The most convincing thing I've read is a paper in which a 96khz or 192khz recording was double blind A/B'd against the same signal through a studio grade ADDA chain at 44.1 khz and no group could consistently identify them or pick a preference. The test included both professional engineers and a number of students from a respected recording engineering program. I'd guess the students were included for their hearing ability in the upper registers.
Maybe it should have been musicians vs. engineers.