When does speaker distortion become audible?

Most of the speakers have overhunged motors (narrow gap, long coil) producing a lot of distortions especially at longer excursions (bass). Underhung motors (big magnet, wide gap, short coil) produce less distortions (are more linear) but are more expensive and not very common. Acoustic Zen Adagios have "underhung" woofers. Underhung is sometimes even used in tweeters to reduce distortions at high power (long excursions) - Morel Supreme 110.

http://www.gattiweb.com/images/delta_design/Supreme110.pdf

According to Robert Lee of Acoustic Zen speakers operating at power higher than 10W can get 10-12% of distortion. Most of it comes from the woofer and is less audible.
Here is fragment of interview with Robert Lee:

"That is a good point and I would like to say this. There are thousands of speaker manufacturers worldwide but so far I have identified less than ten who employ underhung drivers. I think that underhung drivers are the best solution to reduce harmonic distortion in the bass. No one ever talks about the huge amounts of THD in the low frequencies, especially when you are playing things loud. Most drivers create from 10-12% THD when you listen at over 10 watts. If you are listening to 10 watts through your speaker system, you get 5% harmonic distortion. So what are you listening to? When people design amplifiers, they rate them at 0.05% harmonic distortion. Meanwhile even at modest volume levels, your low-frequency drivers put out 5% or greater harmonic distortion so nobody can listen at even low levels and achieve true purity with low distortion, never mind concert levels.
That's why I selected to use both the underhung driver and a ribbon tweeter of my own design."

As for the flux density your assumption seems logical to me but perhaps speaker builders can chime in?

Bruce, I had previously Paradigm Studio 60v2. It had 2 1/2 way xover. One of two 6.5" woofers operated only at the low bass frequencies extending that way frequency response. My current Hyperion 938 have two 8" woofers in much larger cabinet (pretty much like Wilson Puppy) but bass extension is actually worse. What is different, though, is the sound of the bass. I'm not sure if this would come up in any distortion test but bass is more pleasant, more musical. Bass tones are even with realistic attack and decay. I've read that bass reflex speakers might be tuned for extension or for the lowest distortion. People talk of bass as low or as powerful or tuneful but there is something else making it sound real.

Bruce, the worst part of Studio/60 was aluminum dome tweeter. I did try to replace it with Morel Supreme 110 but at the end I failed - it is not as easy as it sounds. Tweeter in Hyperions HPS-938 is so well integrated that you cannot practically tell when sound goes thru it. Fantastic highs with zero sibilants.

Ptss, 10-20% might be noticeable to most but it is also related to frequency. Bass distortions are not as audible as midrange distortions. People often select an amp with 0.01% distortion not realizing that it is most likely due to deep negative feedback at the expense of the sound, while speaker distortions go in many percents.

At 20 Hz it was over 5% but according to this article:
http://www.bksv.com/doc/BO0385.pdf
it is harder to hear distortions below 400Hz. Single tone shows only THD but there are other distortions that might be more audible. Intermodulation, caused by nonlinear motion and membrane bending, produces new frequencies when speaker is driven by two frequencies.

According to this article: http://publications.lib.chalmers.se/records/fulltext/126969.pdf
THD testing does not reveal all nonlinearities of the speaker. Harmonic distortion will alter instruments' overtones adding coloration while intermodulation distortion will be much more audible.