Heh. I admit I perpetrated the fanciful recognition possibilities ("empeg! make me a ham sandwich, and a coke"). Typically the upper set limit (200 phrase-words) is due to memory and processor speed. memory and processor speed affects the quality of the stored "voice prints" and the quality at which the software can discriminate. After a certain point, any new words you add will begin to sound like another word in the list and as you add more similar sounding words you begin to degrade the recognition performance. You can get around that by increasing the resolution (memory & cpu) but there will always be that upper limit kinda "there." At least on this type of system.

The OnStar system will soon be incorporating General Magic's very excellent voice recognition system. Rob, if you're reading this, it might be a good idea to take a peek at http://www.generalmagic.com and take a very close look at the Portico product. General Magic's recognition system is suitable in noisy environments to an extent, and is very close to star trek as much as you can get. You can say stuff like "portico! do a web search for all documents made last year by mr. smith, and fax me the top 15 results to my usual phone number. also, route all my office calls to my cell phone, and reschedule my afternoon appointments to next wednesday."

Of course this takes a huge amount of cpu power, and OnStar's system gets around that by having a cellular link, that when you hit the onstar button, it will connect to a mainframe, and the processing is done offsite, but will appear to be instantaneous on the client end. ;) This functionality is not yet in OnStar I think, but it will be!

Calvin