Re: Absolute pitch discussion

[Martin Braun <nombraun@xxxxxxxxx>]

Dear Etienne and others,

> Is there any chance that the bimodal distribution found in Athos et al.
> could be explained by the way to compute scores ?

Thank you. Things like that can indeed be crucial. First, computational
biasing can never be avoided fully. Here, I think, the authors' choices have
been reasonable.

Second, the two clusters are so sharply defined and so far apart that a
possible influence of computational biasing seems negligible. You
see, between 11.75 and 24.5 the chart (Fig.1) is almost empty.

Third, the inbox in Fig. 1 strikingly demonstrates that the zero-AP cluster
precisely fits random responses. There is not the slightest indication in
the data of a possible continuum from the zero-AP cluster to the AP cluster.

Additional comment: It had been fairly clear from the early days of AP
research that this trait approaches the all-or-nothing quality. Now we have
got it in black-and-white, as good as it can be.

Martin

---------------------------------------------------------------------
Martin Braun
Neuroscience of Music
S-671 95 Klässbol
Sweden
web site: http://w1.570.telia.com/~u57011259/index.htm




----- Original Message ----- 
From: "Etienne Gaudrain" <et.gaudrain@xxxxxxx>
To: <AUDITORY@xxxxxxxxxxxxxxx>
Sent: Saturday, September 01, 2007 11:01 AM
Subject: Re: Absolute pitch discussion


Dear Martin,

Is there any chance that the bimodal distribution found in Athos et al.
could be explained by the way to compute scores ? What I understood is
that exact correct answer gave a score of 1, and an error of a semitone
gave a score of .75. But this way to compute scores does seems to me to
advantage bimodal distribution.

I'm far from beeing a specialist... but I tried to imagine what could be
the different kind of subjects that would populate the distribution :
- Very good AP subject => gives score of 36.
- Quite good AP subject. For example, knows a few absolute tones, and
has very good relative pitch. May do some mistakes, but only a few and
small errors (<1 semitone) => should give a score >36×.75=27, i.e.
considered as AP.
- Not so good AP subject. Knows fewer absolute tones, and has only quite
good relative pitch. For example, we can imagine that for intervals
greater than an octave, the error can exceed one tone. Such a subject
may probably produce a lot a errors limited to 1 tone. But since 1 tone
error count as zero... The overall score would probably be close to zero.
- Subject without AP but good relative pitch. Since the test is a
sequence of tones, the subject with good relative pitch should be able,
if he adpots such a strategy, to refine his judgement along the test. He
should then make smaller errors than the last category.
- Subject without AP and without relative pitch. Should answer
aproximatly at random.

So it seems to me that the continuum between AP expert and subject
without any pitch naming skill exists, but that the way to compute
scores advantages the apparition of a bimodal distribution. Do you think
it is possible ?

Regards,
-Etienne

--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Etienne Gaudrain
Universite Claude Bernard LYON 1
CNRS - UMR5020, Neurosciences Sensorielles, Comportement, Cognition
50, avenue Tony Garnier
69366 LYON Cedex 07
FRANCE
Tél : 04 37 28 74 85
Fax : 04 37 28 76 01
Page web equipe : http://olfac.univ-lyon1.fr/unite/equipe-02/equipe-2-f.html
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~