Anyone who honestly believes that they can hear the difference between the original CD and a 320kbits MP3 should head on over to HydrogenAudio and check out the ABX testing they do.
Long story short, pretty much nobody can hear the difference on music samples. Only those who have trained their ears (and minds) to detect the specific artifacts of different codecs can hear the difference, and even then only on a small percentage of samples.
I was skeptical, so I tried it myself using the Foobar2000 ABX plugin (for the pedants, I was using an AMB Gamma-1 DAC, a Tangentsoft Pimeta headphone amplifier and a pair of Sennheiser HD-25-1-II's.), using LAME to encode CBR MP3's. At 64kbits, the difference was blindingly obvious. At 96kbits, it was less awful, but easily detectable. At 128kbits, many samples were listenable, but not perfect. And at 160kbits, my ability to distinguish the samples from the lossless source fell below the level of statistical significance - ie. maybe I could hear something, but not consistently, and this was when I was listening for problems.
I now use 160kbit Average Bit Rate AAC's (for the Sony) or WMA's (for the car head unit) - since both these codecs are more efficient than MP3, I'm confident that I've avoided nearly all possibility of audible artifacts, without wasting storage capacity. At home, I still store everything lossless, so I can convert it to the appropriate format as required.