Open any home studio forum and the same argument runs on a loop: should you record and mix at 96kHz, or even 192kHz, instead of the old 44.1 or 48kHz? The pitch sounds reasonable. More samples per second, more data, more resolution. It feels like free quality. The honest answer is more interesting than either side usually admits. For everything you can actually hear, a higher rate buys you nothing, and in a real playback chain it can make the sound slightly worse.
What a sample rate actually controls
The Nyquist-Shannon sampling theorem sets a hard ceiling. A signal with no energy above a frequency f can be reconstructed perfectly from samples taken at any rate above 2f. The reconstruction is exact, not an approximation, and not the stair-stepped waveform you see drawn in tutorials. Below Nyquist, the original signal comes back whole.
Run the numbers. 44.1kHz captures everything up to 22.05kHz. 48kHz reaches 24kHz. 96kHz reaches 48kHz, and 192kHz reaches 96kHz. Human hearing tops out around 20kHz, and even that bound is generous. The threshold of hearing climbs so steeply near 20kHz that a tone up there has to be painfully loud before you register it at all.
So 44.1 and 48kHz already cover the entire audible range, with margin left over for the anti-aliasing filter to roll off. Moving to 96 or 192kHz adds no resolution inside the audible band. It only extends the recorded bandwidth into the ultrasonic, into frequencies you cannot hear. The "more samples means a smoother wave" intuition is where people trip up. There is no smoothing to do. The waveform below Nyquist is already reconstructed exactly.
The blind test that settled the audible part
The cleanest evidence comes from a 2007 study by E. Brad Meyer and David Moran of the Boston Audio Society, published in the Journal of the Audio Engineering Society (Vol 55, No 9, pp 775-779). They took high-resolution SACD and DVD-Audio playback and inserted a 16-bit, 44.1kHz analog-to-digital-to-analog loop into the signal path. Listeners had to tell which was which: the full hi-res chain, or the same music squeezed through a CD-quality bottleneck.
Across 554 trials, run over a year on multiple systems with many listeners, the score was 49.82 percent correct. That is statistically indistinguishable from flipping a coin. Not one listener reliably identified the CD-quality stage on any system at normal listening levels. Whatever a higher sample rate is doing, the ear sitting in the room could not detect its absence.
The part that runs backwards
Here is the twist that makes this more than a debunking. Carrying ultrasonic content can degrade the sound you do hear.
Real amplifiers and speakers are nonlinear, and their distortion tends to rise toward the frequency extremes. When ultrasonic energy, the content a 96 or 192kHz file now carries, hits a nonlinearity, intermodulation produces sum and difference frequencies. A 25kHz and a 30kHz component can beat together and drop a 5kHz product right into the middle of the audible band. The result is an uncontrolled spray of distortion smeared across the spectrum, generated by content you could never hear on its own. More bandwidth, in this case, means more audible garbage rather than less.
The important qualifier is that this is system-dependent, not a law. Meyer and Moran's own gear produced no audible intermodulation, part of why their result came out at chance. David Griesinger reproduced the effect and found that his power amplifier introduced audible intermodulation from ultrasonics while his loudspeakers did not. So the risk is real on some chains and absent on others. The takeaway is not "high rates always sound worse," it is "high rates can only hurt here, never help, so the expected value is negative."
Your converter already did the smart thing
A common defense of high rates is that they let the converter use gentler, cleaner anti-aliasing filters. That benefit is real, but you already have it for free. Nearly every modern audio interface uses a delta-sigma converter that oversamples internally at multi-megahertz rates, filters in the digital domain where filters are cheap and phase-linear, then decimates down to whatever rate your project runs at. The clean filtering people credit to "96kHz" is happening inside the box at 44.1 or 48kHz output regardless of your session setting.
Where higher rates genuinely earn their cost
This is not an argument that high sample rates are useless. Two specific cases make them pay off.
The first is aggressive sound design. If you record a source intending to pitch it down an octave or time-stretch it heavily, a 96kHz capture gives you real high-frequency content to work with after the transform, instead of empty air above 10kHz once you have dragged everything down. Field recordists who plan to mangle their material track high for exactly this reason.
The second is nonlinear processing. Saturators, clippers, and limiters generate harmonics above Nyquist, and those harmonics fold back into the audible band as aliasing. An older plugin with no internal oversampling will alias less if the whole project runs at 96kHz, because Nyquist is now further away. The catch is that most current saturation and limiting plugins solve this themselves with a built-in oversampling option, usually switchable between 2x, 4x, and 8x. That gives you the anti-aliasing benefit on precisely the plugins that need it, without doubling the cost of the entire session.
The cost, and the better lever
Cost is the other half of the trade. Compared to 48kHz, a 96kHz project roughly doubles CPU load and file size. 192kHz roughly quadruples it. That means fewer tracks and fewer plugins before the system starts to struggle, in exchange for a difference no blind test has been able to find.
If you want a digital-audio setting that genuinely improves your recordings, it is the other one: bit depth. Recording at 24-bit gives you real headroom and a lower noise floor while tracking and mixing, which matters every time you set a level. 16-bit offers about 96dB of theoretical dynamic range, stretched to roughly 120dB with shaped dither, but 24-bit removes the need to think about it during production. Record 24-bit. The sample-rate question is the one that gets argued; the bit-depth answer is the one that actually helps.
A test you can run yourself
Mastering engineer Ian Shepherd suggests a simple check. Take a finished mix, apply a phase-linear high-pass filter at around 25kHz, and listen. If nothing changes, the ultrasonic content your high sample rate captured is doing nothing for you, and on a nonlinear monitoring chain it may be working quietly against you.
For most music recording and mixing, 24-bit at 44.1kHz (music) or 48kHz (video, film, games) is the rational default. It is transparent, it is efficient, and it sidesteps the intermodulation risk entirely. Reach for a higher rate when you have a concrete reason, such as extreme pitch work or a chain of plugins that cannot oversample on their own. Otherwise the higher number on the screen is buying CPU load and disk space, not sound.