Here's Ian Shepherd's recent podcast with a discussion on these findings and things to consider when looking at this data and/or doing your own tests. There's some interesting stuff to learn amongst all this if you like this side of things
I always find it interesting that studies like these never discuss the material used. Was it a soft orchestral piece? A big band jazz piece? A screaming heavy metal tune? Often exposed solo tracks give away audible clues easier than mixed material, but they seem to lump all audio together. And sure, with a fully mixed song mastered to get every detail at 44.1 it sounds great, but how about the single note samples of a kontakt piano? And once I start pitching those samples around and warping the output with fx and granular goodies will it hold up? I've heard a significant difference in between sample rates with even a minor amount of tweaking, but when I start pitching things down an octave or two the differences become glaring. Yes, I like to experiment more than most probably, but to have samples I can manipulate to extremes and they still hold up is a huge bonus to me. Developers like Mod Wheel that sell instruments at 96k I feel are thinking ahead. When our computers catch up and are able to handle anything, do we really think 44.1 will still be a standard? For mixed material it may survive for a while, but for samples it may eventually be like VHS tape, I can watch it but do I really want to?
The problem is that even if our computers can catch up, our ears can't. Well, not at least without bionics. That would be terribly cool. (Our brains would be next in line to catch up, but they're pretty trainable.)
The data just isn't there to support the conclusion that rendering audio (including sample content) to anything higher than about 48kHz is providing any value. Sure, if you want your signal chain to have higher resolution during editing, I can somewhat see an argument there, depending on what's being done to the audio, but there's no good reason to think that 96kHz sample material is anything but unnecessary bloat. The evidence simply isn't there. (Anecdotes of uncontrolled A/B tests aren't evidence.)
As to this meta-analysis: the effect size is extremely underwhelming, and could easily be explained by not weighting the constituent studies properly according to the biases that were identified. The deal breaker for me is that studies that didn't control for IMD weren't summarily rejected but were instead weighted. That alone could account for the small effect size.
The advantages of recording and producing high-resolution audio, i.e. methods and formats capable of rendering audio beyond CD quality, are heavily debated. To examine the capacity of test subjects to notice a difference between high resolution and conventional, 44.1 or 48 kHz, 16-bit audio, we conducted a systematic review and meta-analysis. Potential study biases, the impact of test technique, experimental design, and stimulus selection were all looked into. The results indicated that test respondents had a small but statistically significant capacity to detect high-resolution material and that this impact grew substantially when they got prolonged training. The general conclusion is that operating above typical levels can influence the perceived quality of an audio recording and playback chain. Pubrica offers you complete publishing support across a variety of publications, journals, and books. You can now morph your concepts into incisive reports with our array of writing services: regulatory writing, Clinical Report Forms (CRF), biostatistics, manuscripts, business writing, physician reports, medical writing and more.
