Keith Hughes wrote:
Mark DeBellis wrote:
Keith Hughes wrote:

I understand that you are *asserting* that to be the case. To be true,
one has to accept that a subject can be able to enjoy A more than B
without being able to make a conclusion about same/difference. I believe
that I have describe, in detail, why I find this totally implausible.

As far as I can see, you find it implausible not because that situation
or parallel ones couldn't occur in real life, but because you would
reject such an outcome as the result of a poorly constructed test.
Isn't that a circular argument?

An hypothesis is typically an explanation of observed phenomena, or a
set of assumptions to be tested by observation. An hypothesis must also,
to have any rigor, be falsifiable. Your 'hypothesis' is found wanting
against these criteria.

Sure the hypothesis is falsifiable. It would be falsified if a subject
rated things differently, but didn't discriminate them in the
same/different judgment test. But you would simply reject such an
outcome as the result of something other than a "properly constructed"
test. It's that rejection that runs afoul of the falsifiablity
criterion, not anything in my hypothesis.


1. Karl is presented with some form of probe signal, "A", followed by a
second, "B". He's asked to rate them for enjoyment.

2. Karl enjoys A significantly more than he enjoys B. Karl *must* know,
at that point, that A and B are *not* the same. This is simple logic; if
AB = 1 (true), then A=B = 0 (false). This is the most basic form of a
truth table.


Possibly, Karl knows this only because he sees that he rated A a "7"
and B a "6."

He doesn't *see* it, he *does* it. That's what the test is. I notice
you also use "6" and "7" presumably to imply small, subtle differences.
Another defect in this type of test - resolution. You need a very large
sample size, under controlled conditions, from each subject, to estimate
the precision of the test.

He does not necessarily, at the end of B, make a direct
mental comparison of the amounts of enjoyment. His knowledge that he
enjoys one more than the other depends upon his coding his enjoyment as
a rating, and comparing the ratings.

Well, you never tire of trying this contextual obfuscation do you? The
test *requires* him to make these comparisons, yet you keep wanting to
talk about what "might" be if it weren't a test, and extrapolate that to
support the idea that the same mechanism exists within the context of
the test. There *is* no "what if he doesn't compare" within the context
of a test for which comparison is the basis.

Not an accurate representation of what I say.



3. 1. Karl is presented with probe signal A, followed by B. He's asked
to determine if they are the same, or different. Barring stipulation of
intrusive test restrictions (re. listening times, manners, order of
presentation, time dista presentations, etc. - which *NO ONE* has
*ever* suggested here to my knowledge), Karl will evaluate the two,


Using the procedure of "rating them for enjoyment" as described above,
yes?

Yes, as a matter of course.


and will, as he has already demonstrated to be the case, enjoy A more than
B. Once again, the logic is irrefutable. If he enjoys one more than the
other, he *knows* they are not the same.


Yes, because the rating system enables him to know *that* he enjoys A
more than he enjoys B. But what are the chances that someone will
actually go about determining whether A and B are the same or different
in this way? Not high;

They don't need to "go about" comparing that way, its part of everyday
perception.

Introspecting about how much you enjoyed something and assigning it a
number is not ordinarily involved in everyday perception.

Do you only enjoy something if you're being asked to rate
it? Do you need a pencil and paper to remember what you like better?

Given that the procedure of introspecting and coming up with a rating
leads to a certain result, whence comes the assurance that making a
same/different judgment in a way that does *not* involve
introspecting-and-rating will give the same result?

It's more likely that they would compare the sounds and not think
about, or encode and keep track of the levels of, their enjoyment. But
unless Karl follows the procedure you outline,

It is not a *procedure*, it is a recognition of normal perception. It is
your proposition that enjoyment can *ONLY* be a factor in a ratings
test, which is patently ridiculous.

I don't make that claim. Moreover, there is a psychological difference
between (1) enjoying something and (2) introspecting about the
enjoyment and representing it on a scale from 1 to 10. A person's
discriminatory abilities and behavior, when he generates, manipulates,
and compares such representations, may or may not be identical to his
abilities and behavior in a situation in which he does not represent
his enjoyment in this way.


there is no assurance
that he will reliably discriminate between A and B even if he rates
them differently in the first test you describe.

There is every assurance that statistically valid discrimination in the
first test will result in the ability to discriminate in a
same/difference test.

This is what I have been asking. Could you please be more specific as
to what this assurance is?

But really, what follows from what you are saying is that, if we're
interested in the question of the listener's enjoyment (and why
shouldn't we be?), everybody, or nearly everybody, is doing the ABX
tests *wrong*. What we should do is to follow exactly the procedure
you outline: "repeat *EXACTLY* what he did above. He say's to himself,
'Hmmmm, I like A better than B, logic dictates that they must be
different'" (Date: 26 Jul 2005 18:10:52 GMT, Message-ID:
). In other words, we should do a
ratings test in which we encode our enjoyment on a scale from 1 to 10,
and then compare the ratings.


You are trying hard to find ways to
compartmentalize perception such that there can be no overlap in the
perceptual mechanisms involved in various forms of audio presentation.
Where's the evidence that such is the case?

The best response modern science can muster here is to demand from me
evidence that they *won't* be the same? Surely there is evidence that
they *will* be the same! So what is that evidence? (If there is no
evidence either way, then that will be an answer too.)


I am not saying that
there isn't *some* way of conducting a same/different judgment test
that would be equally sensitive as a ratings test (you have told us
what it is); I'm suggesting that a normally enacted same/different
judgment test might, for all we know, not be as sensitive, and so far I
haven't seen why that's wrong.

The same argument is just as valid for assuming that we're descended
from space aliens. Just a matter of degree. Hence the falsifiability
requirement for a true hypothesis.

The hypothesis that the audio tests will have the same outcome meets
the falsifiability requirement, since it is a straightforward empirical
claim. Either the results would be the same, in any particular case,
or they would not.

snip

However, responding to A
one way and responding to B in a different way is not equivalent to
judging that A and B are different.

Of course it is.


Well, no, it isn't. The two things should not be identified with one
another.

We may as well make this an aural example. You and I both have perfect
pitch. On Tuesday morning you hear a pitch and say, "It's D." On
Wednesday morning you hear a pitch and say, "It's E." You are asked,
was it the same pitch both days? You say "No."

On Tuesday morning I hear a pitch and say, "It's D." On Wednesday
morning I hear a pitch and say, "It's E." I am asked, was it the same
pitch both days? I say, "I don't know because I don't remember what I
heard yesterday."

Again, and analogy based on a supposition that has no place in any well
designed test. As we've discussed at length.

There is a psychological difference between us, one that is reflected
pretty clearly in our behavior. The main difference is that you can,
and do, judge that the pitches on the two days were different, whereas
I am unable to make such a judgment.

You clearly do not understand what a test is, and concomitant data
analysis methodology. "I don't know, I can't remember yesterday" is a
non-response. It is a clear indication of either a flawed methodolgy, or
an unsuitable subject.


Or that something is going on that the test is not suited to measure.


Whereas the inability to discriminate a
difference in a time-proximate test presentation, *is* indicative of
actual perceptual performance.

Why is the failure to judge, in the situation in which I do not
remember what pitch I heard, any less real (or "actual") as an instance
of perceptual performance?


However, it is true to say of each of us that we responded to A one way
and responded to B in a different way.

And, of course, reminding you that yesterday you identified the sound as
"D" would clearly allow you to make a determination of "different".

Why should that be allowed, since it is giving the subject more
information than he possesses at the time?

Hence the analogy (the original Blue analogy) fails, since there is an
internal construct against which you've already made an identification,
and you *MUST* forget for you not to be able to make a distinction. Do
you fail to see how contrived this is?

It's contrived only if the limitations of memory are irrelevant to
audio, a claim for which you have given no real argument, and seems
doubtful in the extreme.



Therefore responding to A one way and responding to B in a different
way is not equivalent to judging that A and B are different, since the
former is something I do but the latter is not.

If you ignore context, and impose an implausible range of constraints.

I'm not imposing any range of constraints here, just noting that two
things can't be identical if there's a situation in which you have one
but don't have the other. If I understand you, you are saying that if
we limit the context in certain ways, i.e. to "properly constructed"
tests, then such cases won't arise. True, but then it is you who are
imposing constraints.


snip


OK. In general, how do you create a "properly constructed, time
proximate test" that forces a reliable same/different evaluation, if
what we are measuring, in the ratings test, is the subject's perception
of a temporally extended property of the signal?

You presuppose that such exists.

There are innumerable examples of properties of temporally extended
sounds (that are not simply properties of short snippets of those
sounds).


As I've stated to Mr. Lavo on occasion,
you need only do a level matched, blind, test. Do it *any* way you want,
subject to the listed constraints, and do a sufficient number of trials
to get a statistically valid result. If you cannot differentiate, you
have no observation on which to form an hypothesis. To say "well, all my
listening says they sound the same, and analysis of the engineering
behind the articles/recordings/etc. indicate that relative to the
established thresholds of human hearing, they should sound the same, but
maybe they don't...I better speculate on what could be causing these
results", hardly even rises to the level of speculation.

