Some insight into how we decide how fast music should be played

Blogging on Peer-Reviewed ResearchListen to the following three short audio samples. Your job is to say whether the tempo (the rate at which the notes are played) is speeding up or slowing down. Even if it sounds like it's maintaining the same tempo, make your best guess as to whether it's speeding up or slowing down.

[Update: There's a new demo here. And see this correction]

Clip 1:

Clip 2:

Clip 3:

If the results here follow the pattern found in a number of studies, there should be a bias in the responses (and yes, some of the clips really are slowing down or speeding up). I'll explain what's supposed to happen later on in the post.

Music researcher Daniel Levitin (also a former music producer for acts like Carlos Santana and Blue Ãyster Cult) has found, working with Perry Cook, that despite these biases in perceiving tempo for single notes, even untrained individuals have a remarkable memory for the tempo of songs they know well. They can sing them back, a capella, at almost precisely the same tempo as the recordings they hear on the radio.

If some studies have shown biases in the way we perceive tempo, then why are we so accurate with familiar songs? It's probably not due to the subtle variations in tempo produced by artists: One study had pianists play songs at several different tempos -- deliberately too slow or fast, then used a computer to speed up or slow down the tempo of the same music. Even musically trained listeners couldn't tell the difference between the artificially sped-up music and music played faster by the musicians.

Sandra Quinn and Roger Watt recently explored the phenomenon using a different method, which I'll discuss below.

They took 23 Scottish fiddle songs and played them on a synthesizer as marked in their musical scores. They then artificially slowed and speeded each tune's tempo by 10, 20, and 30 beats per minute. Student volunteers listened to each song, as well as the original version, in random order (so they weren't listening to the same song over and over again). They were simply asked if the song sounded too slow or too fast. This graph shows some of the results:

i-791e15301495e70732d94088ebc47b86-quinn1.gif

The graph charts the proportion of students saying each of three excerpts was "too fast." As you might expect, the faster the excerpt was played, the more likely students were to rate it as too fast. But as you can see, for excerpt #1, a tempo of just over 60 beats per minute was rated too fast by nearly all listeners. Meanwhile, 100 beats per minute was seen as "too slow" for excerpt 24. Each song had its own ideal tempo, where half the listeners rated it too slow and half rated it as too fast. These ideal tempos ranged from a low of less than 50 beats per minute (as in Clip 1 above) up to well over 100 beats per minute (as in Clip 2 above).

An analysis of the measurable musical features of the songs found that most features (for example, whether the music was in a major or minor key) bore no significant relationship to the ideal tempo of the song. The only feature that did correlate significantly was the number of descending intervals, which correlated with tempo at r = 0.49.

So how is the optimal tempo picked?

Let's return for a moment to the demo at the start of this post. It's based on a 1978 study by Hans-Henning Schulze which found that people could accurately detect tempo changes -- but only for single notes repeating at 100 beats per minute. Any faster [correctin: a 1997 study by Vos et al. finds], and even tempos were perceived as speeding up, while slower tempos were perceived as slowing down. The first of the three clips was played at 50 beats per minute and maintained a steady tempo. If we replicated Vos et al.'s results, most people should say this one slows. The second clip was again an even tempo: 200 beats per minute. Most people should say this one speeds up. The third clip starts off at 100 beats per minute. Respondents should have been able to accurately say whether it sped up or slowed down. I'll leave discussion of that one for the comments section.

Do we have an internal clock that runs at 100 beats per minute? Quinn and Watt's results suggest that if we do, we don't apply it willy-nilly to every song we hear. Instead, something about content of the songs suggests an appropriate tempo. While their research doesn't give us a definitive answer as to what that tempo might be, they do have some hunches. If a song has many "strong" events -- events that vary simultaneously across several musical dimensions -- then the authors suggest that these sorts of songs might be preferred at a slow tempo, compared to songs filled with weak events. Listeners want to savor those nuances, and can only do so when the song is played slowly enough.

