The other thing worth noting. Carver said his amp was nulling -39 db. One would expect that might be audible. And as JGH said, 4 out of 5 is not significant. Suppose he was just one less. 3 of 5 or one the other way 4 of 5. Many complain when this sort of objection is brought up. One can use the random number function in a spreadsheet to see how often 4 of 5 happens. I just did and in 100 trials, which done 5 at a time is 20 runs, it occurred 5 times that 4 of 5 at random turned up. That is one in four by chance. Suggestive, but simply not enough to be conclusive.
Hence why I mention nothing can be concluded either way in the discussion apart from the engineering and this goes beyond distortion/FR, and that it is very easy to go in unprepared for blind listening
Even JGH states the listening test would need more results to be conclusive, however worth remembering the initial listening (where seemed there were no differences between amps) and the follow up were casual sighted, eventually they went to blind test selection to provide greater validation that unfortunately run out of time with just 5 tests.
Most would use the Clark rule that would need 11-12/15 correct, however even this is way too simplistic because outside of AES other JND research use a sliding scale regarding %; the basis being as something becomes smaller it may still be noticable but accuracy will be much lower in dbt.
I posted several scientific papers or parts from ones I have 1-2 years ago relating to a broad spectrum and diverse range of JND, including biases-determination-weighting-etc, but it is something I am not bothered about doing again as it takes an insane amount of time to ensure context,validation and use is correct.
Anyway worth remembering that initially they (JGH and JA) did not notice differences and was only over time; whether this was because the amp tolerances-spec-performance due to the nature of its build were compromised or the fact greater length of time listening enabled a more methodical approach in terms of music-sounds used that then focused on emphasised traits we cannot say.
However key aspect is JGH mentioning using specific segment music-trait to "engineer" differentiation.
Not disagreeing btw.
Cheers
Orb