there are two tabs in the speech window, <em>speech recognition</em> is the one you are referring too, the other is <em>text to speech, </em>that's the one me thinks... It might be different if you are running leopard. I tried to simulate your problem myself, couldn't do it, I'm running tiger...