how about some Science?
I'd equate the op question with "how much information can the audio system convey"
the engineering answer to that is well established:
Shannon–Hartley theorem - Wikipedia, the free encyclopedia
if for comparison the amplitude is limited to relatively safe SPL levels and we assume audio bandwidth is also limited then the noise and distortion of the system limit the information the channel can convey
source and amplifier noise establish a limit - even with dynamic transducers
for practical headphone listening higher limiting noise includes room noise, microphonics from cup, cable motion
at higher volumes transducer nonlinear distortion ( or distortion from "no feedback", "simple" discrete, and/or tube circuits ) limits the Channel Capacity - IMD distortion products between complex multi-tone signal components (like orchestral music) looks like an increase in the noise floor
the Intermodulation Distortion problem with channels reproducing complex multitones is what drove the boys at Bell Labs to invent negative feedback amplifiers - and more recently the advent of A/DSL modems has driven the development of op amps capable of -100 dB distortion levels driving 25 Ohms up to MHz
Hearing is also complex and nonlinear the "information" we can perceive in sound is affected strongly by Masking - a low frequency sound makes other sound components of higher frequency harder to hear - badly distorting low frequencies will "wipe out" a lot of the rest of the audio spectrum perceptible "information"
For more intellectual fun with hearing/sound and information consider:
By exploiting psychoacoustic Masking phenomena good Codecs like OGG, AAC, and the rest can toss out 75% (@ ~320Kb ) of the Shannon-Hartley definition of information in Redbook CD data and vanishingly few people can hear the difference without training and using specific test sound clips