Quinn, S., Watt, R. (2006). The perception of tempo in music. Perception, 35(2), 267-280. DOI: 10.1068/p5353

More like this

I could tell that no tempo changes took place in the first two samples but there was no option provided for 'no change'. Since I had to choose between 'speeding up' and 'slowing down' I think any results for this survey will be rather skewed.

As to sample 3, since you say that "yes, some of the clips really are slowing down or speeding up" and since clips 1 and 2 remained constant, sample 3 must either be speeding up or slowing down. I guessed 'slowing down'.

James, are you saying you knew for sure that #1 and #2 were constant, but you weren't sure about #3? That's an interesting result. Before I say what's actually going on with #3, I'd like to hear about other readers' subjective responses to the clips.

The first two were constant, and I flipped a coin to determine which button to push. The last one slowed.

I can understand the utility of the forced-choice paradigm in this experiment--in psychophysics, the forced choice is the tool to tease apart bias (anyone who has been tested for corrective lenses, using the "better like this? Or this? ... Better like this? Or this?" method has faced a forced choice between two identical alternatives, if the procedure was properly done). In practice, though, it was very frustrating to know that the first two were constant, and I was not allowed to choose "constant". One would think that if the phenomenon was robust, it would show up even without the forced-choice methodology.

I agree partially with James - the first two seemed to be at a constant tempo to me. I listened to them over and over and still ended up guessing. I picked "slowing down" on the last one half way through hearing it.

Clip #1 and #2 sounded pretty much constant to me, while #3 definitely sounded like it slowed down, to my ears.

But as a napalm death fan I've got to say that the only proper way to play music is as fast as possible.

I could easily tell that clip 3 as slowing down. Clip 2 sounded constant, but there wasn't a choice for that. Clip 1 was a lot harder until I subdivided the beats (old habit from orchestra).

By Tim Buchanan (not verified) on 08 Jan 2008 #permalink

(Different James to one above). I was 90% sure the first two were constant, and I'd reckon that the third clip was slowing down.

For reference I picked 'speeding up' for the first two clips, but please don't take them as showing (or not) the effect you've described.

I also just picked randomly for the first two. I couldn't hear any change in them at all.

Sounds to me as though the 3rd one gets slightly faster to begin with, then slows down very noticeably after the 4th or 5th note, so I went with slowing down. Curious to see if it actually is getting faster at the start or if it's some other trick my ears are playing.

i was also quite certain that the first two clips did not change tempo. i think you should ask respondents about their musical training.

#3 is definitely slowing down, and from the poll results for that question, that much is pretty obvious to most.

#2 to my ears was a steady tempo, though I marked it as speeding up

#1 was the most interesting to me, as although it seemed like the tempo remained constant overall, it felt like some of the notes were slightly behind the beat while others were dead on. In that case, it doesn't really qualify as slowing down, though, does it?

Okay, as of now, we're not getting Schulze's bias effect. I have a guess as to why it might be happening -- I think because the tempi I chose were even multiples or fractions of 100 bpm, people could easily use their "automatic" 100 bpm clock to determine whether speeding or slowing was occuring.

So I've now changed the tempo on those two clips. Clip 1 is now 40 bpm and Clip 2 is now 180 bpm. Let's see if that makes any difference.

Data at the point I made the change:

Clip 1
Speeding up: (47%)
Slowing down: (53%)
Total votes: 72

Clip 2
Speeding up: (51%)
Slowing down: (49%)
Total Votes: 70

Naw, they still sound as if there is no change in tempo. I have no musical training and a terrible ear (for post #9, lila). But to me, the first two clips showed no speeding or slowing.

I am in 100% agreement with Comment #10 Michael.

I too perceived a dragging if the beats in Clip 1 that corrected towards the end.

I voted with Michael.

I listened to it after the change (but not before). I felt that the first two clips were not changing, but that wasn't an option, so I randomly chose one or the other.
The last one was definitely slowing down.

