If science can't measure a difference it is only that, nothing more. People might interprete this differently on their own risk, as its a field with too much vaiables to make projections based on measurements.
Such a situation only becomes intresting to me when science can't measure anything and a significant number of people with a documented good hearing can hear a difference, I don't tink that anything intresting happened yet.
Well, there are perfectly valid scientific methods for dealing with sound that are not measurements. I've done many hundreds of jury tests of sound. An experiment with a skilled clarinetist who doesn't know what he's playing, and an audience of skilled musicians who don't know what's being played, asking them to judge first "same, or different?" and then "Better, equal, or worse?" is a perfectly valid experiment. In fact, given the extreme difficulty of looking at spectra and determining which differences are significant and which are not, a listening jury like I just described, a double-blind test, is probably the best way to assess the actual impact of different constructions. (It's not truly a double blind test if the experimenter knows what's being tested; shall we call it a single and a half blind test?)
Just a little anecdote: my wife recently tried out two different violas. To make her decision, I would play each one - same sequence of notes, short and simple - while switching randomly between the two. Sometimes "first viola" and "second viola" were actually the same instrument. Sometimes A was first in the pair; sometimes B was first in the pair. If I did B or A twice, I'd actually make sure to set the instrument down and then pick it up again so there wouldn't be cues from the amount of time or extraneous noises. She was sitting across the room with her back turned. So this wasn't a double blind experiment because I knew which one I was playing. And I didn't truly randomize the sampling, just made some patterns that looked "kind of" random. But it was a heck of a lot more controlled and random than the usual kind of test where everyone knows what's being played and then they - surprise, surprise - get exactly the results they're expecting.
And the interesting thing is that for two completely different instruments - different makers, slightly different dimensions, different strings, completely different violas - even then, she sometimes was not able to tell when two successive instruments were the same or different, or to identify which was which. Where with total random chance it would be 50% misidentification, I'd say her results were something like 15-20%.
The point is that even in a controlled experiment, the ability to discriminate between things, even things with major differences, is far from 100%. No one would argue that two violas made to different dimensions by different makers wouldn't be expected to sound repeatably different. Now if you take a factor that is inherently going to be have at MOST a tiny effect compared to the variation in other factors, you can pretty much guarantee that the ability of listeners (and players, in this case) to distinguish it will be very close to that of random chance, at BEST.
And yet players and manufacturers continue to make blanket, positive assertions about these supposed effects, and they NEVER NEVER NEVER attach any data. Why is that, do you think?