Father McKenzie Music

Research: Music Cognition

Lyric Intelligibility

Working with David Huron at Ohio State, I conducted as serious of experiments related to the intelligibility of sung lyrics. Anecdotal examples of lyrics being "misheard"—sometimes called a "mondegreen"—are quite common. We've all had it happen to us! In fact, thousands of examples have been documented on pages like KissThisGuy.com and AmIRight.com. Famous examples include:

• Excuse me while I kiss this guy (actually it's kiss the sky)—Jimi Hendrix, Purple Haze
• There's a bathroom on the right (actually it's a bad moon on the rise")—John Fogherty, Badmoon Rising
• The ants are my friends. They're blowing in the wind (actually it's "the answer my friend...")—Bob Dylan, Blowin' in the Wind

Intelligibility in Twelve Song Genres

For this experiment we recruited thirty undergraduate music students from Ohio State University. Each participant listened to 120 short musical excerpts (in randomized order) drawn from 248 different different commercially recorded songs, sampled randomly from twelve different genres: Avant-garde, Blues, Classical, Country, Folk, Jazz, Musical Theater, Pop/Rock, Rhythm and Blues, Rap, Reggae, and "Religious". Each excerpt was a single sung phrase, sampled randomly from within one song. Here are two examples of excerpts we used:

Participants heard an excerpt once, and then were asked to type what they though the lyrics were.

For our analysis, we simply counted how many words in each lyric each participant identified correctly. We estimated word-identification accuracy using a mixed-effects regression model. Our results are summarized in this graph:

As you can see, some genres are indeed much more intelligible than others: people tend to understand over 90% of jazz singers' words, while classical singers are understood less than half the time (on average). However, you can also see that there is lots of variability within genres as well: the little gray dots in the figure show each individual excerpt's intelligibility, and the brackets show the 95% and 80% confidence intervals (kind of like "margin of error") for the estimates for each genre.

Metric "Fake-Outs"

In 2014–2015 I conducted two experiments concerning the perception of what theorist Justin London calls metric "fake-outs." Fake-outs occur when the rhythm at a beginning of a piece seems to suggest one meter, but as the music continues (especially when the drums start playing), you realize you've been fooled and the meter is actually different then you thought. A striking example occurs in the song Wake Up by XTC

At the beginning, it seems like the regular guitar strokes (paned to the right in the stereo mix) are landing on the beat. However, when the drums start playing, it becomes clear that they are actually off beats!

Fake-outs are interesting because they show us something about how music listeners unconsciously infer meter from music. They are unusual edge cases where something goes "wrong" with our metric induction. My most interesting finding, however, is simply how variable people's perceptions of fake-outs are: when two people listen to the same song, it's quite common that one is "faked-out" while the other isn't! This isn't a fascinating example of how variable music perception can be, even at the "basic" level of experiencing meter.

(Exploratory) Experiment 1

Most of these examples did indeed "fake-out" at least some of my participants. However, several examples—notably Metallica's Holier Than Thou and Beethoven's Concerto 1, Movement 3—were close to fifty-fifty splits, with roughly half of my participants faked out, and the other half not!

Experiment 2

Working on and analyzing my first experiment, I noticed that almost all fake-outs involve hearing the first onsets in the music as the downbeat. In other words, fake-outs tended to happen when notes that are "supposed" to be heard as pick-ups are misheard as downbeats. I hypothesized that listeners tend unconsciously assume that the first thing they hear is the downbeat. In order to test this hypothesis, I created edited versions of 14 of my excerpts, where I either removed a pick up from the original, or added a pick-up that was no supposed to be there. Here are two examples of original and edited excerpts:

• the Foo Fighters: All My Life

  Original: Edited:

• Metallica: Ain't My Bitch

  Original: Edited:

I played these edited excerpts, and their matching unedited versions, to a new group of participants and measured how they tapped along to the beat. Sure enough, my results are strongly consistent with the idea that just removing/adding a pick-up can dramatically change our metric perception in some cases. Of course, there are other factors at play so not all songs are equally affected (the two examples I give above were two of the most strongly affected).

For more details about these experiments, and fake-outs in general, you can read my (as of yet unpublished) paper, entitled Beat and Switch.

Other Fake-Outs

I've begun (slowly) tabulating all the examples of songs with fake-outs that I know of. Try listening to them to see if you too are "faked-out" by these tracks: