Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site opus.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!harpo!seismo!hao!cires!nbires!opus!rcd
From: rcd@opus.UUCP
Newsgroups: net.audio
Subject: Re: extra speakers: test methodology
Message-ID: <328@opus.UUCP>
Date: Fri, 6-Apr-84 03:51:56 EST
Article-I.D.: opus.328
Posted: Fri Apr  6 03:51:56 1984
Date-Received: Mon, 9-Apr-84 05:29:24 EST
References: <304@nbires.UUCP>
Organization: NBI, Boulder
Lines: 40

<>
Steve (nbires!winograd) omitted one important part of a test methodology.
The test ABSOLUTELY MUST be done with the help of another person to remove
and replace the extra speaker.  You (or whoever is doing the critical
listening) must not know whether the speaker is present or not.  Use a
blindfold if you need to (honest, it won't affect audio perception that
much!) but try to make the test unbiased.  If you know when the speaker is
present and when it isn't, you may not be able to be objective.

Suggestion:  Blindfold the listener.  Have the assistant go to the room
where the extra speaker is kept, and either bring the speaker out or not.
Try the test selection.  Remove the speaker or not, but the assistant must
go to the room where the speaker would be kept, and return.  The assistant
should be a skeptic of the phenomenon, if possible.  Be aware of subtle
cueing of the assistant's actions.  For example, if the extra speaker is a
substantial one in an enclosure, it may be that the assistant will make
some noises from the exertion of carrying the speaker.  (If you can arrange
for the tester to stop his ears as well while the change is/isn't being
made, so much the better.)   Remember that testing involving human beings
is the most difficult to control, because unconscious cues are often hard
to find.

One variation of the test which has been suggested to me is to compare with
the speaker always present, but sometimes with the terminals shorted and
sometimes not - this SHOULD show a difference, though not as much as
predicted for removing the speaker entirely, if I understand the arguments
right.

Use a pseudo-random sequence of changes.  (Come on, everyone out there
should know how to set this up, right?)  That is, if 1=speaker present and
0=speaker absent, don't use a sequence like 010101010101.  Instead, use
something like 0101111000101011100.  In fact, runs of "speaker present" or
"speaker absent" are good for checking whether the listener really hears a
difference and is convinced of it.
---
"Nothing left to do but smile, smile, smile."
{hao,ucbvax,allegra}!nbires!rcd
-- 
"Nothing left to do but smile, smile, smile."
{hao,ucbvax,allegra}!nbires!rcd