Comparing longer
excerpts will be prey to problems of memory.

Yes, that's why it isn't done by serious people.

Suppose then we force
comparison of corresponding short portions of the signal in a
time-proximate way. How do we know that this will be sensitive to all
differences to which the ratings test is sensitive?

How do we know it is sensitive to all the difference that we imagine
might be there? That we have never observed? That there's absolutely no
basis for supposing?

There is ample basis for supposing that we perceive properties of
temporally extended sounds (i.e., properties of musical passages).

So the question, "How do we know that there will be a difference
between what I perceive in A and what I perceive in B (where A and B
are temporally extended) *only* if there is a detectable difference in
some corresponding short snippets of A and B?" is a perfectly
intelligible one.

What I *think* you are assuming here is basically a supervenience
thesis: no difference in perceptible properties of the entire passage
without a detectable difference in the short snippets. How do we know
that that's true?


The assumption that the person can recognize the identity between what
is contained in one perception and what is contained in another is a
substantive psychological assumption. People don't remember everything
they perceive, and they aren't logically omniscient. And not every
context is one of "properly constructed" testing. There is a lot of
idealization going on in your model of human abilities and behavior.

No, my "model" is based on objectively verifiable phenomena, and
behavior.

Though you suggest that when a subject's actual behavior violates your
assumptions, he or she is "unsuitable."

Your "model" is based on the assumption that if we are not
omniscient, we know nothing for certain. That if we do not fully and
completely understand every nuance of how a perception is generated,
that applying statistical interpretations of the objective results of
perception are of no use, because they cannot *prove* that there are,
and can be, no other interpretations.

My actual view is that we know many things, but very few of them, if
any, for certain. And if a certain methodology is unsuited to give us
information about a certain class of possible phenomena, then we should
not conclude, from the absence of data, very much either way.


Fine for philosophical ponderings,
inimical to observational science.

Sorry, but I fail to see any value to either of us in continuing this
discussion. And, as I have too many protocols to write, and reams of
data to analyze, I'll bid you adieu with this post.

Thank you, I have enjoyed and learned from the discussion.

Mark

On 7 Aug 2005 16:57:18 GMT, wrote:

Stewart Pinkerton wrote:

Why does that mathematician's comment at the HE2005 "debate" spring to
mind.....his comment that in his classes at MIT the engineers were always
separating themselves from the mathematicians and physicists by "jumping to
the end point" and failing to examine the underlying premises"?

Indeed - that's why engineers get things done, while the
mathematicians are still worrying about the origin of that strange
fourth-order term. The engineer simply shrugs and adds another 3%
safety margin.

It's also why they build things from time to time that break or don't
work. Demands of the practical world force compromise on folks with
deadlines and budgets. I just don't understand the blanket denial of
this slop built into real world applications of theoretical ideals.

Actually, it's more often stuff designed by 'pure' scientists that
doesn't work, or breaks - because they failed to leave an adequate
safety margin, or forgot something really basic. More often, however,
it just never gets built at all......

No one is denying the use of approximations. These are not 'slop',
they are *practical* values that allow rapid progress without
agonising over unimportant details. Note also that compromise is the
very essence of engineering, as there are *always* mutually exclusive
requirements in any major project. This is nothing to do with
timescales or budgets, which have their own deleterious effects,
usually exacerbated by management - see NASA for a prime example of
what happens when you let managers run engineering projects.

You should be grateful to engineers - it's because of them that
mid-price hi-fi is now as good as the very best 'cost no object' gear
was twenty years ago. There's an old saying that an engineer is
someone who can build for a dollar what any fool can build for ten.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 7 Aug 2005 20:50:32 GMT, wrote:

Steven Sullivan wrote:
The two ton gorilla in the room that any philosopehr who asserts that
facts are merely social constructs is that stuff based on scientific reasoning
*works*. It can't be mere chance that the technology developed by
science and engineering actually does what it's intended to do.

It obviously isn't mere chance but it obviously is an on going
proccess.

Indeed, theories change when new evidence accumulates to disprove the
old ones.

DBTs *work*.

No,all else being equal they work better. They don't *always* work
though.

Sure they do, they just don't always give the answer you wanted.

And they don't always give the answer that's *desired*.

I agree. Interesting that some objectivists tend to reject such tests
out of hand while not applying any standards of rigor to reported tests
that give desired answers. It's not real science when that happens.

Excuse me? Who on the 'objectivist' side is rejecting DBT results? As
far us we appear to be concerned there is no *desired* answer, only
the actual one.

Certainly scientists' pet theories don't always turn out to be true, either.
So what would it matter
to the endless self-correcting nature of science , if observations weren't
ever 'theory neutral'? The ones that are *wrong* will find themselves
at odds with the rest of the data eventually. That hasn't happened with
DBTs.

Really? Every dbt in the history of science has yielded correct data? I
think not.

Please stop erecting strawmen. If a DBT produces answers which are
wrong, this will be discovered by other DBTs which disagree with the
wrong one, forcing a closer examination of the original test.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

One more example if you will (and I apologize, too late, for the
profusion of them). Could someone please tell me if the following is
possible.

You have two signals, A and B, each constant. Signal A contains a
sound that masks another sound. Signal B is indistinguishable, on
short-snippet comparison, from signal A. But because masking is going
on in signal A, its intensity is greater than that of B. Therefore A
causes auditory fatigue (temporary threshold shift, or TTS) at a
different rate than B does. Therefore, the perceptual effects of A
will differ from those of B. After you hear A for three minutes, a
given sound may be inaudible, whereas the same sound would be audible
after you hear B for three minutes. But if you only compare short
excerpts in the test, A and B will sound identical to you.

Does that make sense, or have I stated something implausible along the
way? Thanks in advance.

Mark

Stewart Pinkerton wrote:
On 7 Aug 2005 20:50:32 GMT, wrote:

Steven Sullivan wrote:
The two ton gorilla in the room that any philosopehr who asserts that
facts are merely social constructs is that stuff based on scientific reasoning
*works*. It can't be mere chance that the technology developed by
science and engineering actually does what it's intended to do.

It obviously isn't mere chance but it obviously is an on going
proccess.

Indeed, theories change when new evidence accumulates to disprove the
old ones.

DBTs *work*.

No,all else being equal they work better. They don't *always* work
though.

Sure they do,


Absolutely wrong. people make mistakes, things go wrong. Anyone with
any knowledge of scientific research should know this.

they just don't always give the answer you wanted.


Indeed.




And they don't always give the answer that's *desired*.

I agree. Interesting that some objectivists tend to reject such tests
out of hand while not applying any standards of rigor to reported tests
that give desired answers. It's not real science when that happens.

Excuse me?


Why?


Who on the 'objectivist' side is rejecting DBT results?


Tom Nousaine. I would think you would remember this since it was your
tests of SS amps he rejected.


As
far us we appear to be concerned there is no *desired* answer, only
the actual one.

Certainly scientists' pet theories don't always turn out to be true, either.
So what would it matter
to the endless self-correcting nature of science , if observations weren't
ever 'theory neutral'? The ones that are *wrong* will find themselves
at odds with the rest of the data eventually. That hasn't happened with
DBTs.

Really? Every dbt in the history of science has yielded correct data? I
think not.

Please stop erecting strawmen.


What straw man? Maybe you should stop the rhetoric. Calling a
legitimate point a straw man is just business as usual for you. Maybe
you should think about stopping yourself.


If a DBT produces answers which are
wrong, this will be discovered by other DBTs which disagree with the
wrong one, forcing a closer examination of the original test.

Yes the data can be checked. That's the way good science works. Why you
see this as a straw man is rather bewildering. I'd think someone who
knows this would be bothered by the claim that DBTs have in effct never
yielded bad data. That clearly is a bogus claim.




Scott Wheeler

wrote:
wrote:
Stewart Pinkerton wrote:

Why does that mathematician's comment at the HE2005 "debate" spring to
mind.....his comment that in his classes at MIT the engineers were always
separating themselves from the mathematicians and physicists by "jumping to
the end point" and failing to examine the underlying premises"?

Indeed - that's why engineers get things done, while the
mathematicians are still worrying about the origin of that strange
fourth-order term. The engineer simply shrugs and adds another 3%
safety margin.


It's also why they build things from time to time that break or don't
work. Demands of the practical world force compromise on folks with
deadlines and budgets. I just don't understand the blanket denial of
this slop built into real world applications of theoretical ideals.

I don't understand the insinuation that blanket denial is what usually
happens.


Why should you? No insinuation was made. I made no comment whatso ever
as to the frequency of this sort of denial.


But I do understand the likely upcoming denial that you are suggesting such. ;-)



I understand you accurtately predicting that I would deny making the
insinuation you just invented. Anyone can make something up and then
accurately predict others will call it as such. ;-)




Scott Wheeler

On 9 Aug 2005 00:01:27 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 20:35:44 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 15:19:45 GMT, "Mark DeBellis" wrote:

Theory is always required.

