Bruce Olsen wrote:
>I must admit that I didn't use it, because I don't think that a
>statistical approach is needed here. [I did actually program that
>data collection into my ABC player program, but didn't have the
>patience to run through a large number of tunes in a variety of
>modes and collect the results. And are you using tune modes or
>scoring modes?] Also Phrygian and Lydian are rather uncommon (and
>Locrian useless) and there are a substantial number of
>non-'Greek' modes that are much more common (see table below).]
I played through and edited all the tunes I used to make sure
that the scoring mode and tune mode were the same.
Since I've had the algorithm working I've discovered quite a few
more Phrygian and Lydian tunes; they are uncommon, but not
quite as rare as I had thought. As for Locrian - it's unused
but not useless. Try this:
X:9
T:The Vampire's Lair
C:Phil Taylor
R:Reel
M:C|
K:D Loc
A,2|\
D2 (FD) ADFD | EFGA (3dcd AF | D2 (FD) ADFD | EFED CA,B,C |
D2 (FD) ADFD | EFGA d2 d2 | (3fed (3edc ddAF | DFEC D2 :|
(dc) | \
dcde ddAd | defd edBd | ({f}e)def eddc | defd d2Ad |
dcde ddAd | defd edBc | dcde dBAF | DFEC D2 :|
>The mode of the tune is determined by the keynote and the notes
>in the tune, and may or may not have anything to do with the
>scoring mode specified in K:key-mode. A program to find keynotes
>for such would appear to me to be more useful. [My ABC player
>program determines scoring mode from the final note and sharps or
>flats on the key signature, but final mode is the tune mode, not
>the scoring mode, and that comes from notes in the tune, and of
>course, isn't necessarily a 7 note 'Greek' mode, and in a slight
>majority of cases it is not.]
>
>Correct me if I'm wrong, but it seems to me the key is the
>'final' note of the tune (96% of those I've stressed note coded),
>except in cases of circular modes and questions only arise in
>connection with keynotes for circular modes (4% of those I've
>stresssed note coded). [This is not strictly true, but exceptions
>are rare. A variable 2nd strain can lead to different modes for
>tunes whose 1st strains are identical. e.g., in 'Sources of Irish
>Traditional Music', prototypes #456 ("And the Kirk Woud Let Me
>Be") and #738 ("Silly Old Man"), are both versions of the Scots
>'Fye let us a' to the Bridal'(#377). An obvious difficulty with
>stressed note coding, one I haven't yet solved, is that the
>keynote comes from the end of the tune, while the stressed note
>code is for the beginning of the tune.]
I think you pay too much attention to the final note. While it's
true that most tunes end on the key note, the key information is
actually distributed through the tune. As an accompanist, I
often have to accompany tunes that I've never heard before. I
can almost always determine the key before the tune has got to
the end the first time through. So while the last note is a useful
indicator, it's by no means essential to determining the key.
Circular tunes (by which I mean those that end in a phrase which
leads back to the beginning) are a bit of a red herring here. There
are also many tunes which end solidly on a stressed note which
is not the key note.
>How well does the 7 note 'Greek' mode model serve us?
>
>Mode distribution: Highest 31 modes (of 179) of the 6601 tunes
>stressed coded in file Comcode.TXT on my website.
>Letters x= a to g are the keynotes for which modes have no sharps
>or flats. In xn, note n of the scale is missing. v = variable, vm
>is variable mth note of scale (natural plus sharp or flat
>depending on scoring mode).
Fascinating statistics. However, all of the non-Greek modes you
classify there can be thought of as derived from one or more of
the seven classical modes. You even describe them that way
yourself - e.g.
> 2 cv7 8 534 ionian with 7b added
It's possible that you have lumped together two different modes
there - ionian with added 7b and mixolydian with added 7#. Those
don't necessarily come to the same thing, since tunes in the two
parent modes give different weights to notes in the different
scale positions. The algorithm I described would split that group
into two, and you would find that the two groups actually sound
different. Tunes in the mixolydian mode tend to place heavy
emphasis on the flattened seventh (it's the characteristic sound
of that mode) while tunes in the major place much less emphasis
on the more dissonant major seventh.
Even highly chromatic tunes can be thought of as derived from a
Greek mode. "Entry of the Gladiators" in my previous post uses
all twelve notes, but does not use them equally. Here's the
percentage note useage histogram (that is the percentage of time
that the tune spends on each note):
A *********************
A# ********
B **********
C *****
C# ********
D ******
D# ******
E **************
F ***
F# ********
G *****
G# ******
You can see that the key note has to be A with E as the fifth.
(Despite the fact that it both starts and ends solidly on D.)
The tune makes much more use of the major third (C#) than the
minor third (C), so it's "major" rather than "minor", which
fits with the way it sounds. The algorithm favours AMix as
the solution (based on the proportions of the above histogram)
but also gives A major as second best. I wouldn't really
argue with either.
>The 'Greek' mode model starts out with a bang on ionian, but
>after that doesn't match very well with what's observed.
>I don't, think, however, that the 'Greek' mode model is going to
>disappear in the near future; it's just too hard to figure out
>anything else. Homer got it wrong; it should be "Beware of
>Greeks bearing modes".
:-)
Phil Taylor
To subscribe/unsubscribe, point your browser to: http://www.tullochgorm.com/lists.html
