lexnasa, you don't seem to know where the extra 8 bits of 24 bit go. They just tack on the bottom and define detail that is already below any meaingful noise floor (this is for delivery only, for recording the extra bits have obvious advantages in level setting on recording, but we are talking normalized at playback). Unless you think you can hear signal at -90db in your listening room, 16 and 24 bit will be identical.
"As you're ignoring sampling frequency, and only addressing bitrate, your argument is fundamentally flawed. If you imagine analogue as being a continous flow of data, and digital as a series of samples that are taken and reconstructed, which do you think would be better? And taking more detailed samples should be good? By your argument 4 bit should sound as good as 16 bit... it doesn't! 4 bit or 16 bit or 24 bit at the same sampling frequency are addressing exactly the same range of human hearing."
You don't take the range of human hearing, in volume terms, and split it into as many samples as you have. You start at pegging the meters and then each bit defines a sound half as loud. The extra 8 bits define the sounds as quite as nose-hair waving and the self-noise of the microphones. Sampling should make a difference from various secondry isses (anti-alasing filter slopes etc), but bit-depth does nothing but give you more signal-to-noise. And as 16 bit is 96db, and there is only about 70db in even the most dynamic recordings, 16 would seem to cover it.