Exceptions allow you to customize the pronunciation of your speech synthesizer, to correct mispronunciations - especially names, to adjust the pacing of your speech synthesizer's speech for specific phrases, and even make it incapable of saying certain words, such as cursing.
Each exception is made up of two parts, the Word that is pronounced incorrectly, and the Exception or phonetic spelling that the speech synthesizer pronounces correctly. For example, on one early speech synthesizer, the word "are" was pronounced like "air", so it needed a pronunciation exception of "r" to make it pronounce "are" correctly. Typically getting a good pronunciation for the exception is a matter of trial and error - try a spelling variation, have the synthesizer say it, try another variation and so on until an optimal variation is found.
Adjusting Synthesizer Pacing
On some synthesizers the pacing of the voice is very mechanical "I - WANT - TO - GO - TO - THE - STORE". Exceptions can be entered that slur words together to make them more natural sounding, so you could enter exceptions for "I want to" like "I wantoo" and "go to" like "gotoo" and "the store" like "thuhstore" to get "I want to - go to - the store" which sounds nicer. For this technique to be successful, a large number of exceptions need to be entered that cover the most common phrases used by the user.
You can prevent your speech synthesizer from saying certain words by entering the words as the Word, and nothing or "fudge" as the exception. It is a good idea if you are doing this to also do the same thing for any spelling variations you can think of.