Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!rutgers!mtune!mtunb!dmt
From: dmt@mtunb.ATT.COM (Dave Tutelman)
Newsgroups: rec.audio,sci.physics,sci.electronics
Subject: Re: Mercury Filled Speaker Wire
Message-ID: <1067@mtunb.ATT.COM>
Date: Wed, 16-Sep-87 07:27:49 EDT
Article-I.D.: mtunb.1067
Posted: Wed Sep 16 07:27:49 1987
Date-Received: Sat, 19-Sep-87 00:53:14 EDT
References: <3827@watdcsu.waterloo.edu> <551@unisoft.UUCP> <3837@watdcsu.waterloo.edu>
Reply-To: dmt@mtunb.UUCP (Dave Tutelman)
Organization: AT&T Information Systems - Lincroft, NJ
Lines: 51
Summary: Double-blind tests
Xref: mnetor rec.audio:3277 sci.physics:2177 sci.electronics:1314

In article <3837@watdcsu.waterloo.edu> bmaraldo@watdcsu.waterloo.edu (Commander Brett Maraldo) writes:
>    Double-blind testing seems to be the most accepted method 
>    by those who understand the importance of accurate test results.  But
>    is double-blind testing the definitive test method?  This is certainly
>    and arguable point, as it is as objective as the golden-ear phenomenon.
	If I understand what the "golden-ear phenomeneon" is, it isn't at
	all objective.  It's totally subjective.  Maybe you could explain
	what you mean, and disabuse me of this opinion.
>
>    For me, a double-blind test must
>    be rigorous, extended, and thorough; that is, there must be a large
>    number of people involved in the evaluation, and the test procedure
>    must be overseen by at least one member of a group of potentially biased
>    individules (one audio engineer, physicist, golden ear, and stereo
>    salesman).  As each procedure is concluded, the overseeres must
>    unanimously agree that the procedure was unbiased before enumerating
>    the results.  The listening environment must differ between anechoic and
>    very 'alive', with each group listening in the different situations;
>    one listening environment does not suffice.  
>
>    ....  This same argument can be overlayed on a poorly conducted
>    double-blind test and most do not meet the requirements of an unbiased
>    controlled environment procedure.
>
	Perhaps we ought to review the DEFINITION of a double-blind test,
	as the description above carries a lot of extra baggage (useful in
	testing audio equipment, but having nothing to do with double-blind)
	and misses the fact that double-blind methodology is INHERENTLY
	unbiased (no "agreement" among the overseers is needed if
	the methodology is double-blind).

In a double-blind experiment, TWO parties are "blind":  (1) the subjects
(you knew that, of course), and (2) the "overseer" of the experiment.
The pattern of the experiment is created by a random number generator,
and administered by a tester that doesn't know which experiment is currently
going.  In a medical experiment, this means that both the patient and
the doctor are ignorant as to whether the patient is in the experimental
or control group.  In an audio experiment, this means that both the
"golden ears" and the person doing the switching are ignorant about
which equipment is being played.  There ARE techniques for assuring this.
If they are used, the test WILL be unbiased.  You can criticize other
aspects of the test (room too "live", statistically insignificant
number of subjects, tone-deaf subjects :-), but it is innocent
of bias.

+---------------------------------------------------------------+
|    Dave Tutelman						|
|    Physical - AT&T  -  Lincroft, NJ				|
|    Logical -  ...ihnp4!mtuxo!mtunb!dmt			|
|    Audible -  (201) 576 2442					|
+---------------------------------------------------------------+