Absolutely true (other than the "understand everything" hyperbole). Of course, there's that annoying "evidence" thing...
It is hard to understand that one can talk about validation and refuses at the same time to accept that any test has to be validated too.
Controlled listening tests are part of the scientific tool box and the rules are strict; a test has to be objective, valid and reliable.
An ABX uses a standard protocol but relies on a human detector (aka listener) and some sort of switching device (if used for amplifiers for example).
Without using positive and negative controls it is simply impossible to fulfill the scientific requirements (means to show that it is valid).
Without that the result of any test is only "null hypothesis could not be rejected" or "null hypothesis could be rejected" but no further conclusions about the reasons for these results can be drawn, which means that the test is quite useless.
Controlled listening tests are part of the scientific tool box and the rules are strict; a test has to be objective, valid and reliable.
An ABX uses a standard protocol but relies on a human detector (aka listener) and some sort of switching device (if used for amplifiers for example).
Without using positive and negative controls it is simply impossible to fulfill the scientific requirements (means to show that it is valid).
Without that the result of any test is only "null hypothesis could not be rejected" or "null hypothesis could be rejected" but no further conclusions about the reasons for these results can be drawn, which means that the test is quite useless.
Last edited:
First sentence: Yes, where appropriate. It depends on the question being asked and the variable examined. Otherwise, this is just a mantra.
Second sentence: The first part is correct (and appropriate to the question, "Can listener X hear the difference between two preamps in a setup where he claims the differences are readily audible?"). Either he can (null hypothesis rejected) or he can't (null hypothesis not rejected). The second part is a nonsequitur- it's very useful for answering questions of this type.
Being specific about the question asked by an experiment is a necessary first step to prevent confusion of the sort you're having.
We are talking about audible differences between two DUTs and we are talking about the test protocols normally used.
One other possible method, where the participants did not know about being tested, i have described before.
Btw, scientific rules are indeed sort of a mantra and for an ABX there exists no other way to show the validity wrt to our discussions.
Of course if an experimenter is researching the hypothesis that a difference is audible under just any condition by every listener, than a positive control is not needed (but a negative control is still mandatory).
I´m sorry but our concern isn´t normally if a specific listener can detect something, but if an audible difference between two DUTs exists and that is a very different question.
That was the reason why i´ve wrote a lot of posts about operationalisation in test design to bring (back) that concept to (your) attention and we know that in the past (in most cases) a definite question wasn´t established before testing for amplifier and other differences.
It does not help (well of course that depends on the experimenters bias) to adjust the question afterwards to meet the results.
Seems perfectly reasonable to me. One clear question.
Now I'm confused.
Jakob has pointed out one question.
Could you be more specific regarding the different questions that are mashed together?
I don´t know if it were different questions in every case, but if you´re asking if a listener can detect a difference in a test (under the premise that he claims to hear a difference under different conditions), then you still would have to show that the independent variable in your test will be only an audible difference but not the disability to detect the difference under test conditions.
It ist only appropriate if you have shown first the validity of the test, otherwise the corrected question would be "can listener X hear the difference between two preamps _under_ _test_ _conditions_ in a setup where he claims the difference are readily audibel?")
It is an answer to these questions, _if_ the test is objective, valid and reliable.
Otherwise you can´t draw any conclusions and the result would just be the "null hypothesis could...." result as descripted.
Last edited:
This isn't my question, this should be the question asked by someone who claims sonic superiority of their product or technology. My question was the one you dismissed.
This isn't my question, this should be the question asked by someone who claims sonic superiority of their product or technology. My question was the one you dismissed.
Ahh, not dismissing something, just sorting out, due to a given uncertainty.
So, did "the" listener claim to hear a difference under the specific blind test conditions?
The listener claims he can hear differences under sighted conditions, unrelated to frequency response and level, with his choice of material and with his system. Like every other listener who has made such a claim, he has yet to demonstrate he can do it by ear alone. Like most others who feel their reputation is at stake, he does not want to test his hypothesis.
Really, there's no need to obfuscate a very simple question.
Really, there's no need to obfuscate a very simple question.
I am afraid, but different test conditions compared to the "claim conditions" require validation of the test. Positive controls are a must have; training sessions are preferrable too.
Listeners do not want to test their (listening)hypothesis, and experimenters obviously do not want to validate their tests.
I totally agree, their is no need for obfuscation; positive/negative controls are mandatory for this sort of test.
if you don't "know your enemy"
..their real theory, intellectual framework, supporting experiments then you aren't participating in a dialog - just political/religious rants trading sound bites, waving the bloody shirt to rally your base
why make statements in ignorance of the literature - someone is sure to point to links disproving your assertion
discriminating very small ~ 0.1 dB level difference - not generally identifiable as a perceived loudness difference - does emerge from the statistics - even if the participants can't articulate why they are identifying A vs B
ABX Amplitude vs. Frequency Matching Criteria
distortion with differing profiles, way below 1% were even ranked, correlated with a simple formula in Geddes work on the GedLee Metric
Perception
psychoacoustic lossy CODECs, "MP3" and better ones are totally DBT driven in their development
MP3'Tech - www.mp3-tech.org
Hydrogenaudio Forums - Portal ran public audio CODEC tests (largely abandoned for higher bit rates - due to lack of significant results as they have improved)
JAES articles do require a uni library or AES membership for cheap access, JASA has more of the stuff that becomes psychoacoustic textbook material
..their real theory, intellectual framework, supporting experiments then you aren't participating in a dialog - just political/religious rants trading sound bites, waving the bloody shirt to rally your base
why make statements in ignorance of the literature - someone is sure to point to links disproving your assertion
discriminating very small ~ 0.1 dB level difference - not generally identifiable as a perceived loudness difference - does emerge from the statistics - even if the participants can't articulate why they are identifying A vs B
ABX Amplitude vs. Frequency Matching Criteria
distortion with differing profiles, way below 1% were even ranked, correlated with a simple formula in Geddes work on the GedLee Metric
Perception
psychoacoustic lossy CODECs, "MP3" and better ones are totally DBT driven in their development
MP3'Tech - www.mp3-tech.org
Hydrogenaudio Forums - Portal ran public audio CODEC tests (largely abandoned for higher bit rates - due to lack of significant results as they have improved)
JAES articles do require a uni library or AES membership for cheap access, JASA has more of the stuff that becomes psychoacoustic textbook material
Last edited:
I guess a different question might be -- is there any other way, than dbt, to hear differences or to hear anything for certain? Which other legit ways are there? Do such ways exist now? If such other tests do exist, what do they test for?
[i've already given examples]
Then ask your self ("self") is there a single - one - test that we could do which would tell us the sum total of all the many years research on masking, harmonic thresholds, acoustics and perceptions etc etc etc? Like the meaning of life itself.... we want the simple answere that covers it all.
-RNM
[i've already given examples]
Then ask your self ("self") is there a single - one - test that we could do which would tell us the sum total of all the many years research on masking, harmonic thresholds, acoustics and perceptions etc etc etc? Like the meaning of life itself.... we want the simple answere that covers it all.
-RNM
Last edited:
After test and measurement vetting I like Rupert Neve's method "I hook it up and play music and do other things" - Passive listening - That is, IME, the only way to distract your brain from worldly things so that that audio zen moment can grab your hearing-brain connection and get a clear evaluation. Otherwise your brain is over analyzing. May not be "legit" but it is how I roll.
- Status
- Not open for further replies.