Arny says the Dolby test is blind. If you searched for the word blind in the Dolby AES paper you find zero references. If you search for statistical analysis of whether the results are random or not, you won't find that either. If you search for ABX, you guessed it, you will be out of luck there too.
What we do find in the paper are the following regarding how the listening test was conducted:"In order to reduce testing time and the number of reversals necessary, a self-regulated up down threshold test was devised. This test shortens the feedback loop of the up-down threshold test by providing the subject with direct control over the amount of jitter added to the program source. This allows the subject to quickly identify and home in on the audible threshold. During this process, the subject was allowed to confirm his or her threshold using an AB comparison box. This box allowed the listener to compare the low-jitter bitstream with the jittered bitstream and further refine his or her threshold.
A function generator was connected to this input and used to set the jitter frequency and level throughout these tests. The jittered output of the JM-1 was connected to input 'B' of the AB comparator box, which is used to compare the two bitstreams.
Self-Administered Threshold Evaluation
[skipping the section on setting the level]
After the output level was adjusted, testing began and the subject was instructed to adjust the jitter level until their threshold of audibility was reached. During testing, the subject was free to use the AB comparator to compare the two bitstreams for audible differences. This allowed for further refinement of the threshold by providing the subject with ready access to a low-jitter reference. Once a threshold was reached the result was recorded. This test was repeated for all of the selected program material."
Note that the tester himself was in charge of changing the level of jitter until he determined the points at which he could or could not hear it. There is no way such a test then would be blind let alone double blind. The tester is changing the level up or down and with it, has that bias factor (i.e. knowledge of whether he was increasing or decreasing jitter). The comparator box was an additional aid to him, to use if he felt the need to refine his levels of detection. This is what the test harness looked like:
Note that it is a simple AB box, NOT ABX. So much for only ABX tests being acceptable in AES or industry testing in general. Even there, we see that the AB test is used sighted. The tester at all times knows which input is which on the AB box. How else would he use it to refine his detection levels?
If the above is the definition of a double blind ABX test, then we have to say the following cable test is also good:
Tester is given 10 different audio cables and a short zip cord. He is given a box where he can insert his audio cables. He puts them in one and a time until he hears a difference. We note the brand and identity of that cable. We also give him an AB test where he can confirm his answer by comparing said cable against the short zip cord. He does not have to use that box if he does not want to. It is up to him.
Let's see who will accept the result of the above cable test. Anyone? I suspect none of you will accept the results of that test. Yet Arny goes around promoting the Dolby test as "blind" and authoritative.
The bottom line here is very clear: when the tests agree with our preconceived notions, any test will do. When it doesn't, then no amount of rigor is sufficient. I don't call this objectivity. You have to have equal standards here. As soon as you shift them with the outcome, then it really doesn't matter what objections you bring to the table. It is a biased assessment worst than a biased test itself.