No, *facts* are required, *observations* are required, and
experimental *evidence* is required. Once you have those, you can then
hypothesise, and *test* your hypothesis against the experimental
results.

OK. Clearly you take a different view from that of many writers in the
philosophy of science, who argue that there is no such thing as
"theory-neutral" observation, e.g., N. R. Hanson and W. V. O. Quine. I
myself find that view pretty persuasive.

I have no overt interest in the philosophy of Science, only in its
practical implementation.

Fair enough, but the question of how theory and observation are related
is, after all, a central issue in the philosophy of science.

OTOH, does any scientist actually *care* about all this armchair
philosophising by non-scientists? Does it have any relevance to actual
Science?
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 9 Aug 2005 00:03:16 GMT, "Mark DeBellis" wrote:

Steven Sullivan wrote:
Stewart Pinkerton wrote:
On 6 Aug 2005 20:35:44 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 15:19:45 GMT, "Mark DeBellis" wrote:

Theory is always required.

No, *facts* are required, *observations* are required, and
experimental *evidence* is required. Once you have those, you can then
hypothesise, and *test* your hypothesis against the experimental
results.

OK. Clearly you take a different view from that of many writers in the
philosophy of science, who argue that there is no such thing as
"theory-neutral" observation, e.g., N. R. Hanson and W. V. O. Quine. I
myself find that view pretty persuasive.

I have no overt interest in the philosophy of Science, only in its
practical implementation.

The two ton gorilla in the room that any philosopehr who asserts that
facts are merely social constructs is that stuff based on scientific reasoning
*works*. It can't be mere chance that the technology developed by
science and engineering actually does what it's intended to do.

DBTs *work*.

OK, so if we want to measure the perception of signals that occupy a
span of time, like a musical passage, i.e., temporally extended sounds
or sequences of sounds, you are saying DBTs work. In particular,
quick-switch same/different judgment tests work. And they are better
the shorter the snippets are that are compared, yes?

Generally, yes.

What does "work" mean here?

By now, you should not have to ask that question. Work' means that
they have proven to be the most effective method for differentiating
small byr *real*, acoustic differences which cannot be differentiated
by other means (aside from measuring, of course).

How do we know that they work?

Many decades of experiment.

Why do they work?#

Bottom line, who cares?

But basically, it's because audio memory is short, and any form of
sighted comparison is totally swamped by expectation bias.

Suppose I think that "DBTs work" means this: for any property x, if I
perceive passage A as having property x, but do not perceive passage B
as having property x, then I will be able to differentiate A and B in
the quick-switch same/different judgment test. Is there some
scientific explanation of why DBTs work (assuming it is true)? Or is
it just a brute fact that they work, not something that can be
explained?

See above.

I mean, I think it's kind of remarkable that they work, if they do.

Why would you think that? Given that audio memory is shown to be
short, and that expectation bias is easily demonstrated, why would
they *not* work?

Things could have been otherwise.

But they are not. I do wish that you'ld pack in this endless 'but
suppose' posturing, when all these suppositions are in *fact* without
substance.


In other words, I don't see why they *should* work. Maybe that's the
kind of explanation I'm asking for.

Your failure to understand something that has been explained to you at
least a dozen times, is not a failure on the part of the test.

About the idea that "facts are merely social constructs," I don't think
that philosophers who say that observations are "theory-laden" are
necessarily saying that. Most philosophers are pretty sensible.

Evidence, please? :-)
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 9 Aug 2005 00:03:52 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 20:35:21 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 15:20:28 GMT, "Mark DeBellis" wrote:

Suppose he does reliably report the presence of x in that way. Do we
know that if we conduct a different test, a quick-switch test where
Karl has to say if the two sources sound the same or different to him,
that he will reliably diferentiate them? If so, how do we know that?

How do we know? We *test* him, of course, using the alternative
protocol. Have you forgotten that this is the entire premise of the
thread?

OK, I am asking whether we have some theory that predicts that if Karl
reports a difference then the quick-switch test will also show a
difference.

We don't have 'a theory', we have hundreds of DBTs performed every day
by major audio manufacturers. Over many decades of research by
psychoacousticians, quick-switched short-term tests have *proven* to
be the most sensitive for the differentiation of small, but *real*,
acoustic differences.


OK, good point, thanks.


To be precise, ABChr is currently considered to be the ultimate DBT,
and is the basic model used by those developing advanced lossy
compression codecs such as AAC and MP3.

If Karl is unable to do this, then ipso facto x does *not* exist in
the physical soundfield,

Why should we think that?

Because it is the logical conclusion.

What are the premises from which it follows as a matter of logic? I
mean deductively.

If he does not sense it, it is *by definition* not audible, and he
cannot report it. That he might not *report* a difference that he did
sense, would be perverse.

In the above, x is a property, not a difference.

Of *course* it's a difference, if it's not possessed by B! Sheesh!

The question is
whether if he perceives the property x he would be able to report the
presence of x, not whether if he perceives a difference he would be
able to report a difference. (The perception of a difference is not
the same as a difference between perceptions.)

That is utter rubbish. If he perceives the presence of x in A but not
in B, then *of course* he detects a difference.

In case there was some ambiguity in my earlier statement, there are of
course many *measurable* differences in the physical soundfield which
are below audibility.

How do we know that if Karl perceives a
certain property of the signal, then he will be able to report that he
perceives that property?

Because if he could not report it, he did not perceive it. I believe a
'duuh' should be inserted around here somewhere.

What would you say about cognitive psychologists' theories of pitch
height and chroma (say), which they say listeners perceive; very few
people, if any, could *report* what they perceive in these terms. The
psychologists' route to knowing about them is far more circuitous than
through reliance on direct report.

What psychologists say people perceive, tends to drift with time.....

Well, if what they say implies that people are not always able to
report what they perceive, and you say people are always able to report
what they perceive, I'm going to believe the psychologists.

You seem to be misinterpreting what they say. That someone may not be
able to *describe* a difference, doesn't mean that they don't report
the *existence* of that difference. I may well be able to tell A from
B without being able to pinpoint exactly what is the difference.
Largely, that's what separates the experienced audiophile, and audio
engineer, from the layman with equally acute hearing. Note however
that both will detect the difference.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 9 Aug 2005 00:06:13 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 20:35:21 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 15:20:28 GMT, "Mark DeBellis" wrote:

Suppose he does reliably report the presence of x in that way. Do we
know that if we conduct a different test, a quick-switch test where
Karl has to say if the two sources sound the same or different to him,
that he will reliably diferentiate them? If so, how do we know that?

How do we know? We *test* him, of course, using the alternative
protocol. Have you forgotten that this is the entire premise of the
thread?

OK, I am asking whether we have some theory that predicts that if Karl
reports a difference then the quick-switch test will also show a
difference.

We don't have 'a theory', we have hundreds of DBTs performed every day
by major audio manufacturers. Over many decades of research by
psychoacousticians, quick-switched short-term tests have *proven* to
be the most sensitive for the differentiation of small, but *real*,
acoustic differences.

To be precise, ABChr is currently considered to be the ultimate DBT,
and is the basic model used by those developing advanced lossy
compression codecs such as AAC and MP3.

p.s. What are those DBTs compared *with*, in order to show that they
are more sensitive? Is the comparison systematic or anecdotal?

Over the years, with sighted listening, with single-blind listening,
and with non-time-proximate AB tests. In the world of professional
audio and psychoacoustics, the comparisons are of course systematic.

If there is a large body of research on this, then if you or anyone
could point me to a survey article, or a representative study say, I'd
appreciate it.

See anything by Sean Olive.

--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 9 Aug 2005 00:08:48 GMT, "Mark DeBellis" wrote:

Steven Sullivan wrote:

The two ton gorilla in the room that any philosopehr who asserts that
facts are merely social constructs is that stuff based on scientific reasoning
*works*. It can't be mere chance that the technology developed by
science and engineering actually does what it's intended to do.

DBTs *work*.

p.s. If DBTs work, then the fact that they work depends on certain
things in psychology being true, and it is interesting to think what
they must be.

Interesting to you, perhaps, but not of relevance to those who use
them as tools on a daily basis.

For instance, a short-comparison test will work, or be most sensitive,
only if a masking effect does not fade over the course of a long,
constant signal. If it did, then the perceptual effect of longer
signals would not be predicted by the comparison of short snippets.
Yes?

As with most of your arguments and suppositions, that is irrelevant,
because it is not in fact the case.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

Stewart Pinkerton wrote:
On 9 Aug 2005 00:01:27 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 20:35:44 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 15:19:45 GMT, "Mark DeBellis" wrote:

Theory is always required.

No, *facts* are required, *observations* are required, and
experimental *evidence* is required. Once you have those, you can then
hypothesise, and *test* your hypothesis against the experimental
results.

