Sunday, March 14, 2010

Scientific Analysis of Swing Ratio

My friend Andre Majaes brought this paper entitled "Jazz Drummers' Swing Ratio in Relation to Tempo" to my attention. It provides a scientific look at the ratio of the long-short pairs in swing 8th notes.

What is not known is the exact relation between the long and the short note. It is not specified in the score and music students are often advised to learn it by listening to recordings. Such knowledge would contribute to the understanding of the perception and production of music. It would also be useful for generating synthetic performances on a MIDI sequencer, as well as being helpful for students who wish to learn how to swing.

The essential conclusions they come to are that:

  • Swing 8th notes are not triplet-based nor are they dotted-8th and 16-note pairs. The feel varies from drummer to drummer and, as is obvious to the ear and well-known, becomes more straight as tempo increases.

    It is interesting to note that Jack DeJohnette, compared to the other drummers used in the study, actually stays closer to a 2:1 triplet ratio regardless of tempo than the other subjects.

  • The length of the shorter 2nd note of each pair seems fairly constant regardless of tempo- around 100ms. According to the data provided, it appears that the length of the 2nd note only deviates from this fixed amount at slower tempos.

(note: images above from the original paper)

I'm glad to see there is finally somewhat believable scientific evidence that swing 8ths really do vary from player to player and are not absolutely triplet-based.

I do think this study could be done better, however.
  • Better Data Visualization. More data points per drummer, or per-drummer breakdowns, instead of lumping all drummers together. Yes it is visible based on the shape of the data point, but it's had to get a clear picture from it.
  • Control Cases. Comparisons with the same drummers playing straight 8ths and triplet-based feel to use an accuracy baseline.
  • Don't average lengths, but average ratios. I think it is erroneous to average out lengths of notes over the context of an entire song, since variations in tempo (i.e.: speeding up or slowing down) will skew the data. Perhaps this is already done in the original study, but it's not specified.
  • More data. I'd like to see more drummers included in this, and comparisons of relative 8th note feels in different contexts. Also, why not include other instrumentalists as well? I'd also like to see the same drummer at the same tempo, but from different recordings. Perhaps we'll have a clearer justification for why certain recordings "swing like a MOFO" more than others.
It might also be interesting to see a second study exploring the use of accents in swing feel.

[ source: Jazz Drummers' Swing Ratio in Relation to Tempo ]