Sample rate and the Nyquist limit
Sample rate determines the highest frequency your recording can capture. The Nyquist theorem states that the highest frequency you can accurately record is half the sample rate. At 44.1kHz (CD quality), the highest frequency is 22.05kHz, which exceeds human hearing (top out around 20kHz) by a comfortable margin. At 96kHz, the highest frequency is 48kHz, providing a 4x safety margin and allowing less aggressive anti-aliasing filters in the converter.
In practice, 44.1kHz is sufficient for final delivery (streaming, CD). Many professionals record at 48kHz (film/video standard) or 96kHz for the headroom and smoother anti-aliasing. Higher sample rates add no audible difference in the final mix if sample-rate conversion is done correctly at the end. The real benefit of higher sample rates is engineering margin during production, not final user experience.
Bit depth, quantization, and dither
Bit depth determines the resolution of each sample. 16-bit allows 65,536 discrete amplitude levels, with the smallest step (1 LSB) being approximately 0.0015dB. This resolution is sufficient that quantization noise (the error from rounding to discrete levels) sits below the noise floor of most analog gear. 24-bit (16 million levels) provides much finer resolution, useful for recording in lower-level tracking and for extensive processing chains where accumulated rounding errors could accumulate.
Dither (adding small random noise to the audio before quantizing to a lower bit depth) is essential when exporting from 24-bit to 16-bit. Without dither, quiet signals quantize to silence, causing digital artifacts. With proper dither, quiet signals remain audible. Most DAWs handle dithering automatically during export, but understanding the concept helps explain why a 16-bit mix with dither can sound better than a 24-bit mix without it, and why 24-bit production is standard despite 16-bit being adequate for distribution.