OK. Clearly you take a different view from that of many writers in the
philosophy of science, who argue that there is no such thing as
"theory-neutral" observation, e.g., N. R. Hanson and W. V. O. Quine. I
myself find that view pretty persuasive.

I have no overt interest in the philosophy of Science, only in its
practical implementation.

Fair enough, but the question of how theory and observation are related
is, after all, a central issue in the philosophy of science.

OTOH, does any scientist actually *care* about all this armchair
philosophising by non-scientists? Does it have any relevance to actual
Science?

I've met some colleagues who are aware of Quine, Feyerabend, and some
other philosophers' work science'. And plenty have at least read some
Popper, who actually had *useful* things to say about science.



--

-S
"God is an asshole!" -- Ruth Fisher, 'Six Feet Under'

wrote in message
...
Stewart Pinkerton wrote:
On 7 Aug 2005 20:50:32 GMT, wrote:


snip, to sharpen focus



If a DBT produces answers which are
wrong, this will be discovered by other DBTs which disagree with the
wrong one, forcing a closer examination of the original test.

Yes the data can be checked. That's the way good science works. Why you
see this as a straw man is rather bewildering. I'd think someone who
knows this would be bothered by the claim that DBTs have in effct never
yielded bad data. That clearly is a bogus claim.


I think he'd also be bothered by the small number of peer-reviewed published
dbt's in the audio world, since it might be many millennia until the
"corrective effect" provided enough checks and balances to assert itself.

Stewart Pinkerton wrote:
On 9 Aug 2005 00:01:27 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 20:35:44 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 15:19:45 GMT, "Mark DeBellis" wrote:

Theory is always required.

No, *facts* are required, *observations* are required, and
experimental *evidence* is required. Once you have those, you can then
hypothesise, and *test* your hypothesis against the experimental
results.

OK. Clearly you take a different view from that of many writers in the
philosophy of science, who argue that there is no such thing as
"theory-neutral" observation, e.g., N. R. Hanson and W. V. O. Quine. I
myself find that view pretty persuasive.

I have no overt interest in the philosophy of Science, only in its
practical implementation.

Fair enough, but the question of how theory and observation are related
is, after all, a central issue in the philosophy of science.

OTOH, does any scientist actually *care* about all this armchair
philosophising by non-scientists? Does it have any relevance to actual
Science?

Not the same question ... philosophy of science helps us (or ought to
help us) understand what science is doing.

Whether any given scientist cares is another question. Compare
philosophy of language. We all use language, refer to things, etc.
Philosophers of language ask what a language *is*, what it is for a
word to refer to something, what it is for words to have the same or
different meaning, etc. Not every language user, obviously, is
interested in those philosophical aspects. The language user engages
in the activity while the philosopher is asking what is the nature of
that activity, to understand what it is.

Philosophers of science, or good ones, may often be non-scientists, but
do know a lot about it.

Also, thanks for the reference to Sean Olive.

Mark

Stewart Pinkerton wrote:
On 9 Aug 2005 00:03:52 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 20:35:21 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 15:20:28 GMT, "Mark DeBellis" wrote:

Suppose he does reliably report the presence of x in that way. Do we
know that if we conduct a different test, a quick-switch test where
Karl has to say if the two sources sound the same or different to him,
that he will reliably diferentiate them? If so, how do we know that?

...

If he does not sense it, it is *by definition* not audible, and he
cannot report it. That he might not *report* a difference that he did
sense, would be perverse.

In the above, x is a property, not a difference.

Of *course* it's a difference, if it's not possessed by B! Sheesh!

A difference is a relation, not a property belonging simply to A or B.
"Sensing" a difference is not the same thing as sensing a quality or
property (like pitch, color, loudness, etc.).


The question is
whether if he perceives the property x he would be able to report the
presence of x, not whether if he perceives a difference he would be
able to report a difference. (The perception of a difference is not
the same as a difference between perceptions.)

That is utter rubbish. If he perceives the presence of x in A but not
in B, then *of course* he detects a difference.

No, that is why time-distal testing is unreliable.


Well, if what they say implies that people are not always able to
report what they perceive, and you say people are always able to report
what they perceive, I'm going to believe the psychologists.

You seem to be misinterpreting what they say. That someone may not be
able to *describe* a difference, doesn't mean that they don't report
the *existence* of that difference. I may well be able to tell A from
B without being able to pinpoint exactly what is the difference.
Largely, that's what separates the experienced audiophile, and audio
engineer, from the layman with equally acute hearing. Note however
that both will detect the difference.

By "perceive" I meant perceiving a property, not a difference, so maybe
we are just talking about different things here.

Mark

Stewart Pinkerton wrote:
On 9 Aug 2005 00:08:48 GMT, "Mark DeBellis" wrote:

Steven Sullivan wrote:

The two ton gorilla in the room that any philosopehr who asserts that
facts are merely social constructs is that stuff based on scientific reasoning
*works*. It can't be mere chance that the technology developed by
science and engineering actually does what it's intended to do.

DBTs *work*.

p.s. If DBTs work, then the fact that they work depends on certain
things in psychology being true, and it is interesting to think what
they must be.

Interesting to you, perhaps, but not of relevance to those who use
them as tools on a daily basis.

For instance, a short-comparison test will work, or be most sensitive,
only if a masking effect does not fade over the course of a long,
constant signal. If it did, then the perceptual effect of longer
signals would not be predicted by the comparison of short snippets.
Yes?

As with most of your arguments and suppositions, that is irrelevant,
because it is not in fact the case.

Granted.

But do you see what I'm getting at, though?

Why is there any necessity that stimuli that persist over time, and
that are in short corresponding portions indiscriminable, should have
identical perceptual effects when heard over longer time spans, since
small effects can accumulate over time?

If there is never an actual case of this in hearing, then either it's
pure luck, or it's due to some very interesting property of the
auditory system.

I bet it isn't true of other senses. I bet there are light sources A
and B that look alike when examined short-term, but where one causes
sunburn (or vision loss, or some perceptual effect) and the other
doesn't, over longer time spans. Why should this type of disparity
*never* occur in the case of hearing? That small effects can build up
over time is so pervasive and familiar a phenomenon across other
sensory modalities, it seems like it would be a miracle if it never
occurred in hearing. Hearing would have to be *unique* in this
respect. (It is a familiar part of daily life that things that differ
only microscopically, and hence imperceptibly, can differ in their
cumulative effects.) At any rate, I would be interested to know why,
if something like this never occurs in hearing, it *isn't* a miracle,
but a consequence of some known property of the auditory system.
(Concerning "short audio memory," I ask about this in another post,
which see, please.)

Anyway, what about my suggestion that signals that are indiscriminable
in short corresponding portions can sometimes induce TTS (temporary
threshold shift, or auditory fatigue) at different rates? Correct or
incorrect? Thanks.

Mark

[1] By "point" I don't mean interval of zero length of course, I mean
short interval.

Stewart Pinkerton wrote:
On 9 Aug 2005 00:03:16 GMT, "Mark DeBellis" wrote:


What does "work" mean here?

By now, you should not have to ask that question. Work' means that
they have proven to be the most effective method for differentiating
small byr *real*, acoustic differences which cannot be differentiated
by other means (aside from measuring, of course).

It is not obvious that the goal of differentiating real acoustic
differences is equivalent to the goal of modeling auditory perception.



How do we know that they work?

Many decades of experiment.

Why do they work?#

Bottom line, who cares?

We ought to.


But basically, it's because audio memory is short...

This is interesting. How short? The reason I ask is that I suspect
there is more than one kind of aural memory at work. For instance
there is a kind of musical memory, in which we make sense of phrases
and periods, antecedent/consequent relationships, and so on. That sort
of musical memory is relevant to perception, although perhaps in a
different way than is the audio memory you refer to.

Mark

Mark DeBellis wrote:
One more example if you will (and I apologize, too late, for the
profusion of them). Could someone please tell me if the following is
possible.

You have two signals, A and B, each constant. Signal A contains a
sound that masks another sound. Signal B is indistinguishable, on
short-snippet comparison, from signal A. But because masking is going
on in signal A, its intensity is greater than that of B. Therefore A
causes auditory fatigue (temporary threshold shift, or TTS) at a
different rate than B does. Therefore, the perceptual effects of A
will differ from those of B. After you hear A for three minutes, a
given sound may be inaudible, whereas the same sound would be audible
after you hear B for three minutes. But if you only compare short
excerpts in the test, A and B will sound identical to you.

Does that make sense, or have I stated something implausible along the
way? Thanks in advance.

Mark

I think you should read up on two things:

1.) masking. This is the phenomenon, that *louder* levels in a frequency
band "swamp" low levels in adjactent bands, so these details can not be
perceived in presence of the louder sound. This is independent of time.
Please note that the masked sounds must be much smaller in level than the
masking sound. Thus the masked sounds do not contribute to the overall
level, the difference will be so small if at all even measurable. Since the
frequency is really close those close tones will not sound pleasureable,
rather dissonant if audible.
We can explain that with the basilar membrane being stimulated in different
areas, on one side the deep tones are perceived, on the further side the
treble. When a section of the membrane resonates with a high amplitude, the
nearby parts will still move a bit and thus small levels in these areas can
*not* be perceived by the auditory nerve.
2.) TTS. We will not encounter such high levels which would cause a TTS
during a listening test. We will also perceive the too high of a level as
painful, certainly nobody will accept to participate or continue with the
test.

