So theres essentially 4 types of microphones, piezo transducers, electret, electrostatic and MEMs.

Piezo Transducers.

     These are generally flat discs that uses a piezoelectric effect, essentially when it is distorted by sound waves, it generates a small amount of electricity.  The disadvantage with these is they are generally only used for a narrow frequency range, eg around 40kHz.

Electret Elements.

    These are the generic microphone element, used in everything from camera microphones to toys.  Cheap and readily available, they are rated for around 20 to 20kHz, but some do respond well to up to 100KHz.  A popular one for bat detection is the Panasonic WM-61A, 

http://www.panasonic.com/industrial/components/pdf/em06_wm61_a_b_dne.pdf.

  The only problem with this is there is no hard data on the response above 20kHz.

Electrostatic Elements.

  These are essentially capacitors where the sound vibrations cause a change in the distance between the plates. They have a good range, up to 200kHz, but need a high voltage, around 200V to use.  This isn't something I want to strap to my head....

MEMs Elements.

   These are the new kids, small surface mount packages with the element etched directly onto the silicon wafer.  There are only a few that have their high frequency response documented, one such example is SPU0410LR5H from Knowles.

http://www.knowles.com/eng/Products/Sensors/Ultrasonic

This has the the response described up to 80kHz, and is also a very small package, being only around 3mm2, and only 1.1mm high.

Conclusions. 

  Well, all things considered I think I will take a 2 pronged approach to this. I will design my PCBs to use the Knowles MEMs elements, but for testing and development purposes I will buy a couple of Panasonic electret elements, this means I can develop the system whilst the PCBs are being made.