The link doesn't work for me, but I assume you read it: is there any control for distortion? IMD can create distortion in the audible spectrum from content above it. I don't know whether the level can be high enough to be audible, but since the main difference being examined, the sampling rate, is also considered to be below the threshold of audibility by most, it seems really important to examine whether what the subjects are hearing are just unfortunate artifacts of the chain or not.
I remember reading a study from Japan a few years ago that had zero mention of managing distortion at all.
