MP3 and AAC are designed to keep lots on information in areas we hear best (midrange) and lose information in areas that we humans generally aren't attuned to as much - lower end bass and the high-end of the spectrum.
The more compressed, the smaller that "sweet spot" of good sounding stuff gets. The human voice can be really, really compressed with pretty good sound quality. The more dynamic the source, the less compressed it needs to be to still sound "good."
Most of the time you can notice an MP3 or AAC by muddy bass and "swishy" sounding high-end stuff. Cymbals especially seem to suffer.
It's all a lot more complicated than that, but there it is in a nutshell. I usually stick to AAC at 192 or better, and that seems to be the best compromise between size and sound quality, or at least to my ears. I still use MP3 as well sometimes if I know it's not going to be played on a mac or ipod - again, I just use 192bps.
