There are three sides to consider for this debate:Soundmind?? said:Recently read an article by one of Apogees designers. Being an expert in digital audio and conversion, he should know what he's talking about. He pointed out that the human ear can't hear much above 18-20khz and that since 48K gives you ability to record up to 24khz
1) The Nyquist theory
The max frequency that you can sample is half the sampling rate.
But this theorem is only valid for ideal converters and filters.
Furthermore, at this max frequency, you only have two samples per complete sine waveform, which is quite not really acceptable in pro audio.
2) The inputs to most soundcards are low-passed anyways.
Look, wether you select 44.1k or 96k, you cannot physically change the input circuit of your soundcard, right ? In order to avoid undersamplig at 44.1k, sound card manufacturers have to low-pass everything below about 20.5kHz. Thus, even when using 96k, you still ain't going to get those 48kHz harmonics many people talk about. Sampling higher only enables you to have "more" fidelity at frequencies higher than about 10kHz (but still lower than 20.5kHz).
3) The higher the sampling frequency, the easier the filtering
Without going too far into sampling theory, a higher sampling frequency reduces the constraints on output filtering, which in return leads to higher fidelity.
Bottom line is that you should use the highest possible/reasonable sampling rate and get over it.
And don't use less than 24 bits for all the good obvious reasons mentionned earlier.
I can actually hear a difference between 44.1k, 48k and 88.2k. Above that, I don't hear a difference.