In the course of developing the final version of this project, I looked into producing the sounds using IC sound chips instead of the Macintosh. While this solution didn't work out in the end, I gathered a good deal of information and a few half-finished C programs along the way. Details follow.
Thanks immensely to Geoff Smith for introducing me to the PIC chip and the ISD sound chips. Without the material I learned in his Small Scale Interaction course and in conversation with him, this wouldn't have been possible.
The sound chips I experimented with were from Information Storage Devices, Inc. They make a variety of sound chips varying in storage length and sound quality. None of the ones I've worked with are better than the quality of the average telephone answering machine, but for many applications, that's just fine. The most useful resource ISD has is their data book. Call and ask them for a copy; they may still be supplying it free, and it's an invaluable resource if you plan to use their chips. It seems like a daunting book to the beginner, but it's a good reference, with examples and detailed descriptions of their product line.
I was working with the ISD 33000 and the ISD2560. Both have approximately the same sound quality, but different storage lengths. the big difference between them is the interface. The 33000 uses a synchronous serial protocol (SPI), whereas the 2500 series are basically a parallel interface. While the latter can be controlled without a microchip, the former are easier to program for if you are using a chip (my opinion only). In particular, the 33000 series are easier to program to playback from a random address.
Both of these ISD chips set one pin on the chip high to signal the end of a message. This is the main way for a programmer to know when messages are over; you have to tie the end-of-message (EOM) pin to an interrupt pin on your microcontroller Because of this, the BASIC stamp is not an appropriate controller, since it does not have an external interrupt pin that's fast enough to sense the EOM pulse. The BX-24 and BX-01 chips from NetMedia could presumably work, as they have hardware interrupts, but I have not tried them in this application.