That's the ridiculous. Subjectivity has nothing to do with it, either a certain human being can perceive a difference or they cannot
Subjective means specific to one particular subject (i.e. one listener), as opposed to something which is true for all subjects. Whether there is a definite fact about what the subject perceives is irrelevant â€” it's the "certain human being" part that makes it subjective, not that subject's certainty about the content of his own experience (that should never be in question). That said, you address this in your next point...
That's why ABX tests are always done by more than one person. It's a basic scientific principal - results from any experiment must be repeatable.
If five people with undamaged hearing and good knowledge of how to listen to high-end audio can't differentiate two sources in an ABX test, you can be 99.9% sure that no-one can.
I will quibble and say that a sample size of five isn't enough to draw the margin of error you desire, but I think we basically agree. I see no reason why the ABX process necessarily requires multiple samples, which is why I have been arguing against it, but, yes, in principle ABXing could be used to draw (somewhat) objective conclusions if it was used as part of a large scale study. But, note that the objectivity comes from the sample size, not the ABX test in itself. You could obtain similar results without ABXing if a certain percentage of your test subjects were able to identify a similar difference between two sources (you would need to take steps to avoid samples "contaminating" each other with the power of suggestion).
I was suggesting that it would be worth doing an ABX test to determine if there is any difference between pure bit-perfect TrueHD and down-sampled TrueHD from a Radeon 4800. To be honest that is probably beyond the scope of an SPCR review... It's more like something AV Forums should do, but probably won't.
Yes, well beyond, since this is essentially a test of whether people can hear the difference between 48 and 96 kHz audio. SPCR definitely doesn't have the resources to run a large scale survey like this. I don't see how AVS forum could either, since there's no practical way to standardize a playback system (and space) over the internet.
That said, I think the point is already proven. I've seen / read about enough recording engineers who can point out differences between the two to believe that there is an audible difference, at least in a studio setting. (In other words, I have seen a big enough sample size to draw at least a minimally objective conclusion.) However, I wouldn't generalize this outside of the studio â€” studio quality playback systems (including room acoustics) are simply too rare in the rest of the world. I believe that most people could be trained to distinguish 48 vs. 96 kHz audio in a studio, but I know that outside of these tightly controlled conditions, the difference disappears quickly.
You're incorrect about the 96 khz claim. It's far from conclusive that people can actually regularly hear the difference at normal listening levels. For example the JAES study http://www.aes.org/e-lib/browse.cfm?elib=14195
was rather controversial but it showed most people can't tell the difference between CD quality, and higher word length and sampling rate audio and this included audio engineers. There's also some related information on Hydrogen audio http://www.hydrogenaudio.org/forums/ind ... opic=40134
. Now it may be true that people can be trained or you can come up with specific samples in specific conditions but that's generally not of great interest for the real world.
As for your claim "you could obtain similar results without ABXing if a certain percentage of your test subjects were able to identify a similar difference between two sources (you would need to take steps to avoid samples "contaminating" each other with the power of suggestion)." either I don't understand what your are saying or you don't understand what ABX is. ABX is basically a double blind test.
The reason why double blind is so important in the scientific context is because we are only interested in differences that actually exist (this isn't a very good description but it's good enough for the context). In a medical context, we want to know if a drug is actually better then no drug or a previous drug. We don't want to be confused because people think the drug is making them better (or worse) aka the placebo effect or their doctors think likewise.
The same in the audio context. The fact that people think a 96khz sample should be better then a 48khz sample and there consistently say it's different is of no interest to us. (Or the fact the person administrating the test does and therefore promotes the 96 khz sample more heavily).
Or if we're comparing two brands of products, the fact that the product branded HIGH QUALITY MADE IN USE is consistently said to be different then LOW QUALITY MADE IN CHINA by people when they know the brands is of no interest to us.
If you think asking people to identify the difference is going to be enough you're mistaken. It's quite easily possible that people will come up with the same nonsense about why something sounds different particularly since audio is an area most people probably lack any real ability to describe differences other then with mostly meaningless phrases even more so since it's easily possible the person administering the test will end up helping the person describe the differences when they are trying to work out what the person is actually trying to say.
Ultimately of course, the key thing is it makes little sense to go through all these risks and potential confounding factors, when an ABX test is simple and based on a basic scientific principle and can be analysed statisticly in a relatively straighforward and well understood manner.
What we want to know is, can people actually hear a difference? The fact that they just claim there's a difference when they believe there should or the people administering the test (inadvertently perhaps) lead them to believe there should be is of no interest to us.
Of course you're right that fact that people can hear a difference doesn't mean one is better.
P.S. Of course you don't actually need a double blind test. You could do an AB test for example. But ABX generally allows you to get results with a similar confidence interval with a smaller sample size