Currently while reading a document using speech the user is unaware of changes in attributes of text. For example if the following line:
"Here is some underlined text"
had the words "some underlined" were underlined the user must currently check each character of the line to obtain this information.

I feel orca could do with a mode which the user could enable which would read the example line as:
Here is "underline on" some underlined "underline off" text

Obviously this needs to be possible to turn on and off as needed. While I have specific times this would be very useful, I don't imagine I would want it enabled for all applications and may be not all the time. Current times I need it are within openoffice writer and firefox, but if adobe reader were to become useable I would probably use it there (question whether such information could really be got from adobe reader).

While my example has been for underline, I feel any attribute orca can detect should be on the list. I imagine italics being important for me, sometimes websites mark errors on forms with colours, etc.

Currently Braille is slightly better in this respect as the Braille display marks attributes (even though it doesn't say what it is, at least you know where to check).