Well you have to make a bunch of assumptions about the magnitute of the difference and the chance that a random person hearing it actually detects it, but I ran lots of values and, very roughly, to have 80% power, i.e., 80% chance of detecting the difference with 95% certainty that it really IS a difference, you need about 15 subjects.
Allocating 30 minutes to each test, we can do this in one full one day at a meet, taking 15 volunteers who feel they have "golden ears". I would love to take a top-of-the-line digital system, and play the same passage using MP3's of various bit rates, then full redbook 16/44.1, then 24/192 LPCM. This mimics a test the pros did and published. They claimed no one could hear the difference between redbook and hi-res LPCM.
I bet we can. Would love to know.
I'll bet we can, too!
Without testing there is no Scientific Method.