Where's the evidence that they didn't cherry-pick examples that are intentionally hard to tell apart? It's also dishonest for them not to emphasize how important speakers/amps/etc matter in the test. Of course someone with Apple earbuds or "Beats by Dre" won't be able to tell the difference.
I didn't mean to suggest that the npr article presented _the_ authoritative objective test on the topic. I was only attempting to answer the original question of "how to prove it" - in that - I think its possible to construct such a test.