There is an implicit inference involved when one tests using pure sine waves, and then *infers* the results about audiophilic listening. It is probably best to remove this inference step, and--if one wants to make conclusions about what we perceive when we listen to music--then we should test these perceptions *with music.*
The point is not simply academic: as reported in the Journal of Neurophysiology (
http://jn.physiology.org/cgi/content/abstract/83/6/3548) and cited in lay context here
http://www.ocf.berkeley.edu/~ashon/audio/Ultrasonics.htm, measurable brain physiological response can be measured in individuals when exposed to music with a high pass filter of 22kHz. It is an impressive study, using special recording equipment, speakers with a flat response to 100kHz, baseline measurements for controls, electroencephalograms (EEG), positron emission tomography (PET), and measurements of cerebral blood flow to detect brain response. The PDF is available: well worth reading.
There is no explicit support for or against differences of 0.1dB or 0.25dB at 22kHz being discernable--I'll leave it to the reader to make his/her own inferences from what is explicitly tested.
Besides the importance of musical context, the study also found that short 15-20 second clips were not enough. Subjects listened for 200 seconds in much of the testing, and these longer listening sessions were required for a response to be noted.
This does lend some support to the old audiophile adage that you really have to sit and listen *for a while,* and then--you can't quite explain it--but there really may be a reason why you prefer A over B.