Originally Posted by raddle
There you go, he tested preference, not accuracy. I'd like to see some work correlating measurements with accuracy.
Other tests by the Harmon group and others generally show that the more accurate setups (especially with speakers) are preferred. They're evaluating accuracy primarily on FR, where wild swings are agreed to be bad. Of course, there is more different to speakers and headphones than just the FR differences, so you can't be absolutely convinced (based on just this data, though others seem to corroborate it) that it's the FR that is making the difference. For headphones, it's based on in-ear response, which is different for the same headphone on different people because of different ear and head shapes. The exact most preferred to subjectively accurate response averaged across people is still a matter of ongoing research, but the hypothesis that headphones are perceived to be better when following an in-ear response corresponding to what you get from flat speakers in a room seems very plausible (at least with the test tracks used, though that is probably indicative), and the data seems to support that.
I think the results you might be most interested in is this:
the Listener Comments slide
It's not quite the word "accurate" but "neutral / low coloration" and then "good spectral balance" comments correlate strongly with preference ratings for the listeners tested on the tracks tested in the conditions tested etc. A lot of the other research seems to indicate that as a whole different groups don't actually have different preferences in sound. You can check the measurements of the headphones and check the author's blog for later presentation slides, particularly with respect to preferred and measured responses (and measured more accurate by FR) when tweaking effective headphone FR by EQ.