I can see that you throw in arguments without understanding their
implications, please stay with what you know and experienced yourself!
--
ciao Ban
Bordighera, Italy

Stewart Pinkerton wrote:
On 9 Aug 2005 00:03:16 GMT, "Mark DeBellis" wrote:

In other words, I don't see why they *should* work. Maybe that's the
kind of explanation I'm asking for.

Your failure to understand something that has been explained to you at
least a dozen times, is not a failure on the part of the test.

What's been stated over and over is *that* they work.


About the idea that "facts are merely social constructs," I don't think
that philosophers who say that observations are "theory-laden" are
necessarily saying that. Most philosophers are pretty sensible.

Evidence, please? :-)

Working on it.

Mark

Stewart Pinkerton wrote:
On 9 Aug 2005 00:08:48 GMT, "Mark DeBellis" wrote:

Steven Sullivan wrote:

The two ton gorilla in the room that any philosopehr who asserts that
facts are merely social constructs is that stuff based on scientific reasoning
*works*. It can't be mere chance that the technology developed by
science and engineering actually does what it's intended to do.

DBTs *work*.

p.s. If DBTs work, then the fact that they work depends on certain
things in psychology being true, and it is interesting to think what
they must be.

Interesting to you, perhaps, but not of relevance to those who use
them as tools on a daily basis.

For instance, a short-comparison test will work, or be most sensitive,
only if a masking effect does not fade over the course of a long,
constant signal. If it did, then the perceptual effect of longer
signals would not be predicted by the comparison of short snippets.
Yes?

As with most of your arguments and suppositions, that is irrelevant,
because it is not in fact the case.

I mean, not to quibble, but doesn't it at all confer understanding to
see that a given test works because the auditory system is a certain
way, and that if things were different in a certain respect then the
test wouldn't work? That it depends on certain assumptions being true?
And a contrary-to-fact example can't be helpful in illustrating that?
(Just to clarify, I didn't know when I wrote it whether it was contrary
to fact or not.) It is contrary to fact, but not irrelevant to
understanding, at least not in my case.

When you say the question of what psychological facts the efficacy of
the tests depends on is "not of relevance to those who use" the tests,
well, you would know, of course, and I would not, but I'm surprised at
the idea (which I take you to be expressing) that it is either possible
or desirable to engage in some technical endeavor without
understanding, or being interested in, why it works.

Mark

Harry Lavo wrote:
wrote in message
...
Stewart Pinkerton wrote:
On 7 Aug 2005 20:50:32 GMT, wrote:


snip, to sharpen focus



If a DBT produces answers which are
wrong, this will be discovered by other DBTs which disagree with the
wrong one, forcing a closer examination of the original test.

Yes the data can be checked. That's the way good science works. Why you
see this as a straw man is rather bewildering. I'd think someone who
knows this would be bothered by the claim that DBTs have in effct never
yielded bad data. That clearly is a bogus claim.


I think he'd also be bothered by the small number of peer-reviewed published
dbt's in the audio world, since it might be many millennia until the
"corrective effect" provided enough checks and balances to assert itself.

The task of 'correction' generally falls to those who for whatever reason
are most unconvinced by the current models. So, *where* are the
data from subjectivists?



--

-S
"God is an asshole!" -- Ruth Fisher, 'Six Feet Under'

On 10 Aug 2005 03:02:51 GMT, "Harry Lavo" wrote:

wrote in message
...
Stewart Pinkerton wrote:
On 7 Aug 2005 20:50:32 GMT, wrote:

snip, to sharpen focus

If a DBT produces answers which are
wrong, this will be discovered by other DBTs which disagree with the
wrong one, forcing a closer examination of the original test.

Yes the data can be checked. That's the way good science works. Why you
see this as a straw man is rather bewildering. I'd think someone who
knows this would be bothered by the claim that DBTs have in effct never
yielded bad data. That clearly is a bogus claim.

Good job no one made such a claim, then. That he, and you, would
translate my claim that DBTs always work, into a self-serving strawman
that no DBT has ever yielded bad data, is utterly typical. Of course
it's possible to conduct such a test badly - that does not change the
value of the *test*, it simply indicates incompetence in that
particular example.

I think he'd also be bothered by the small number of peer-reviewed published
dbt's in the audio world, since it might be many millennia until the
"corrective effect" provided enough checks and balances to assert itself.

As ever, you have this completely wrong, Harry. If indeed there *were*
opposing evidence, then the effectiveness of DBTs would not be so
bleedin' obvious as it currently is, and such results certainly
*would* be published in peer-reviewed journals such as the AES. Shame
that you and all your ilk have been totally unable to provide any such
evidence..................

--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 10 Aug 2005 00:20:36 GMT, wrote:

Stewart Pinkerton wrote:
On 7 Aug 2005 20:50:32 GMT, wrote:

Steven Sullivan wrote:
The two ton gorilla in the room that any philosopehr who asserts that
facts are merely social constructs is that stuff based on scientific reasoning
*works*. It can't be mere chance that the technology developed by
science and engineering actually does what it's intended to do.

It obviously isn't mere chance but it obviously is an on going
proccess.

Indeed, theories change when new evidence accumulates to disprove the
old ones.

DBTs *work*.

No,all else being equal they work better. They don't *always* work
though.

Sure they do,


Absolutely wrong. people make mistakes, things go wrong. Anyone with
any knowledge of scientific research should know this.

Indeed - but this has no bearing on the validity of the test protocol
itself. If you insist on such pedantry, as is utterly typical of you,
let me be more specific - *properly conducted* DBTs always work.

they just don't always give the answer you wanted.

Indeed.


And they don't always give the answer that's *desired*.

I agree. Interesting that some objectivists tend to reject such tests
out of hand while not applying any standards of rigor to reported tests
that give desired answers. It's not real science when that happens.

Excuse me?

Why?

Because it certainly is 'real science' to re-examine a test which
gives unexpected results. OTOH, one is not inclined to re-examine
tests which give 'mainstream' results. This would of course include
any test which showed that 'cable sound' does not exist. Now, if you
can show reliable and repeatable *evidence* for 'cable sound', that
really would be interesting!

Who on the 'objectivist' side is rejecting DBT results?

Tom Nousaine. I would think you would remember this since it was your
tests of SS amps he rejected.

He did not 'reject' my results, indeed he did the absolutely correct
thing for a skeptic - he asked for details in order that he could
replicate my test. That's how *real* science works - cold fusion,
anyone?

As
far as we appear to be concerned there is no *desired* answer, only
the actual one.

Certainly scientists' pet theories don't always turn out to be true, either.
So what would it matter
to the endless self-correcting nature of science , if observations weren't
ever 'theory neutral'? The ones that are *wrong* will find themselves
at odds with the rest of the data eventually. That hasn't happened with
DBTs.

Really? Every dbt in the history of science has yielded correct data? I
think not.

Please stop erecting strawmen.

What straw man? Maybe you should stop the rhetoric. Calling a
legitimate point a straw man is just business as usual for you. Maybe
you should think about stopping yourself.

Typical IKYABWAI. No one is suggesting that every DBT ever run was
flawless. This does not affect the validity of the test protocol, just
the competence of the tester, or the quality of the equipment being
used.

If a DBT produces answers which are
wrong, this will be discovered by other DBTs which disagree with the
wrong one, forcing a closer examination of the original test.

Yes the data can be checked. That's the way good science works. Why you
see this as a straw man is rather bewildering. I'd think someone who
knows this would be bothered by the claim that DBTs have in effct never
yielded bad data. That clearly is a bogus claim.

Good job then, that no one made such a claim. That's why it's a
strawman.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 10 Aug 2005 00:17:38 GMT, "Mark DeBellis"
wrote:

One more example if you will (and I apologize, too late, for the
profusion of them). Could someone please tell me if the following is
possible.

You have two signals, A and B, each constant. Signal A contains a
sound that masks another sound. Signal B is indistinguishable, on
short-snippet comparison, from signal A. But because masking is going
on in signal A, its intensity is greater than that of B. Therefore A
causes auditory fatigue (temporary threshold shift, or TTS) at a
different rate than B does. Therefore, the perceptual effects of A
will differ from those of B. After you hear A for three minutes, a
given sound may be inaudible, whereas the same sound would be audible
after you hear B for three minutes. But if you only compare short
excerpts in the test, A and B will sound identical to you.

Does that make sense, or have I stated something implausible along the
way? Thanks in advance.