By MiddleO'Nowhere (not verified) on 08 Jan 2008 #permalink

In the second-to-last paragraph, you say that most people should perceive the 200bpm clip as slowing down. Do you mean speeding up?

I would vote the same way after the change (fast fast slow), although I still don't actually percieve a difference in the first two.

Perhaps of interest is that at first I thought that every other note on the third clip was being played staccato, but after a second listen I no longer think that's the case.

It seemed obvious to me that clips 1 and 2 were at a constant rate, so the fact that your responses are split pretty much 50,50 on those makes a lot of sense. I'm interested in what was really happening in clip 3. It definitely seemed to slow down at the end (it stopped long before the end of the clip), but it could have been speeding up slightly at the beginning. I'm thinking you were playing with us. :)

Echoing other respondents: 1 and 2 were obviously constant (although I confess to paying conscious attention, I suspect that was the desire?) which I confirmed by taking the best from the first couple of tones, then using it to predict further tones, and finding that I got a match.

3 is (I think) clearly slowing, but I can see where the perception of initial speeding comes from; I didn't replay to check.

I agree with the frustration of not being able to pick "constant" for the first two clips. For clip 1, rather than "doubling" the beat to get to an internal 100 bpm, I did my usual waltz tempo after the first 3 beats and ... it was constant. Ah, well. The second was more assuredly constant, while the third was slowing down.

then again, compared to Hank's taste in music (dude, go for early Painkiller!), I listen to a lot of meta-metric music, where multiple time signatures are going on at once. You get used to drifting over the rhythm intervals and attuning yourself to miniscule changes in tempo.

More on metametric music, if you wish. I wonder what the musicologists and composers would tell the psychologists...

I have another guess as to why people aren't getting the effect: visual feedback. They can use the smooth motion of the cursor to monitor the rate of audio playback. I'm going to see if I can figure out how to make a demo without visual feedback. Hopefully I'll have it up later today.

(any suggestions on how to code this demo would be extremely helpful!)

I had the same experience as James. I listened to the first two clips and they sounded like they had even tempo, but when I listened to the third one it sounded like it was slowing down.

i dont agree with the visual feedback idea (@22). i had my eyes closed, as i tend to do when focusing on sounds, and i still could tell that the first 2 clips held a steady tempo.

I have no musical training, and I was 100% certain that the first two clips were played at an even tempo.

I believe the sound clips I heard via Live Bookmarks (is that RSS, or something like it?) were at a different tempo than the ones I heard on this website itself. On this website, I think the 1st and 2nd clips were constant in tempo, but the 3rd one slowed down. I got a different sense from the RSS-like link.

By --Lisa S. (not verified) on 09 Jan 2008 #permalink

1 and 2 seemed very even. I was a little less sure about 2, because it went so fast, but I was pretty certain on both. #3 obviously slowed down.

Clip 1 and 2 sounded mostly as if they were playing at a constant rate. After playing #1 several times, I followed the instruction to "Even if it sounds like it's maintaining the same tempo, make your best guess as to whether it's speeding up or slowing down." and due to a very slight sensation that it might be speeding up, voted for speeding up. Playing #2 repeatedly still left me with no sensation that it was speeding up or slowing down, and so I did not vote. #3 definitely sounded like it was slowing down.

By Dangerous Dan (not verified) on 10 Jan 2008 #permalink

On sample 1 I could hear no tempo change. I then used visual cues from the cursor to verify that there was indeed, no tempo change. I voted as speeding up, probably because there was seemed to be some rushing on a few notes.
On sample 2, I could hear no tempo change. The visual cue was much harder to integrate. I did not vote.
sample 3 was obviously speeding up.

By debotcher (not verified) on 10 Jan 2008 #permalink

#1 speed up
#2 Don't know, maybe speed up
#3 slow down

By Ulysse005 (not verified) on 27 May 2008 #permalink