Originally Posted by aerius
Human hearing isn't linear. If it were, making a given sound 10dB louder would make it sound 10 times as loud, in actuality it roughly doubles the perceived volume.
Well, it would sound about 3x louder if human hearing was linear, since we are referring to dB in perspective of amplitude(log20).
|I've done a lot of testing which involved matching volume levels before starting the test, once the difference is less than 3dB or so it becomes very hard to tell which is louder. Matching to within 1dB by ear is damn near impossible, and requires an SPL meter.
While it may be generally difficult to match 1dB, and this will be a very subtle(at best) difference - this 1dB difference is 
demonstrated to be detectable almost all of the time in direct comparison of signals in controlled perceptual tests. Actually, significanlty lesser differences have been shown to be audible in such testing dependant on the program material(white noise being the most sensitive).
-ChrisFootnotesThe Subjective Loudness of Typical Program Material
Gilbert A. Soulodre, Michel C. Lavoie, and Scott G. Norcross
AES Preprint: 5892