You have stated that they are of different intensity, which
disqualifies them from the likelihood of being indistinguishable. You
also *claim* that the sound which contains a low-level masked
component 'therefore' causes auditory fatigue, but you offer no
evidence that this is true. As ever, you create an artificial scenario
and make baseless assertions about it.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

Stewart Pinkerton wrote:
On 10 Aug 2005 00:20:36 GMT, wrote:

Stewart Pinkerton wrote:
On 7 Aug 2005 20:50:32 GMT, wrote:

Steven Sullivan wrote:
The two ton gorilla in the room that any philosopehr who asserts th=
at
facts are merely social constructs is that stuff based on scientifi=
c reasoning
*works*. It can't be mere chance that the technology developed by
science and engineering actually does what it's intended to do.

It obviously isn't mere chance but it obviously is an on going
proccess.

Indeed, theories change when new evidence accumulates to disprove the
old ones.

DBTs *work*.

No,all else being equal they work better. They don't *always* work
though.

Sure they do,


Absolutely wrong. people make mistakes, things go wrong. Anyone with
any knowledge of scientific research should know this.

Indeed - but this has no bearing on the validity of the test protocol
itself.


Nice straw man. Who was talking about protocols? No one said test
protocols. We have been talking about *tests.*


If you insist on such pedantry, as is utterly typical of you,
let me be more specific - *properly conducted* DBTs always work.


Even when you are more specific you still make a false statement. Even
"properly conducted" dbts are vulnerable to error. Again, anyone with
any reasonable knowledge of science would no this. You seem to be
incapabale of escaping claims of absolutism. Very unscientific.




they just don't always give the answer you wanted.

Indeed.


And they don't always give the answer that's *desired*.

I agree. Interesting that some objectivists tend to reject such tests
out of hand while not applying any standards of rigor to reported tes=
ts
that give desired answers. It's not real science when that happens.

Excuse me?

Why?

Because it certainly is 'real science' to re-examine a test which
gives unexpected results.


What does that have to do with your bogus claim that dbts always work
and your ridiculous defense of that bogus claim?



OTOH, one is not inclined to re-examine
tests which give 'mainstream' results. This would of course include
any test which showed that 'cable sound' does not exist.


To bad there are no such peer reviewed published tests. So as it
stands, there really is no such thing as a "mainstream" result.


Now, if you
can show reliable and repeatable *evidence* for 'cable sound', that
really would be interesting!


If you can show any "evidence" that is reliable and repeatable for no
cable sound that would be very interesting.





Who on the 'objectivist' side is rejecting DBT results?

Tom Nousaine. I would think you would remember this since it was your
tests of SS amps he rejected.

He did not 'reject' my results,


Yes he did.


indeed he did the absolutely correct
thing for a skeptic - he asked for details in order that he could
replicate my test.


And then he rejected them.


That's how *real* science works


No, Toms protocols are not how real science works. funny how when
someone reports a no difference result Tom never asks for the details
in order to replicate the test. That's called a double standard. It is
terribly unscientific but quite typical of the objectivist camp in
audio.


- cold fusion,
anyone?

As
far as we appear to be concerned there is no *desired* answer, only
the actual one.

Certainly scientists' pet theories don't always turn out to be true=
, either.
So what would it matter
to the endless self-correcting nature of science , if observations =
weren't
ever 'theory neutral'? The ones that are *wrong* will find themsel=
ves
at odds with the rest of the data eventually. That hasn't happened=
with
DBTs.

Really? Every dbt in the history of science has yielded correct data?=
I
think not.

Please stop erecting strawmen.

What straw man? Maybe you should stop the rhetoric. Calling a
legitimate point a straw man is just business as usual for you. Maybe
you should think about stopping yourself.

Typical IKYABWAI. No one is suggesting that every DBT ever run was
flawless.


I suggest you work on your comunication skills. dbts always work is a
fairly absolute claim and a fairly ridiculous one for that very reason.



This does not affect the validity of the test protocol, just
the competence of the tester, or the quality of the equipment being
used.


Funny that you would claim IKYABWAI and then turn around and do it
yourself. if you an find any claims one way or another about "test
protocol" please cite it. otherwise please stop obfuscating.




If a DBT produces answers which are
wrong, this will be discovered by other DBTs which disagree with the
wrong one, forcing a closer examination of the original test.

Yes the data can be checked. That's the way good science works. Why you
see this as a straw man is rather bewildering. I'd think someone who
knows this would be bothered by the claim that DBTs have in effct never
yielded bad data. That clearly is a bogus claim.

Good job then, that no one made such a claim. That's why it's a
strawman.


Wrong unless you have another definition of work than what is commonly
used. "DBTs always work" your words repeated a few times on this
thread. Here's what the dictionary says. Main Entry: 3work
Function: verb
Inflected Form(s): worked /'w&rkt/; or wrought /'rot/; work=B7ing
Etymology: Middle English werken, worken, from Old English wyrcan; akin
to Old English weorc
transitive senses
1 : to bring to pass : EFFECT work miracles

If they *always* "bring to pass" they never fail to do so. When dbts
result in inaccurate data they don't "bring to pass" they fail in their
task. They don't always work. Period. What next? A personal definition
of "always?"




Scott Wheeler

Steven Sullivan wrote:
Harry Lavo wrote:
wrote in message
...
Stewart Pinkerton wrote:
On 7 Aug 2005 20:50:32 GMT, wrote:


snip, to sharpen focus



If a DBT produces answers which are
wrong, this will be discovered by other DBTs which disagree with the
wrong one, forcing a closer examination of the original test.

Yes the data can be checked. That's the way good science works. Why you
see this as a straw man is rather bewildering. I'd think someone who
knows this would be bothered by the claim that DBTs have in effct never
yielded bad data. That clearly is a bogus claim.


I think he'd also be bothered by the small number of peer-reviewed published
dbt's in the audio world, since it might be many millennia until the
"corrective effect" provided enough checks and balances to assert itself.

The task of 'correction' generally falls to those who for whatever reason
are most unconvinced by the current models. So, *where* are the
data from subjectivists?


In the same place as the evidence from the objectivists. Unfortunately
as things stand some objectivists engage in the very unscientific
habbit of picking and choosing their favorite results and ignoring the
rest.





Scott Wheeler

Ban wrote:
Mark DeBellis wrote:
One more example if you will (and I apologize, too late, for the
profusion of them). Could someone please tell me if the following is
possible.

You have two signals, A and B, each constant. Signal A contains a
sound that masks another sound. Signal B is indistinguishable, on
short-snippet comparison, from signal A. But because masking is going
on in signal A, its intensity is greater than that of B. Therefore A
causes auditory fatigue (temporary threshold shift, or TTS) at a
different rate than B does. Therefore, the perceptual effects of A
will differ from those of B. After you hear A for three minutes, a
given sound may be inaudible, whereas the same sound would be audible
after you hear B for three minutes. But if you only compare short
excerpts in the test, A and B will sound identical to you.

Does that make sense, or have I stated something implausible along the
way? Thanks in advance.

Mark

I think you should read up on two things:

1.) masking. This is the phenomenon, that *louder* levels in a frequency
band "swamp" low levels in adjactent bands, so these details can not be
perceived in presence of the louder sound. This is independent of time.
Please note that the masked sounds must be much smaller in level than the
masking sound. Thus the masked sounds do not contribute to the overall
level, the difference will be so small if at all even measurable. Since the
frequency is really close those close tones will not sound pleasureable,
rather dissonant if audible.
We can explain that with the basilar membrane being stimulated in different
areas, on one side the deep tones are perceived, on the further side the
treble. When a section of the membrane resonates with a high amplitude, the
nearby parts will still move a bit and thus small levels in these areas can
*not* be perceived by the auditory nerve.

Thanks, I appreciate the explanation. OK, so you are saying that when
masking occurs, the masked sound makes a negligible difference to the
intensity of the result, and hence won't make a difference to TTS.

So perhaps masking isn't the best example, but is it theoretically
possible for there to be two signals that sound alike, but which cause
TTS at different rates? What happens if you take an audible sound and
add a ton of ultrasonics? Will the result sound the same as the
original? Could it cause TTS at a greater rate than the original?

2.) TTS. We will not encounter such high levels which would cause a TTS
during a listening test. We will also perceive the too high of a level as
painful, certainly nobody will accept to participate or continue with the
test.

I guess I'm missing something then, because there is a graph showing
TTS as measurable from sources going as low as 20 dB (in Moore, An
Introduction to the Psychology of Hearing, p. 148).


I can see that you throw in arguments without understanding their
implications, please stay with what you know

If I had done that, then I wouldn't have profited from your helpful
explanation! Thanks again.

Mark

Stewart Pinkerton wrote:
On 10 Aug 2005 00:17:38 GMT, "Mark DeBellis"
wrote:

One more example if you will (and I apologize, too late, for the
profusion of them). Could someone please tell me if the following is
possible.

