If you couldn't tell them apart in such tests, how could you (or anyone) tell them apart in normal listening? That's what I'm confused about. Human auditory memory is only a few seconds, after all.
Don't get me wrong, anyway. I never compress my own files. I only listen to music at home, own a 12 TB (12,000 GB) external hard drive array, and even use uncompressed formats.
Ah, yes, I remember this listening test idea of yours. But how would it go about proving anything objectively? To me, it seems like it introduces far too many variables instead of testing only one variable.
Good question. This gets into the differences between actually
hearing a sound and
remembering a sound. These are two
very different processes.
A) Hearing -
Our entire body picks up vibrations through thousands of hair follicles, the touch of other vibrating items, and specially designed nerves inside our joints to detect a wide range of vibration wavelengths. Our chest reacts to pressure changes. All of these signals are sent by our nervous system to our brains - the auditory cortex to be exact - located near the center and wrapped around the thalamus (the brain's home for emotion and pain).
Meanwhile our outer ears collect and amplify sound waves into our middle ear, where the eardrum and ossicles bones again amplify (and perhaps attenuate) the sound and send it into the inner ear.
The inner ear is a magnificent place: the sound travels down a long spiral-shaped cochlea into the organ of corti, where thousands of microscopic hairs in triangle-wedge-shaped clusters react to the sound waves. Above these hairs hangs a suspended pool of liquid which appears to compress/limit and even expand the sound quality in real time! There is very little known about how this liquid moves, works, or where it comes from, but researchers believe it is able to localize a hair cluster and form over it, either thickening or thinning itself and effecting the sound reception of the hairs below. Basically a real-time fluid-based EQ and compressor/limiter, of which we have no electrical or digital equivalent.
At this point the brain combines the sound signals from both ears with the vibration and pressure signals from the rest of the body, works out timing differences for spatial awareness, and you are "hearing".
If you do this whole thing in headphones, you can remove the body and the outer ear and start the process from the middle ear, as the little headphone speakers inject tiny vibrations into our ear. But the rest is the same. Sound comes down the cochlea and is processed by the auditory cortex, mainlining into the emotional center of your brain.
B) Remembering Sound -
Memory of sound doesn't just come from other areas of the brain, it bypasses the auditory cortex completely. The memory of a sound, much like a taste or smell, is stored far away from the original detection mechanism. A memory does not need to be detected by the senses again and it cannot be studied in real-time. Your nose doesn't remember smells, your eyes don't remember sites, your ears don't remember sound.
It is a snapshot stored into memory as a reaction only because it does not contain vibration or sound, only the memory of your reaction to it. You cannot recall the actual sound from memory, you can recall your reaction to it, and you can use your imagination to attempt to re-construct it, but it is not sound, vibration, or data as we know it, it's human memory, something infinitely difficult to quantify.
See the differences? Two totally different processes, mentally, this hearing and remembering.
BUT you can still compare sound to memory with some accuracy. It's true. It's just not easy or obvious. Accuracy is very low which is why I hate ABX tests - they ultimately conclude that no one can hear anything of detail and we know that is false. All signs point to the test being fatally flawed. Hence ABX is only used to shoot down quality, not to improve it, and only seems to be referenced by so-called objectionists trying to show something doesn't exist. Perfect test for that, it can't show that anything does indeed exist.
How can you improve your accuracy? Become a better listener and educate yourself about sound. Play with an EQ, parametric if possible, and learn the frequencies. Learn how to visualize sound in the soundscape - where are the instruments, how far away, how big is the room they were recorded in. Learn about fundamental frequencies and harmonics. Learn your room, the acoustic properties of it, so you can better put into context what is coming out of the speakers in that room. Pay attention to delays, reverb trails, and instrument timbre. Learn the tells of low-resolution: crunchy cymbals, undefined high-hat, no air or breath in voice (you can't 'hear' when the singer smiles), lots of up-front compression and hyper-movement. Those types of things.
Even producers/mixers compare back and forth all the time. When you are just checking presets or twisting a knob to hear what you like best this is not an ABX test. It's not a test at all, it's just a choice. No one else cares what you pick ,you won't be graded on it. So you can switch through things and compare from memory. Again, no one is grading you, you are just picking something, there is no wrong answer.
Forming a test with a right and wrong answer is not the way to go about it. Our auditory cortex does not work this way, we cannot compare and contrast two sound items accurately. We have to compare the live sound item with the memory, or two memories.
If you want statistical mush (to keep status quo) you use ABX tests. If you want to actually detect a difference you need to invest more time and change the test format.
Lossy codecs do amazing things but it's still LOSS. Dress it up however you want, it's only needed if you have a bandwidth crunch.
