I've done the test a few times and I tend to get around 4/6.  My primary beef is that there is no "reference" sample so I have no idea how a lot of this stuff was supposed to sound.  I can absolutely ABX a difference inbewteen all three samples however because I have no frame of reference on how it originally sounded I have to try and listen for obvious encoder issues like ambiance/attack/decay, tonality etc...  Even worse, I suspect that the encoder is adding some gain in a few of the samples; The classical one in particular I feel the 128Kbps mp3 sounds louder (and thus better) then the wav. 

Finally, we have the music itself which is half heavily compressed/occasionally distorted pop music.  Even the classical selection is pretty poor (for ABX use) as the recording (to me) sounds rolled off/distant and not super dynamic.  Still, I feel it's fair since the music is representative of what the general public is listening to....  unfortunately.

Anyhow, bottom line is that modern encoders at 320Kbps are pretty darn good.  Not transparent, but definitely good for portable devices where storage space/quality trade off makes sense.

Ironically I did better (5/6) at work with hifiman RE600 out of realtek laptop headphone jack vs home with LCD-3F though Yggdrasil/Ragnarok.

I think the RE600 are better then LCD-3F in this type of test due to the FR difference and isolation.  I can hear ambiance/decay a little better.  For fun I should try again at home with the RE600.

What I found really interesting was that in most of the samples I eliminated the 320Kbps almost right away and A/Bd between the 128Kbps mp3 vs the WAV with the WAV winning due to air/ambiance/tonality.

This is also what I found with the LCD-3F's, and exactly why I am ditching mine for HD-800's. This is all part of that whole resolution thing, and a huge deal breaker for me.

I also used the same queues to try and hear the differences, but I found the music to be quite compressed. The first three were the easiest for me, but I only managed 4/6. I was surprised to pick up on differences between 320 and lossless. I'm mostly a believer that nobody can always hear the difference, but of course I stick with FLAC whenever possible.

I'm struggling between 320 and lossless here. For some reason vices on 320 seems to have a little bit more "clear" or may be it's somehow emphasised. And I can never get the Neil Young track and as for the Jay-Z track, they all seem to be different kinds of bad to choose from. Will try with a GO 1000 + HD800 later and see if I get any better.

I tried this a couple of weeks ago. I got better as I tried it more. What is odd is I have a friend with an external HD. He has let me borrow it several times. It has anything from 128 to WAV. The 128kbps files stand out like a sore thumb. 256 not as much. But still not hard to tell. But this NPR "test" is weird. So you might be onto something. I kept picking the 128 file on the last song every time for many trials. The modern highly compressed songs where difficult at first too.  I suspect something with the loudness levels of the rip. Perhaps NPR had a preconceived idea of the outcome? Maybe they compressed the 128 even more.

I tried this a couple of weeks ago. I got better as I tried it more. What is odd is I have a friend with an external HD. He has let me borrow it several times. It has anything from 128 to WAV. The 128kbps files stand out like a sore thumb. 256 not as much. But still not hard to tell. But this NPR "test" is weird. <snip>

He probably has a lot of stuff created using old encoders.  MP3 has been around since ~1994 and the original encoders are absolutely awful in comparison to today.

IMO the current top AAC/MP3 encoders (Apple AAC/LAME etc...) have only gotten half decent at 128Kbps over the past few years.