Yet, human ear (or should I say brain) is much more attuned to intermodulation distortion and can spot this at much lower levels.The frequency bandwidth of the distortion also matters quite a bit, as does the placement of the distortion along the frequency range.
But all this complicates measurements, in fact making them really difficult to do for headphones.
But if one wanted to look at (or to measure) distortion that matters, one would go for IMD, not THD. Actually specific type of IMD at low volumes, low-masking probability, high non-linearity.
EDIT: In fact, IMD is one of the crucial masking factors for complex passages loosing definition and becoming muddy, jumbled or hard to separate the sounds from each other.
Edited by halcyon - 8/22/11 at 7:53am