There are many, probably the majority of published papers say the opposite with respect to IMD, it’s audible, and it’s significant in audio perception.
Edit: All sorts of modulation distortion like IMD or Doppler are nonlinear, before someone else brings them up as "other types".
- Nonlinear distortion: Added spectral contents to the signal (extras tones). Nonlinear distortions are hard to predict and cancel. Not very audible.
It appears to be one of the areas associated with low speaker design and measurements, there doesn’t appear to be a consensus among the science-audio engendering communities. I have read many papers from AES and Joirnal of Acoustics that have concluded on opposite ends of the debate. Here is a quick clip of what Klippel says about it (along with several papers published by AES):
Loudspeakers and other electro-acoustical transducers generate significant intermodulation distortions in the audio band which have a significant impact on the perceived sound quality. The inductance nonlinearity L(x) produces low harmonic distortions at low frequencies but generates wide-band intermodulation which rise to significant values (30 %) at higher frequencies. Intermodulation distortion can easily be detected by using a two-tone signal with varying frequencies of the excitation tones (“bass tone sweep” or “voice tone sweep”).