Cognitive Daily

Yesterday’s demonstration about perception of tempo simply didn’t work. If we had successfully replicated ~~Schultze’s~~ Vos et al.’s study, we would have seen a systematic bias in the results.

I’d like to give this one more shot (I promise this will be my last attempt!). There are a couple reasons why the original might not have worked. First, people might have been using the visual progress of the audio player to help determine whether the clips were slowing. I’ve corrected that by narrowing the width of the player so you can’t see the progress of the clip as it plays. Second, because the notes were long piano notes, listeners might have heard differences in the tonal quality of the notes as they were played longer. I’ve corrected that by using a synthesizer voice, and now all the notes are staccato, and therefore the same length. I’ve also added a “held steady” option to the polls, in case you believe the excerpt has neither sped up or slowed down.

Otherwise, the instructions are the same as before. Play each clip, and listen to see if the tempo (rate the notes are played) is speeding up or slowing down. Some of them really are speeding or slowing (and they’re different from yesterday), so listen carefully. After we’ve got some results, I’ll let you know which ones actually sped up or slowed down, and we’ll see if we’ve managed to replicate Vos et al.’s study.

Clip 1:

Clip 2:

Clip 3:

Clip 4:

Comments

#1 James R
January 9, 2008

Ok, definitely noticing an effect now in the first two (and the first seems subtle to me, so I’m willing to believe that it’s not doing what I think it is).

I think those staccato notes have probably made the difference.
#2 Chris
January 9, 2008

These examples are very obvious. Two changed tempi, two stayed the same. I won’t even bother with mentioning which did what.
#3 Dave Munger
January 9, 2008

Yep. The survey results are very accurate as to which is which. Schultze’s experiment used a forced-choice paradigm, but I don’t think we would have seen results even in that case.

I’m not going to run any more versions of this, but I’m curious as to whether readers have any ideas why this didn’t work.
#4 Gav
January 9, 2008

Maybe the original Schultze study was flawed. It seems rather an odd finding, whatever.

A question that’s often intrigued me regarding perceptions of tempo and the “right” tempo for a picece, is Beethoven’s metronome marking of minims (half notes) = 138 for the first movement of the Op 106 Sonata. It actually works quite well at half that tempo – could it have been a (rare) typo on Beethoven’s part?
#5 Freiddie
January 9, 2008

Maybe the actual experiment is longer than this.
#6 Alvin
January 9, 2008

In the original Schultze study, were subjects allowed to hear the sample more than once? The player controls in this demonstration allow for participants to play the clips however many times they desire, which would lead to the ultra-accurate results coming from this experiment. The listener in this case can simply tap along to the rhythm as it ends, and then replay the clip. If listener is now tapping faster than the clip’s initial rhythm, it is assumed that the sample had sped up, and so forth.
#7 Eric
January 9, 2008

I didn’t need to play them more than once, but I’m also a musician – though I’ve never thought myself to have a great sense of tempo. But it was pretty obvious which were which, in my opinion.
#8 Freiddie
January 9, 2008

Maybe the actual experiment is longer than this.
#9 bobalu49
January 10, 2008

The changes made certainly made it easier to tell which direction the tempo was going.
Some possible explanations:
Schultze’s sample was more homogenous – certainly more under the control of the experimenter.
The study design and conditions were not equivalent to the earlier study – probably the most likely explanation. On the other hand, the conditions of this online study provide some good information on how to communicate tempo accurately over the web.
I read of a replication of Asch’s social conformity experiment using subjects isolated from the rest of the group and communicating only via email or the equivalent (IM, chat session) but in which the other participants were actually programmed in advance. Just how that relates to this study is not totally clear, but maybe giving contrary results to the scores would affect the later responses – in other words, if the subject is told they are wrong, will they continue to trust their own ears or will they get confused? Preventing replays would be a must. Perhaps using the original piano tone to make the actual tempo a little more difficult to discern might affect responses. Also would be necessary to have a questionnaire to see people’s reactions to being given contrary responses to their input.

Thanks for putting this study out on the internet, even if the results didn’t support the earlier study. That is interesting data to know, regardless of the reasons why.