Just for the sake of those who read old threads ... the necessary sampling rate is determined by the bandwidth of the analog signals, and by your objectives in making the recording, since the sampling rate needs to be greater than twice the highest frequency that you wish to preserve. Since human hearing is limited to 20 kHz, a sampling rate of 44.1 kHz is adequate, while higher sampling rates are ... more than adequate.
Whether "more than adequate" translates into "sounds better" depends on the particular equipment that you're using. There's no inherent reason that a higher-than-necessary sampling rate should sound better, but sometimes it does for what we could call (to use the technical term for it) stupid reasons.
For example in the mid-to-late 1980s there was a widely-sold Sony studio DAT recorder that had separate analog filters for its three sampling rates (32, 44.1 and 48 kHz), and the filters that they happened to use at 44.1 sucked rather badly. Back then PCs weren't used in studios yet, there was no digital audio software like we have today, no sound cards that would let you digitize analog signals. This particular deck was the first piece of digital recording equipment most engineers ever got their hands on that would let them directly compare input vs. output at different sampling rates. Many engineers did make the obvious listening comparisons, and became thoroughly convinced that "48 kHz sounds a lot cleaner than 44.1 kHz" because they believed that they were hearing the difference between two sampling rates, rather than the difference between two analog filters of widely differing quality.
Turns out, if you installed good 44.1 kHz replacement filters (Apogee started out selling plug-in replacement filters for early Sony digital gear) then there was really no audible difference any more. So a lot depends on whether a particular piece of gear lives up to the potential of its medium or not. Back then a lot of gear didn't, but some people insisted on judging the CD medium or digital recording in general by those examples nonetheless. That's just a huge honking fallacy.
--Jerryfreak, please don't take this amiss but when you wrote "digital audio is an approximation of a continuous analog waveform. more numbers= better approximation. an analog signal has an essentially infinite sample rate, and a 'bit rate' could be roughly correlated to the s/n level of the source" I have to say that you were reiterating a well-known, fundamentally mistaken viewpoint. I can't claim to have a perfect safety record as a driver when in fact I don't drive; I have no safety record at all as a driver. Similarly, a continuous signal has no sampling rate at all--not an infinite one. If you had to assign it an Fs, it would be zero.
Your first sentence is spot on, though--digital recording is indeed one way of approximating a continuous waveform. In that respect it's just like analog recording, which isn't exact, either. The goal is to minimize the audible difference between input and output. Digital recordings produce continuous waveforms; the use of discrete-time sampling is an internal, technical step which should not be confused with the final result.
Once the requirement to sample at more than twice the highest frequency of interest has been met, the requirement can't be "more met" by making room for still higher frequencies, especially if they're not there; it's either met or it's not. Speaking of Mets, it's like, if you hit a home run ball, your team gets a point when you reach home safely. After that, though, running extra laps around the bases wouldn't earn your team any further points; you've already fulfilled the requirements, and you'd only be holding up the game.
--best regards