On 2006-01-21, Iain Churches wrote:
"Jim Lesurf" wrote in message
...
As has recently been posted elsewhere, some 'ABX' / DBT tests *have* been
reported - *some* of which apparently show some component differences are
audible.

Indeed. As we found. But only a few participants had
sufficiently good levels of audio perception to reliably and repeatly
pick out the changes. ...

May I point out gently the trap into which Stereophile is reported to
have fallen in this respect?

It has been reported that some years ago they tested about 3,000 people
for their ability to discern whether a CD had been tweaked or not
(green felt pens came into this somewhere as I recall). Apparently,
Stereophile pointed to a small sub-set of the test subjects who had
got statistically significant scores as evidence that some people could
distinguish the tweaked CDs.

However, even with purely random guessing, some of the 3,000 will
score that well from just statistics. So the Stereophile conclusion
was incorrect. They should have re-tested the top scorers and only
if they scored well again should their conclusion have been drawn.
They might alternatively have looked at the size of the sub-set and
checked if it too was statistically significant, but a re-test would
have been more convincing.

--
John Phillips