Basically take Tyll's methodology and instead of measuring once during the measurement interval, make multiple measurements and average the result... this would get rid the margin of error inherent in just making one measurement per measurement interval. Also, make absolutely sure the headphone is not moved in any way during the experiment, and that the experiment is replicated with the same make/model of headphone (can be done concurrently), which would account for flukes and provide insight to if this change could be generalized to more than just one individual pair of phones.
With that I believe one could provide a conclusive result.
Edited by Dr. Strangelove - 8/31/11 at 10:37pm
