[Harry Lavo]

"Scott" wrote in message
...
On Apr 1, 4:40 pm, "Harry Lavo" wrote:
"Andrew Haley" wrote in message

...

Scott wrote:
On Apr 1, 4:59?am, Andrew Haley
wrote:

snip

So it is not unfair much less grossly unfair to make this
charcterization when Arny pulls out the science flag. It's only
better to have more people doing "science" so long as they are doing
it up to the standards set by the scientific community.

There, I agree totally. What matters is how well the experiment is
done. But it's a matter of degree: some experimental controls are
surely better than none, even if the experiment isn't perfect.

Not necesarily. If the controls that aren't there are crucial to the
validity of the test, or the design of the test itself is not valid
(stimulus, measurements, intervals, training, intervening technology,
etc.)

If the controls "aren't there" then you have "none" by definition.

No, then the controls are inadequate. There is a difference. Sometimes
"inadequate" controls can slip by the designer, as can validity-destroying
intervening variables. That's why careful peer review is important.


Conventional ABX'ng has never been shown to be valid in evaluating MUSIC
differences that other approaches (the aforementioned Oohashi test) and
even
the ABC/hr test have proven better at. Yet ABX is the test that Arny
developed a computerized version of, and has relied on.


If the construct of the test itself intereferes with the normal
evaluative
process, you can almost be guaranteed that it will not produce valid
results. One of the principles of testing in any field of human endeavor
is
to try to emulate as much as possible the conventional context of the
variable under test.

How does ABX interfere in a way that ABC/hr does not? Neither
methodology is particularly more or less like the "normal evaluative
process" if there is such a singular thing. I can't go there with you
Harry. If done well ABX should do the trick. Sure any given ABX test
may miss an audible difference that is present and not specifically
being listened for. But I have to side with the DBT advocates that
when used to test claims of audibility those making the claims should
already know what specifically to listen for. ABX doen right does not
make audible differences go away. I think "done right" is the issue
not ABX per se.

I don't like either, although ABC/hr takes a timid step in the direction of
musical evaluation.