first results of genetic perfect pitch study

http://www.pnas.org/cgi/content/abstract/0703868104v3

Not the genetic bit, just the survey data. It turns out
G# is misidentified a lot. I'm not sure I believe their
explanation.

-Carl

... In fact there isn't a genetic bit at all - what kind of genetic
study can you do over the internet?

The only possible connection with genetics is the apparent result of a
bimodal distribution. However, I can think of half a dozen other
(non-genetic) reasons why one might find such a distribution with
their survey methods. For example people with mediocre pitch
recognition might be more reluctant to take the survey than either
those with accurate AP or with none at all. Or AP might be correlated
with the number of words of Mandarin the person knows, which is also a
bimodal distribution, but not genetic! The problem is that they have
no control at all over the consistency of sampling.

What they wanted to find is the distribution of AP ability. (Over what
population exactly?...) What they have is this distribution, times a
function which encodes the desire, ability and opportunity of people
to take the online test within the survey period. Without knowing
something about this second function, the actual distribution of
ability is, to say the least, uncertain.

~~~T~~

--- In tuning@yahoogroups.com, "Carl Lumma" <clumma@...> wrote:
>
> http://www.pnas.org/cgi/content/abstract/0703868104v3
>
> Not the genetic bit, just the survey data. It turns out
> G# is misidentified a lot. I'm not sure I believe their
> explanation.
>
> -Carl
>

--- In tuning@yahoogroups.com, "Tom Dent" <stringph@...> wrote:
> ... In fact there isn't a genetic bit at all - what kind of genetic
> study can you do over the internet?

It's actually quite easy to send a cheek swab in the mail.
I don't know how they plan to do that bit -- perhaps in person
on a smaller cohort.

> The only possible connection with genetics is the apparent
> result of a bimodal distribution. However, I can think of
> half a dozen other (non-genetic) reasons why one might find
> such a distribution with their survey methods. For example
> people with mediocre pitch recognition might be more
> reluctant to take the survey than either those with accurate
> AP or with none at all. Or AP might be correlated with the
> number of words of Mandarin the person knows, which is also
> a bimodal distribution, but not genetic! The problem is that
> they have no control at all over the consistency of sampling.

I don't like interent studies either. In any case, we'll
know about the genetic bit soon enough.

I must say (if I didn't already) their explanation of the
G# anomaly was entirely unconvincing.

-Carl

--- In tuning@yahoogroups.com, "Carl Lumma" <clumma@...> wrote:
>
> perhaps in person on a smaller cohort.
>

Yes, I suspect they will select out the cream of the sample and then
actually do some genetic tests.

> their explanation of the
> G# anomaly was entirely unconvincing.

It was (like much of the rest of the press release) just vague
handwaving - but can you think of any better reason why G# should be
so often taken for A? (Or why C and D were identified with such
unusually small or infrequent errors? - which they don't mention...)

~~~T~~~

> > their explanation of the
> > G# anomaly was entirely unconvincing.
>
> It was (like much of the rest of the press release) just vague
> handwaving

I didn't read the press release, I read the paper.

> but can you think of any better reason why G# should be
> so often taken for A?

Hard to say without a closer look at their methods.
My current guess is their data is just wrong.

-Carl