Hi Alan,
Thanks for reporting in more detail on those listening tests. According to Wikipedia, AAC wasn't released as a public standard until 1997, so I doubt it would have played a big role in a 1992 test. Its implementations have also been continually improved since 1997, so a test from that long ago might not show its current capabilities anyway.
Wikipedia says that 128 kbps AAC MPEG-2 is considered "transparent" for stereo; would be interesting to see if there's a believable basis for that, and also if the current MPEG-4 implementation has changed the situation. Maybe such updated codec listening tests would be a fun project for Axiom :-). You certainly have all the technology and expertise required, and it could lead to some good press since large numbers of people want to know how to encode their music as efficiently as possible without sacrificing fidelity.
For now, I think I'll stick to my Apple Lossless encoding--files are about half the size of uncompressed, though much larger than MP3 or AAC, but no worries about fidelity or the complications of determining just how to encode.
