complex and/or more familiar. Of course, these arousal-increasing and -decreasing factors are entirely subjective, according to Berlyne; what one person perceives as complex (e.g. BeBop to a non-jazz listener) may be simple to another person (e.g. Bebop to a jazz musician).

While I’ve kind of boiled down Berlyne’s theories to what is perhaps a very obvious summary (akin to the story of Goldilocks), I think they can still help us predict what will happen in the case of repetition in a song.

With the pilot study, I had folks listen to a bunch of choruses and rate them on how complex they thought the choruses were and how much they liked them. Here are some examples of different choruses:

		Lucinda Williams: “Right In Time” (chorus)
		Halo Benders: “Don’t Touch My Bikini” (chorus)
		Def Leppard: “Too Late For Love” (chorus)

I used a total 18 different choruses from a variety of different artists and genres to hopefully avoid for any trends related to a specific style. The overall result was that there was a positive correlation between complexity and preference. In other words, the more complex the participants rated a chorus, the more they also seemed to prefer the chorus. To look at the results another way, it seems that overall, none of the excerpts were overly complex enough to create a decrease in preference. In general, the choruses were not “too complex;” if anything, they were too simple.

These results regarding preference versus complexity for the choruses of the 18 songs help to predict what would happen if the choruses then become repeated. Since the complexity/preference ratings imply that arousal levels were at or below optimal levels, then an increase in the familiarity of the chorus (through repetition) should cause a decrease in preference. Remember that familiarity is an arousal-decreasing factor and that we are already assuming that arousal levels are generally low to medium already.

So what occurred next is that participants rated different versions of the 18 songs. Some versions had the chorus appearing once, some versions had the chorus appearing twice (one repetition), and other versions had the chorus appearing four times (three repetitions). Here are some examples of the different versions:

		Kanye West: “Touch The Sky” (chorus 1x)
		Kanye West: “Touch The Sky” (chorus 2x)
		Avril Lavigne: “Mobile” (chorus 1x)
		Avril Lavigne: “Mobile” (chorus 4x)
		Liz Phair: “Never Said Nothing” (chorus 2x)
		Liz Phair: “Never Said Nothing” (chorus 4x)

As it turned out, there was no overall significant difference in the preference ratings for the 1x versions of the songs as compared to the 2x versions. However, the overall preference ratings for the 4x versions of the songs were significantly lower than either the 1x or 2x versions. In other words, listeners didn’t mind having something repeated once (that perhaps adds a sense of unity and balance on a phrase level), but listeners did seem in general to like a song less when the chorus was repeated more than once.

There were and are a lot of potential confounds to this pilot study, and the full results are a bit more nuanced than I’ve presented them here. But considering that it was a pilot study (and not a full-blown experiment), I think the results are promising. Having gone through the process, I think a good experiment could be designed to hopefully shown such effects more clearly.

Of course, each song is its own entity, so making a blanket statement that repeating something more than twice necessarily causes a decrease in preference would certainly be a naive assertion. Let me point out, firstly, the parts of these songs undergoing repetition were fairly long (8-13 seconds for each chorus iteration). Shorter subsections of phrases or motives I think can easily stand high levels of repetition. Where that cut-off point lies is hard to say.

Moreover, we can probably all think of cases where parts of songs get pretty much literally repeated over and over again without a decrease in the songs effectiveness. The ends of songs (often a fade out) is one good example. But a lot of good songs, I’ve noticed, still tweak the instrumentation or embellish the melody in these cases of high repetition, perhaps to keep the listener’s appetite for variety versus unity in balance.

As my own outro, I should point out that the 4x chorus version of the Liz Phair excerpt above is the original version as recorded by Liz Phair. Do you prefer this 4x version to the 2x version that I created? I personally think the 2x version is more interesting, especially since a single iteration of the chorus already involves internal harmonic repetition. Could Liz Phair’s song thus have been structurally improved upon?