You have two signals, A and B, each constant. Signal A contains a
sound that masks another sound. Signal B is indistinguishable, on
short-snippet comparison, from signal A. But because masking is going
on in signal A, its intensity is greater than that of B. Therefore A
causes auditory fatigue (temporary threshold shift, or TTS) at a
different rate than B does. Therefore, the perceptual effects of A
will differ from those of B. After you hear A for three minutes, a
given sound may be inaudible, whereas the same sound would be audible
after you hear B for three minutes. But if you only compare short
excerpts in the test, A and B will sound identical to you.

Does that make sense, or have I stated something implausible along the
way? Thanks in advance.

You have stated that they are of different intensity, which
disqualifies them from the likelihood of being indistinguishable.

OK, I guess I'm missing something then. Suppose sound A masks sound B
to such an extent as to render B inaudible. In such a case, won't A+B
be indistinguishable from A alone? And won't A+B have a greater
intensity than A alone?

You
also *claim* that the sound which contains a low-level masked
component 'therefore' causes auditory fatigue,

I was saying, therefore they will cause auditory fatigue *at different
rates*. If the intensity is different, the rates of fatigue are apt be
different. Is that an artificial assumption?

but you offer no
evidence that this is true. As ever, you create an artificial scenario
and make baseless assertions about it.


Mark

p.s. I'm still a doubter. What seems to me has not been ruled out, for
all anyone has said, is the possibility that properties of sounds or
sound sequences that take up spans of time are perceived, and that
there can be differences between what is perceived in two signals in
this respect that are not discriminable in short-snippet comparison.

If this is going to be ruled out, two issues, to my mind, have not been
resolved.

(1) The issue of auditory memory. If the "audio" memory you refer to
is very short, there must be more to auditory memory than that, because
we make sense of music. So it's not clear to me that the claim that we
perceive properties of sounds or sound patterns extended over spans of
time would be defeated by a supposed limitation of memory.

(2) The conception of what it is to perceive a property is excessively
limited, it seems to me, by the supposed requirement that it be
amenable to direct test via discrimination of that property. Yes,
falsifiability is an issue, but evidence can be indirect.

Seems to me this is a classic instance of the sort that generates
skepticism and the witholding of judgment; at least for me.

Mark

Stewart Pinkerton wrote:
On 9 Aug 2005 00:06:13 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:

To be precise, ABChr is currently considered to be the ultimate DBT,
and is the basic model used by those developing advanced lossy
compression codecs such as AAC and MP3.

p.s. What are those DBTs compared *with*, in order to show that they
are more sensitive? Is the comparison systematic or anecdotal?

Over the years, with sighted listening, with single-blind listening,
and with non-time-proximate AB tests. In the world of professional
audio and psychoacoustics, the comparisons are of course systematic.


If we compare tests A, B, C, and D, and observe that A is the most
sensitive of them, isn't it going beyond the data to conclude that
there can't be some other test E, waiting right around the corner,
perhaps constructed in a different way, that would be more sensitive
than all of them?

If we have a choice between asserting that there can't be such a test
E, and remaining agnostic about it, which is less overreaching? Is it
not dogmatic to assert that there can't be such a test?

There are lots of hypotheses we can have ample reason to reject, but I
don't see how this is one of them. Suppose I go to a circus sideshow
and the performer guesses the number that an audience member has
written down. Do we have reason to reject the hypothesis that it was
mental telepathy? Yes, because not only has no one described a
plausible mechanism for it, our background theory entails that it is
extremely unlikely that there could be any such mechanism. If there
were such a mechanism, then it would involve transmission in the space
between the people, but no such transmission has been detected; and we
have excellent theoretical reasons for supposing that if there were
some such transmission then it would be detected.

In the audio case, we have statistics, but not a larger theory that
would explain (at any rate, no one here has convincingly explained) why
we shouldn't think that a more sensitive test, or other evidence, could
be waiting for us around the corner.

Mark

p.s. That's why we should care "why it works," because that tells us
what the statistics *mean*.

Stewart Pinkerton wrote:
On 10 Aug 2005 00:17:38 GMT, "Mark DeBellis"
wrote:

One more example if you will (and I apologize, too late, for the
profusion of them). Could someone please tell me if the following is
possible.

You have two signals, A and B, each constant. Signal A contains a
sound that masks another sound. Signal B is indistinguishable, on
short-snippet comparison, from signal A. But because masking is going
on in signal A, its intensity is greater than that of B. Therefore A
causes auditory fatigue (temporary threshold shift, or TTS) at a
different rate than B does. Therefore, the perceptual effects of A
will differ from those of B. After you hear A for three minutes, a
given sound may be inaudible, whereas the same sound would be audible
after you hear B for three minutes. But if you only compare short
excerpts in the test, A and B will sound identical to you.

Does that make sense, or have I stated something implausible along the
way? Thanks in advance.

You have stated that they are of different intensity, which
disqualifies them from the likelihood of being indistinguishable.

You
also *claim* that the sound which contains a low-level masked
component 'therefore' causes auditory fatigue,
but you offer no
evidence that this is true. As ever, you create an artificial scenario
and make baseless assertions about it.


p.s. So you are saying that we know, in fact, that *whenever* two
sounds cause TTS at different rates, then the sounds must be
distinguishable from one another in a short comparison?


Mark

"Stewart Pinkerton" wrote in message
...
On 10 Aug 2005 03:02:51 GMT, "Harry Lavo" wrote:


snip, not relevant to below


I think he'd also be bothered by the small number of peer-reviewed
published
dbt's in the audio world, since it might be many millennia until the
"corrective effect" provided enough checks and balances to assert itself.

As ever, you have this completely wrong, Harry. If indeed there *were*
opposing evidence, then the effectiveness of DBTs would not be so
bleedin' obvious as it currently is, and such results certainly
*would* be published in peer-reviewed journals such as the AES. Shame
that you and all your ilk have been totally unable to provide any such
evidence..................


Notice the sidestep. Tests purporting to be "evidence" not published, or
made open and transparent, because they are "obvious". Accordingly no
response by "real" scientists able to critique test. Somehow that is the
fault of "me and my ilk", right Stewart?

On 10 Aug 2005 23:57:50 GMT, "Mark DeBellis"
wrote:

Stewart Pinkerton wrote:
On 9 Aug 2005 00:03:52 GMT, "Mark DeBellis" wrote:

Stewart Pinkerton wrote:
On 6 Aug 2005 20:35:21 GMT, "Mark DeBellis" wrote:

If he does not sense it, it is *by definition* not audible, and he
cannot report it. That he might not *report* a difference that he did
sense, would be perverse.

In the above, x is a property, not a difference.

Of *course* it's a difference, if it's not possessed by B! Sheesh!

A difference is a relation, not a property belonging simply to A or B.
"Sensing" a difference is not the same thing as sensing a quality or
property (like pitch, color, loudness, etc.).

This is not philosophy, it is mere sophistry. If Karl senses that A
possesses x, and that B does not possess x, then he can tell that they
are different. QED.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 11 Aug 2005 00:05:17 GMT, "Mark DeBellis"
wrote:

Stewart Pinkerton wrote:
On 9 Aug 2005 00:08:48 GMT, "Mark DeBellis" wrote:

Steven Sullivan wrote:

The two ton gorilla in the room that any philosopehr who asserts that
facts are merely social constructs is that stuff based on scientific reasoning
*works*. It can't be mere chance that the technology developed by
science and engineering actually does what it's intended to do.

DBTs *work*.

p.s. If DBTs work, then the fact that they work depends on certain
things in psychology being true, and it is interesting to think what
they must be.

Interesting to you, perhaps, but not of relevance to those who use
them as tools on a daily basis.

For instance, a short-comparison test will work, or be most sensitive,
only if a masking effect does not fade over the course of a long,
constant signal. If it did, then the perceptual effect of longer
signals would not be predicted by the comparison of short snippets.
Yes?

As with most of your arguments and suppositions, that is irrelevant,
because it is not in fact the case.

Granted.

But do you see what I'm getting at, though?

Why is there any necessity that stimuli that persist over time, and
that are in short corresponding portions indiscriminable, should have
identical perceptual effects when heard over longer time spans, since
small effects can accumulate over time?

Can they? What evidence do you have for this?

If there is never an actual case of this in hearing, then either it's
pure luck, or it's due to some very interesting property of the
auditory system.

Not really, as it doesn't actually apply to any of other perceptions.

I bet it isn't true of other senses. I bet there are light sources A
and B that look alike when examined short-term, but where one causes
sunburn (or vision loss, or some perceptual effect) and the other
doesn't, over longer time spans.

That is not a difference in *perception*, merely in their
physiological effect.

Why should this type of disparity
*never* occur in the case of hearing?

You have demonstrated no such disparity in other senses.

That small effects can build up
over time is so pervasive and familiar a phenomenon across other
sensory modalities, it seems like it would be a miracle if it never
occurred in hearing.

No, that is just another of your baseless claims. Show *evidence* that
such an effect exists for other senses, or withdraw.

Hearing would have to be *unique* in this
respect. (It is a familiar part of daily life that things that differ
only microscopically, and hence imperceptibly, can differ in their
cumulative effects.)

Such as? You have a habit of making wild assertions about audio, based
on some equally wild assertion about something else.

