Plenty of people have said that dither is audible and important. But whenever I ask for an example showing dither as being audible, they never have it and instead make excuses.
Round and round and round and round.. ugh.
There is a HUGE DIGFFERENCE between asking if people can hear a difference between dithered and truncated and asking them to ID which is which. And when you through a bunch of camoflague in in the form of irrelevant files such a low-word length files and MP3s, it just clouds the issue when you are asking them to ID which is which.
Talk about back-pedaling! So now you're saying that dither adds a quality and a color effect, yes?
No no no NO NO NO! Geez, Ethan, you either are just not listening, or you just don't GET IT.
Please read the following paragraph carefully.
Dither IS NOT NOISE in the classical analog way of thinking. The application of dither does NOT ADD SOUND to the signal. Dither only adds "noise" in the definition of digital information theory; i.e. it removes information - meaning it removes regular patterns in the signal (such as truncation clipping and the resulting harmonic distortion.) In this way, it actually REMOVES sound from the signal, not add to it. How much sound there actually IS TO BE REMOVED and how well of a job the dithering will do in removing it is entrely dependant upon the combination of the nature of original content and the type of dither used to combat it. Some signals have less audible or measurable harmonic distortion caused by truncation than others, and some types of harmonic distortion are more strongly smoothed by some types of dither than others. This is why "dither" cannot be just lumped together as a single entity.
An overly-simplistic analogy might be if you had a 60Hz hum and you wanted to try and attack it with EQ. If you only have a 5-band EQ, say, with the closest bands centering around 40 and 100 Hz, it's not going to have much of an effect on the hum. To conclude from that instance that EQ is useless for reducing hum or other noises is just ridiculous. And to conclude that because the 40 and 100 Hz EQ cuts impart no particular color of their own (it's a good quality EQ
) that they are uselessly inaudible devices is also ridiculous. Finally to use such examples as representative of "EQ" in general - i.e. to lump every kind of equalization together as a single entity, pretending that a wide Q bandpass, a high shelf, a notch and a cascading harmonic filter are all the same thing because they are all "EQ" is just plain wrong.
Okay, so let's see your example. As I said earlier, forget the blind test and just post two otherwise identical files where one is dithered and the other is not.
You've already done that Ethan, and plenty of people have responded and indicated that they can hear a difference. You're just throwing them the curve ball of asking them to identify *which is which*.
Again, the effects of dither are not always audible by everyobdy in all situations, but they are in some by some. Here's another analogy that may help. I's like trying to find a red box sitting on a blue table in a low light situation where only the rods and not the cones in our eyes are working and we can only see things in very low contrast shades of black and gray. Some people will be able to see the box on the table by noticing a very subtle difference in shading betwee the box and the table. Others won't see the box; there won't be the contrast enough for them to see anything but a dim gray table surface. What you are asking those who claim to see the difference between the box and the table which one is red and which one is blue, and when they can't tell which is which for sure, you conclude that they aren't really seeing the difference after all.
But the only way this will die is for someone to post a pair of files where all in attendance can listen and agree they sound different.
Why does an effect have to be audible and identifiable by everybody for it to be valid? My mother can often not hear a difference between a 24bit WAV file and 33.3RPM vinyl. The difference between a 2" tape and a 1/4" tape would be absolutely lost on her. Your average Joe Citizen thinks 128bit MP3 sounds just fine, (and frankly for casual listening - I have news for you - so do I.) Does that mean that anything better sounding than a 1/4" tape or a vinyl album or an MP3 is irrelevant or even mythical?
If it's all or nothing, then anything above 1950s quality of reproduction technology is a waste of time because there are some people out there that can't tell the difference. I think you'll find a few besides myself who will find that a ridiculous argument.
If it's not all or nothing, Ethan, then where do you draw the line? What percentage is the make or break point of what makes something worthwhile? Does it have to even be a majority? If only 49% of the people could appreciate a difference, does that mean it's not worthwhile because 51% don't? Even if it's only one out of a hundred, or even one out of a thousand, if it costs virtually nothing in time or money to to go all the way and make it better for that one in a thousand, what's wrong with that?
G.
D sorry, couldn't help it 
