[Les]

I might mess with this some. Seems interesting. As an acoustics type, I would consider the following factors:

1) using an unbaffled speaker
This should work fine, as it will function as a pressure gradient microphone, like a ribbon. It will have a figure 8 pattern and major proximity effect.

2) size
Smaller should be fine as long as the resonant frequency is low enough. Response will rapidly drop off below Fs. A smaller cone will likely resond better to high freqiencies as well.

3) Damping
An unbaffled unloaded speaker may have a high q resonance, resulting in a very peaky response as a microphone. So the damping at resonance might be more critical. an easy way to control this could be with a variable shunt resistance.

4) Output
Should be huge. A divider prob would be good to get it down to typical mic level. Part of that divider would also be the shunt resistance mentioned above in 3).

So, I think I would pick a smallish speaker...perhaps 6 inches or so, with an Fs at or below 40 Hz with a fairly low Qes, less than .3.

Running unbaffled, I would use a shunt resistance such that total Q is 0.7-0.9. The shunt resistance would also be part of a resistive divider network feeding a balanced microphone input.

Just some thoughts...