Resumption of Posts from Brian

From: mclaren
Subject: psychoacoustics versus aesthetics
--
In tuning digest 755, John Chalmers wrote:
"The JI versus ET or nj-net debate is unlikely to be resolved,
especially since Roberts and Mathews found that there are
two types of listeners, those that prefer JI and those that
prefer the richness of intervals tuned sharp or flat in triads. Frankly,
given the over-learning of 12-tet that musically trained subjects will
have been subjected to and the self-selection that musically naive
persons will have undergone for the lack of musical discrimination
abilities, I tend to be skeptical of many preference studies.
My advice is for composers and performers to use the tuning
they find best for their musical purposes."
"I suspect that their audiences will usually agree." -- John Chalmers
This advice for composers and performers seems sound
and accords with common sense. However, the rest of John's
post reflects a troubling misunderstanding of the psychoacoustic
research.
If someone as smart as John Chalmers is unclear on these
points, it's clear that explanation is required.
The study of triadic tuning preference cited by John Chalmers is
not in the same class with the rest of the pscyhoacoustic
research which I have quoted. Properly speaking, the Mathews/
Roberts studies on "pure" and "rich" listeners belong
to the subset of studies on music aesthetics. Such studies are
seldom done nowadays because they are fraught with peril:
as John Chalmers points out, it's unclear whether a test of
interval preference indicates (A) cultural brainwashing
or (B) innate proclivity for this or that tuning. This applies
to aesthetic "preference" studies of Asian, African, South
American tunings as well as to "preference" studies of
western 12-TET intervals.
I have seldom cited such aesthetic research because of
the potential problems. In particular, Carol Krumhansl's
studies seem to be deeply flawed because of their
inherently aesthetic bent--Krumhansl talks about "western
tonal hierarchies," but it's not clear whether her data are
produced brainwashing among the test subjects or
some innate supracultural tendency. (Krumhansl's attempts
to address this issue raise more problems than they solve,
inasmuch as it's clear that Krumhansl actually believes that
Indian musicians use *exactly the same pitches* as the white
notes of a Western piano. Put on *your* CD of Indian music,
and see if *you agree.*) Much of the bias could even be
generated by Krumhansl's test structure. I'd like to see
Krumhansl's probe-tone tests repeated for high-limit JI
scales and exotic high-limit probe tones.
My guess is that the results would be *radically* different.
By contrast, the studies I have cited are *not* concerned with
preference. I have not talked about "preference" since
tuning digest 28 (& if the word crept into my posts, it's the result
of my falling into bad habits--in any case I've endeavoured
strictly to avoid such prejudicial and inaccurate language).
Most psychoacoustic studies do not seek to uncover
"preferences." Modern psychoacoustic studies are concerned,
rather, to compare what the *listener perceives* with
what the *instruments measure.*
To say (as Mathews and Roberts do in one single isolated
study) that some listeners prefer triads tuned with plenty
of beats while other listeners prefer triads tuned as
beatless as possible...well, that's one thing.
But to say, as 99.99999+% of modern psychoacoustic research
does, that listeners predominantly perceive the p5 interval
as "pure" when it is sharp of the just value by an average
of 5-7 cents for the "perceived" 3/2...
That's something *entirely* different.
Such studies make no claims about preferences.
These studies do *not* say "such-and-such a tuning is
a true fifth," or "such and such a fifth sounds better
than another `fifth'." They don't even say "such-and-such
interval is preferred by listeners."
That's a complete misunderstanding of the psychocaoustic
research.
Instead, what these studies say is that there is a significant
persistent universal difference between classes of *perceived*
intervals and classes of *acoustically measured* intervals.
Such results have *nothing* to do with preferences.
When Johan Sundberg points out that "it is important not
to confuse with the *perceptual* octave with the *acoustical*
octave," this is NOT a statement that the 2:1 is "better"
or "worse" than the 1215-cent octave. That's not it at all.
That's not the point.
"Better" or "worse" cannot be applied to well-designed
psychoacoustic experiments. Such experiments seek
*only* to quantify the human sensorium--in particular,
the human ear/brain system.
A good example of how complex this is (and how completely
irrelevant and inappropriate terms like "preference" and
"better/worse" are to this process of attempting to
quantify human auditory percepts) can be found in the
controversy over the mel scale of pitch during the
early 1950s.
The idea behind the mel pitch scale was to quantify
pitch perception in the same way Fletcher & Munson had
quantified loudness perception during the 1930s
for different pitch regions.
Because Fletcher and Munson discovered that
perceived loudness was dependent *both* on
the physical amplitude of the signal *and* on
its frequency, someone thought that a similar
scale of pitch height might prove useful.
Alas, the mel scale proved largely useless
because it provided a measure of pitch which
changed drastically depending on where the
interval was in the frequency spectrum. Moreover,
the mel scale applied accurately only to sine wave
signals, and broke down when more complex acoustic
inputs were tested.
It would be a gross error and a complete
misunderstanding to say that listeners "prefer"
the mel scale for sine wave signals, or that
listeners "prefer" a quasi-logarithmic
scale for complex harmonic signals. There is
no question of "preference"--one scale of
measurement is not "better" or "worse" than
another. The mel scale is useful for quantifying
sine signals, but it is almost never used nowadays
simply because the vast majority of acoustic
phenomena which psychoacousticians seek to
measure and quantify are not sine waves signals.
"Useful for such-and-such a class of auditory
phenomena" is *not* the same as "better."
"Meaningless when applied to such-and-such a
class of auditory inputs" is *not* the same as
"worse."
It's important to be clear about these distinctions,
because there's a persistent undercurrent of
discussion on this forum which conflates studies
of musical *preference* (which might have limited
validity, depending on the very limited cultural
context one restricts oneself to) with studies
of music *perception* (which in general are
valid across cultural boundaries, and which
furthermore often tell us something basic
about the organization and operation of the human
ear/brain system).
If psychoacoustics were nothing but a set of
tests which produce "I like it/I don't like it"
answers, the science would indeed be of little
concern to members of this tuning forum.
That is not the case, however. This is a
complete misrepresentation of psycoacoustics.
Instead, psychoacoustic studies (when
done well--not all such research is competent
or adroit) bypass such subjective aesthetic
and emotional reactions and reach directly
into the details of how the human auditory
system operates.
This latter issue of some concern when dealing
with intonation, since it's vital to disentangle
*what the listener perceives* from
*the measured acoustic data.*
If we do not disentangle percepts from
prejudices, questions of intonation are apt
to reduce to the trivial level of "such-and
-such tuning is more natural," "such-and
-such tuning is purer," and so on. In short,
into meaningless wrangling over
buzz-words.
--mclaren


Received: from ns.ezh.nl [137.174.112.59] by vbv40.ezh.nl
with SMTP-OpenVMS via TCP/IP; Fri, 30 Aug 1996 01:11 +0200
Received: by ns.ezh.nl; (5.65v3.2/1.3/10May95) id AA21387; Fri, 30 Aug 1996 01:12:48 +0200
Received: from eartha.mills.edu by ns (smtpxd); id XA06971
Received: from by eartha.mills.edu via SMTP (940816.SGI.8.6.9/930416.SGI)
for id QAA18551; Thu, 29 Aug 1996 16:12:46 -0700
Date: Thu, 29 Aug 1996 16:12:46 -0700
Message-Id:
Errors-To: madole@ella.mills.edu
Reply-To: tuning@eartha.mills.edu
Originator: tuning@eartha.mills.edu
Sender: tuning@eartha.mills.edu