And we're not talking about the cumulative effects of say heavy metal
poisoning here, but of *perception*.

At any rate, I would be interested to know why,
if something like this never occurs in hearing, it *isn't* a miracle,
but a consequence of some known property of the auditory system.
(Concerning "short audio memory," I ask about this in another post,
which see, please.)

So why are you asking if it isn't a miracle, when you already have the
answer?

Anyway, what about my suggestion that signals that are indiscriminable
in short corresponding portions can sometimes induce TTS (temporary
threshold shift, or auditory fatigue) at different rates? Correct or
incorrect? Thanks.

Certainly, there's no evidence in support of such a theory.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 11 Aug 2005 00:05:50 GMT, "Mark DeBellis"
wrote:

Stewart Pinkerton wrote:
On 9 Aug 2005 00:03:16 GMT, "Mark DeBellis" wrote:


What does "work" mean here?

By now, you should not have to ask that question. Work' means that
they have proven to be the most effective method for differentiating
small byr *real*, acoustic differences which cannot be differentiated
by other means (aside from measuring, of course).

It is not obvious that the goal of differentiating real acoustic
differences is equivalent to the goal of modeling auditory perception.

Irrelevant to the question.

How do we know that they work?

Many decades of experiment.

Why do they work?#

Bottom line, who cares?

We ought to.

Who's 'we'? Most drivers do not understand the Otto cycle, but their
cars still work, and that is what is important.

But basically, it's because audio memory is short...

This is interesting. How short?

Seems to be only a few seconds. Best ask the psychoacousticians -
that's what they *do*, basically. Bottom line - quick-switched DBTs
*work* better than time-distal presentations.

The reason I ask is that I suspect
there is more than one kind of aural memory at work. For instance
there is a kind of musical memory, in which we make sense of phrases
and periods, antecedent/consequent relationships, and so on. That sort
of musical memory is relevant to perception, although perhaps in a
different way than is the audio memory you refer to.

Why do you suspect such a thing? What *evidence* do you have that it
exists?
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 11 Aug 2005 00:07:33 GMT, "Mark DeBellis"
wrote:

Stewart Pinkerton wrote:
On 9 Aug 2005 00:03:16 GMT, "Mark DeBellis" wrote:

In other words, I don't see why they *should* work. Maybe that's the
kind of explanation I'm asking for.

Your failure to understand something that has been explained to you at
least a dozen times, is not a failure on the part of the test.

What's been stated over and over is *that* they work.

And yet, you seem to be scrabbling around for some strange theory that
would run counter to this. Why?

About the idea that "facts are merely social constructs," I don't think
that philosophers who say that observations are "theory-laden" are
necessarily saying that. Most philosophers are pretty sensible.

Evidence, please? :-)

Working on it.

Actually, despite being told on many occasions to go read up on
psychoacoustics, you appear *not* to be working on it. That is not
sensible for someone who claims to seek a deeper understanding of why
DBTs work.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 11 Aug 2005 00:09:20 GMT, "Mark DeBellis"
wrote:

Stewart Pinkerton wrote:
On 9 Aug 2005 00:08:48 GMT, "Mark DeBellis" wrote:

Steven Sullivan wrote:

The two ton gorilla in the room that any philosopehr who asserts that
facts are merely social constructs is that stuff based on scientific reasoning
*works*. It can't be mere chance that the technology developed by
science and engineering actually does what it's intended to do.

DBTs *work*.

p.s. If DBTs work, then the fact that they work depends on certain
things in psychology being true, and it is interesting to think what
they must be.

Interesting to you, perhaps, but not of relevance to those who use
them as tools on a daily basis.

For instance, a short-comparison test will work, or be most sensitive,
only if a masking effect does not fade over the course of a long,
constant signal. If it did, then the perceptual effect of longer
signals would not be predicted by the comparison of short snippets.
Yes?

As with most of your arguments and suppositions, that is irrelevant,
because it is not in fact the case.

I mean, not to quibble, but doesn't it at all confer understanding to
see that a given test works because the auditory system is a certain
way, and that if things were different in a certain respect then the
test wouldn't work? That it depends on certain assumptions being true?

It does *not* depend on assumptions, it depends on what *is*. *You*
are the one who wants to assume things that are *not* the case.

And a contrary-to-fact example can't be helpful in illustrating that?
(Just to clarify, I didn't know when I wrote it whether it was contrary
to fact or not.) It is contrary to fact, but not irrelevant to
understanding, at least not in my case.

Your understanding would be greatly improved if you read up on
psychoacoustics, rather than just throwing out non-factual
suppositions.

When you say the question of what psychological facts the efficacy of
the tests depends on is "not of relevance to those who use" the tests,
well, you would know, of course, and I would not, but I'm surprised at
the idea (which I take you to be expressing) that it is either possible
or desirable to engage in some technical endeavor without
understanding, or being interested in, why it works.

Really? You think that Michael Schumacher understands, or cares about,
fluid dynamics or materials science?
--

Stewart Pinkerton | Music is Art - Audio is Engineering

Mark DeBellis wrote:

1.) masking. This is the phenomenon, that *louder* levels in a
frequency band "swamp" low levels in adjactent bands, so these
details can not be perceived in presence of the louder sound. This
is independent of time. Please note that the masked sounds must be
much smaller in level than the masking sound. Thus the masked sounds
do not contribute to the overall level, the difference will be so
small if at all even measurable. Since the frequency is really close
those close tones will not sound pleasureable, rather dissonant if
audible.
We can explain that with the basilar membrane being stimulated in
different areas, on one side the deep tones are perceived, on the
further side the treble. When a section of the membrane resonates
with a high amplitude, the nearby parts will still move a bit and
thus small levels in these areas can *not* be perceived by the
auditory nerve.

Thanks, I appreciate the explanation. OK, so you are saying that when
masking occurs, the masked sound makes a negligible difference to the
intensity of the result, and hence won't make a difference to TTS.

So perhaps masking isn't the best example, but is it theoretically
possible for there to be two signals that sound alike, but which cause
TTS at different rates? What happens if you take an audible sound and
add a ton of ultrasonics? Will the result sound the same as the
original? Could it cause TTS at a greater rate than the original?

2.) TTS. We will not encounter such high levels which would cause a
TTS during a listening test. We will also perceive the too high of a
level as painful, certainly nobody will accept to participate or
continue with the test.

I guess I'm missing something then, because there is a graph showing
TTS as measurable from sources going as low as 20 dB (in Moore, An
Introduction to the Psychology of Hearing, p. 148).



FWIK this is not true. With 75dB exposure over several hours test showed
around 6dB TTS. I can not perceive of anybody having that with 20dB SPL.
Maybe that 20dB was not an absolut level? I don't have that book but at that
low level a TTS seems improbable.
What *is* important is the kind of stimulus used. With a pleasant musical
material the TTS turns out to be much less than some technical sounds like
nasty squeaking noise or (filtered) white noise of the same level.

--
ciao Ban
Bordighera, Italy

In article ,
Stewart Pinkerton wrote:

On 11 Aug 2005 00:05:50 GMT, "Mark DeBellis"
wrote:

Stewart Pinkerton wrote:
On 9 Aug 2005 00:03:16 GMT, "Mark DeBellis" wrote:


What does "work" mean here?

By now, you should not have to ask that question. Work' means that
they have proven to be the most effective method for differentiating
small byr *real*, acoustic differences which cannot be differentiated
by other means (aside from measuring, of course).

It is not obvious that the goal of differentiating real acoustic
differences is equivalent to the goal of modeling auditory perception.

Irrelevant to the question.

How do we know that they work?

Many decades of experiment.

Why do they work?#

Bottom line, who cares?

We ought to.

Who's 'we'? Most drivers do not understand the Otto cycle, but their
cars still work, and that is what is important.

But basically, it's because audio memory is short...

This is interesting. How short?

Seems to be only a few seconds. Best ask the psychoacousticians -
that's what they *do*, basically. Bottom line - quick-switched DBTs
*work* better than time-distal presentations.

The reason I ask is that I suspect
there is more than one kind of aural memory at work. For instance
there is a kind of musical memory, in which we make sense of phrases
and periods, antecedent/consequent relationships, and so on. That sort
of musical memory is relevant to perception, although perhaps in a
different way than is the audio memory you refer to.

Why do you suspect such a thing? What *evidence* do you have that it
exists?

There is an interesting discussion of such things in Howard Gardner's
"Frames of Mind". It has been some time since I've read it, but I'll
review it this weekend and report here.

Mark DeBellis wrote:

The reason I ask is that I suspect
there is more than one kind of aural memory at work. For instance
there is a kind of musical memory, in which we make sense of phrases
and periods, antecedent/consequent relationships, and so on. That sort
of musical memory is relevant to perception, although perhaps in a
different way than is the audio memory you refer to.



Umm...didn't this get addressed in this very same thread months ago?
See my post #348, from June 24th @ 10:29am. Yes, musical memory and
audio memory are totally different things.