A weblog by Will Fitzgerald

Interestingness of parts

There has been an interesting discussion about composing on the Sacred Harp mailing list; one of the subtopics has been a perennial favorite: what makes a ‘good’ Sacred Harp tune. One of the desiderata is that parts be interesting to sing as themselves; i.e., a kind of polyphony. I’m curious how one could model ‘interestingness’ computationally. For example, a more varied tune is generally more interesting than a less varied one, so perhaps the variance of interval changes could be a simple model of interestingness. As a small test: I compared the variance of interval changes in the alto part of the arrangement of New Britian (i.e. Amazing Grace), in the 1191 Denson revision of The Sacred Harp to the version in the Christian Reformed Church Psalter Hymnal (Midi version). Sure enough, the variance in the Sacred Harp is 1.95 vs. 1.47 for the Psalter Hymnal: by this measure, the Sacred Harp version is about 1/3 ‘more interesting.’ I suspect this would hold up over all.

One comment made in the mailing list discussion is that most modern hymnbooks are written (as I think Sam Sommers put it when I talked to him about this at the Goshen monthly singing) ‘by keyboardists, for keyboardists’–that is, the music isn’t arranged for the singers so much as for the benefit of the piano or organ player. I’m interested in finding a good a capella hymn book.


4 responses to “Interestingness of parts

  1. Natalia September 6, 2005 at 2:14 am

    It’s certainly true that tenor parts, which are the “melody,” jump around a lot more than the treble or alto parts.

  2. Will September 6, 2005 at 8:06 am

    Yes — I thought that the alto part, which (as you know) was added later, would show less variance, making any difference in variance more remarkable. Still, of course, this is just a bit of exploratory investigation.

  3. Jesse September 7, 2005 at 11:39 am

    Neat idea. I wonder whether there might be other characteristics of interestingness, certain intervallic jumps, for example.

  4. Will September 7, 2005 at 12:37 pm

    No doubt! Variance is a very weak measure.

%d bloggers like this: