I paid special attention to the first 12 seconds before the singing attempt began, particularly the 9th though 12th notes, which have the sort of content which might demonstrate a difference. After several trials on each of the four samples, nothing was heard would support anything beyond a pure guess.
And as I've said, that first test was really absolutely worse case, and I couldn't tell the difference.
Did you check out the second set of tracks? (You might like the music too.) I'm glad that I had the option for 8- and 12-bit output, or I would have never believed that the dithering software I was using was even doing anything (actually, I can see it doing something on the VU meter at the higher depths, but I can't hear it).