It's tempting to think of testing with screen-readers as being like testing with browsers. With browser testing, you're checking to see how a particular piece of software deals with the code you're throwing at it. A screen reader is a piece of software too, so it makes sense to approach it the same way, right?
I don't think so. I think it's really important that if someone is going to test your site with a screen reader, it should be someone who uses a screen reader every day.
Think of it this way: you wouldn't want a designer or developer to do usability testing by testing the design or code on themselves. That wouldn't give you any useful data. They're already familiar with what problems the design is supposed to be solving, and how the interface works. That's why you need to do usability testing with someone from outside, someone who wasn't involved in the design or development process.
It's no different when it comes to users of assistive technology. You're not trying to test their technology; you're trying to test how well the thing you're building works for the person using the technology.
Don't think of screen-reader testing as a form of browser testing; think of it as a form of usability testing.
This was originally posted on my own site.